Construction zone sign detection

Abstract

Methods and systems for detection of a construction zone sign are described. A computing device, configured to control the vehicle, may be configured to receive, from an image-capture device coupled to the computing device, images of a vicinity of the road on which the vehicle is travelling. Also, the computing device may be configured to determine image portions in the images that may depict sides of the road at a predetermined height range. Further, the computing device may be configured to detect a construction zone sign in the image portions, and determine a type of the construction zone sign. Accordingly, the computing device may be configured to modify a control strategy associated with a driving behavior of the vehicle; and control the vehicle based on the modified control strategy.

Description

where X_i is the feature vector for instance i, β_k is a vector of weights corresponding to category k, and score(X_i,k) is the score associated with assigning instance i to category k.

As an example, a training computing device may be configured to receive training data for a plurality of driving situations of a given vehicle. For example, for each of the plurality of driving situations, respective training data may include respective image-based information, respective LIDAR-based information, respective RADAR-based information, respective traffic information, and respective map information. Also, the training computing device may be configured to receive positive or negative indication of existence of a respective construction zone corresponding to the respective training data for each of the driving situations. Further the training computing device may be configured to correlate, for each driving situation, the positive or negative indication with the respective training data; and determine parameters (e.g., vector of weights for equation 1) of the classifier based on the correlations for the plurality of driving situations. Further, in an example, the training computing device may be configured to determine a respective reliability metric for each source of information based on the correlation. The parameters and respective reliability metrics of the plurality of sources of information may be provided to the computing device configured to control the vehicle 402 such that as the computing device receives the information, from the plurality of sources of information, relating to the detection of the construction zone, the computing device may be configured to process the information through the classifier using the determined parameters of the classifier to determine the likelihood.

In one example, the likelihood may be qualitative such as “low,” “medium,” or “high” or may be numerical such as a number on a scale, for example. Other examples are possible.

Referring back to FIG. 3, at block 306, the method 300 includes modifying, using the computing device, a control strategy associated with a driving behavior of the vehicle, based on the likelihood.

The control system of the vehicle may support multiple control strategies and associated driving behaviors that may be predetermined or adaptive to changes in a driving environment of the vehicle. Generally, a control strategy may comprise sets of rules associated with traffic interaction in various driving contexts such as approaching a construction zone. The control strategy may comprise rules that determine a speed of the vehicle and a lane that the vehicle may travel on while taking into account safety and traffic rules and concerns (e.g., changes in road geometry due to existence of a construction zone, vehicles stopped at an intersection and windows-of-opportunity in yield situation, lane tracking, speed control, distance from other vehicles on the road, passing other vehicles, and queuing in stop-and-go traffic, and avoiding areas that may result in unsafe behavior such as oncoming-traffic lanes, etc.). For instance, in approaching a construction zone, the computing device may be configured to modify or select, based on the determined likelihood of the existence of the construction zone, a control strategy comprising rules for actions that control the vehicle speed to safely maintain a distance with other objects and select a lane that is considered safest given road changes due to the existence of the construction zone.

As an example, in FIG. 4, if the likelihood of the existence of the construction zone is high (e.g., exceeds a predetermined threshold), the computing device may be configured to utilize sensor information, received from on-board sensors on the vehicle 402 or off-board sensors in communication with the computing device, in making a navigation decision rather than preexisting map information that may not include information and changes relating to the construction zone. Also, the computing device may be configured to utilize the sensor information rather than the preexisting map information to estimate lane boundaries. For example, referring to FIG. 4, the computing device may be configured to determine locations of construction zone markers (e.g., the construction zone cone(s) 406) rather than lane markers 418 on the road 404 to estimate and follow the lane boundaries. As another example, the computing device may be configured to activate one or more sensors for detection of construction workers 420 and making the navigation decision based on the detection.

In an example, a first control strategy may comprise a default driving behavior and a second control strategy may comprise a defensive driving behavior. Characteristics of a the defensive driving behavior may comprise, for example, following a vehicle of the vehicles 414A-B, maintaining a predetermined safe distance with the vehicles 414A-B that may be larger than a distance maintained in the default driving behavior, turning-on lights, reducing a speed of the vehicle 402, and stopping the vehicle 402. In this example, the computing device of the vehicle 402 may be configured to compare the determined likelihood to a threshold likelihood, and the computing device may be configured to select the first or the second control strategy, based on the comparison. For example, if the determined likelihood is greater than the threshold likelihood, the computing device may be configured to select the second driving behavior (e.g., the defensive driving behavior). If the determined likelihood is less than the threshold likelihood, the computing device may be configured to modify the control strategy to the first control strategy (e.g., select the default driving behavior).

In yet another example, alternatively or in addition to transition between discrete control strategies (e.g., the first control strategy and the second control strategy) the computing device may be configured to select from a continuum of driving modes or states based on the determined likelihood. In still another example, the computing device may be configured to select a discrete control strategy and also may be configured to select a driving mode from a continuum of driving modes within the selected discrete control strategy. In this example, a given control strategy may comprise multiple sets of driving rules, where a set of driving rules describe actions for control of speed and direction of the vehicle 402. The computing device further may be configured to cause a smooth transition from a given set of driving rules to another set of driving rules of the multiple sets of driving rules, based on the determined likelihood. A smooth transition may indicate that the transition from the given set of rules to another may not be perceived by a passenger in the vehicle 402 as a sudden or jerky change in a speed or direction of the vehicle 402, for example.

In an example, a given control strategy may comprise a program or computer instructions that characterize actuators controlling the vehicle 402 (e.g., throttle, steering gear, brake, accelerator, or transmission shifter) based on the determined likelihood. The given control strategy may include action sets ranked by priority, and the action sets may include alternative actions that the vehicle 402 may take to accomplish a task (e.g., driving from one location to another). The alternative actions may be ranked based on the determined likelihood, for example. Also, the computing device may be configured to select an action to be performed and, optionally, modified based on the determined likelihood.

In another example, multiple control strategies (e.g., programs) may continuously propose actions to the computing device. The computing device may be configured to decide which strategy may be selected or may be configured to modify the control strategy based on a weighted set of goals (safety, speed, etc.), for example. Weights of the weighted set of goals may be a function of the determined likelihood. Based on an evaluation of the weighted set of goals, the computing device, for example, may be configured to rank the multiple control strategies and respective action sets and select or modify a given strategy and a respective action set based on the ranking.

These examples and driving situations are for illustration only. Other examples and control strategies and driving behaviors are possible as well.

Referring back to FIG. 3, at block 308, the method 300 includes controlling, using the computing device, the vehicle based on the modified control strategy. In an example, the computing device may be configured to control actuators of the vehicle using an action set or rule set associated with the modified control strategy. For instance, the computing device may be configured to adjust translational velocity, or rotational velocity, or both, of the vehicle based on the modified driving behavior.

As an example, in FIG. 4, controlling the vehicle 402 may comprise determining a desired path of the vehicle, based on the likelihood. In one example, the computing device may have determined a high likelihood that a construction zone exists on the road 404 on which the vehicle 402 is travelling. In this example, the computing device may be configured to take into account lane boundary indicated by the lane markers 418 on the road 404 as a soft constraint (i.e., the lane boundary can be violated if a safer path is determined) when determining the desired path. The computing device thus may be configured to determine a number and locations of the construction zone cone(s) 406 that may form a modified lane boundary; and may be configured to adhere to the modified lane boundary instead of the lane boundary indicated by the lane markers 418.

As shown in FIG. 4, the vehicle 402 may be approaching the construction zone on the road 404, and the computing device may be configured to control the vehicle 402 according to a defensive driving behavior to safely navigate the construction zone. For example, the computing device may be configured to reduce speed of the vehicle 402, cause the vehicle 402 to change lanes and adhere to the modified lane boundary formed by the construction zone cone(s) 406, shift to a position behind the vehicle 414A, and follow the vehicle 414A while keeping a predetermined safe distance.

In one example, in addition to determining the likelihood of the existence of the construction zone, the computing device may be configured to determine or estimate a severity of changes to the road 404 due to the existence of the construction zone. The computing device may be configured to modify the control strategy further based on the severity of the changes. As an example, in FIG. 4, the computing device may be configured to determine, based on the construction equipment 410A-B, number and locations of the construction zone cone(s) 406 and barrel(s) 408, how severe the changes (e.g., lane closure, shifts, etc.) to the road 404 are, and control the vehicle 402 in accordance with the defensive driving behavior. In another example, the construction zone may comprise less severe changes. For example, the construction zone may comprise a worker that may be painting on a curb lane on a side of the road 404. In this example, changes to the road 404 may be less severe than changes depicted in FIG. 4, and the computing device may be configured to reduce the speed of the vehicle 402 as opposed to stop the vehicle 402 or cause the vehicle 402 to change lanes, for example.

These control actions and driving situations are for illustration only. Other actions and situations are possible as well. In one example, the computing device may be configured to control the vehicle based on the modified control strategy as an interim control until a human driver can take control of the vehicle.

As described above with respect to FIGS. 3 and 4, the computing device may be configured to determine the likelihood of the existence of the construction zone based on identification or detection of a construction zone sign (e.g., the construction zone sign 412A) that may be indicative of the construction zone.

FIG. 5 is a flow chart of a method 500 for detection of a construction zone sign, in accordance with an example embodiment. FIGS. 6A-6B illustrate images of a road and vicinity of the road the vehicle is travelling on, in accordance with an example embodiment, and FIGS. 6C-6D illustrate portions of the images of the road and the vicinity of the road depicting sides of the road at a predetermined height range, in accordance with an example embodiment. FIGS. 5 and 6A-6D will be described together.

The method 500 may include one or more operations, functions, or actions as illustrated by one or more of blocks 502-512. Although the blocks are illustrated in a sequential order, these blocks may in some instances be performed in parallel, and/or in a different order than those described herein. Also, the various blocks may be combined into fewer blocks, divided into additional blocks, and/or removed based upon the desired implementation.

At block 502, the method 500 includes receiving, at a computing device configured to control a vehicle, from an image-capture device coupled to the computing device, one or more images of a vicinity of a road on which the vehicle is travelling. The computing device may be onboard the vehicle or may be off-board but in wireless communication with the vehicle, for example. Also, the computing device may be configured to control the vehicle in an autonomous or semi-autonomous operation mode. Further, an image-capture device (e.g., the camera 134 in FIG. 1 or the camera 210 in FIG. 2) may be coupled to the vehicle and in communication with the computing device. The image-capture device may be configured to capture images or video of the road and vicinity of the road on which the vehicle is travelling on.

FIGS. 6A-6B, for example, illustrate example images 602 and 604, respectively, captured by the image-capture device coupled to the vehicle 402 in FIG. 4. In an example, the image-capture device may be configured to continuously capture still images or a video from which the still images can be extracted. In one example, one or more image-capture devices may be coupled to the vehicle 402; the one or more image-capture devices may be configured to capture the images from multiple views to take into account surroundings of the vehicle 402 and road condition from all directions.

Referring back to FIG. 5, at block 504, the method 500 includes determining, using the computing device, one or more image portions in the one or more images, and the one or more image portions may depict sides of the road at a predetermined height range. In some examples, the predetermined height range may correspond to a height range that is typically used for construction zone signs. In many jurisdictions, construction zones on roads are regulated by standard specifications and rules, which may be used to define the predetermined height range. An example rule may state that a construction zone sign indicating existence of a construction zone on the road may be placed at a given location continuously for longer than three days and may be mounted on a post on a side of the road. Further, another rule may specify that a minimum sign mounting height for a temporary warning construction zone sign, for example, may be 1 foot above road ground level. In other examples, in addition to or alternative to the minimum sign mounting height, a height range can be specified, i.e., a height range for a temporary warning construction zone sign may be between 1 foot and 6 feet, for example. In some locations where the construction zone sign may be located behind a traffic control device such as a traffic safety drum or temporary barrier, the minimum height may be raised to 5 feet in order to provide additional visibility. Additionally or alternatively, a height range can be specified to be between 5 feet and 11 feet, for example. These numbers and rules are for illustration only. Other standards and rules are possible. In some examples, the predetermined height range or minimum height of a typical construction zone sign may be dependent on location (e.g., geographic region, which state in the United States of America, country, etc.).

The computing device may be configured to determine portions, in the images captured by the image-capture device, which may depict road sides at the predetermined height range of a typical construction zone sign according to the standard specifications. As an example, in FIG. 6A, the computing device may be configured to determine a portion 606 in the image 602 depicting a side of the road 404 at a predetermined height range specified for typical construction zone signs according to the standard specifications. Similarly, in FIG. 6B, the computing device may be configured to determine a portion 608 in the image 604 depicting another side of the road 404 at the predetermined height range. FIG. 6C illustrates the image portion 606 of the image 602 illustrated in FIG. 6A, and FIG. 6D illustrates the image portion 608 of the image 604 illustrated in FIG. 6B.

Referring back to FIG. 5, at block 506, the method 500 includes detecting, using the computing device, a construction zone sign in the one or more image portions. The standard specifications also may include rules for shape, color, pattern, and retroreflective characteristics of typical construction zone signs. As an example for illustration, the standard specifications may specify that a typical construction zone sign may be a 48 inches×48 inches diamond shape with black letters of symbols on an orange background having a standard type of reflective sheeting. These specifications are for illustration only, and other specifications are possible.

As an example, referring to FIGS. 6A-6D, the computing device may be configured to detect candidate construction zone signs, such as the sign 412A and the sign 416B in the image portions 608 and 606, respectively. The computing device further may be configured to determine, using image recognition techniques known in the art for example, whether a candidate construction zone sign relates to a construction zone, based on one or more of the shape, color, pattern, and reflective characteristics of the candidate construction zone sign as compared to the standard specifications of typical construction zone signs. For example, the computing device may be configured, based on the comparison, to determine that the sign 412A is a construction zone sign, while the sign 416B is not.

In an example to illustrate use of image recognition, the computing device may be configured to compare an object detected, in the one or more image portions, to a template of the typical construction zone sign. For example, the computing device may be configured to identify features of the object such as color, shape, edges, and corners of the object in the one or more image portions. Then, the computing device may be configured to compare these features to orange/yellow color, diamond shape with sharp edges, and corners (i.e., “corner signature”) of the typical construction zone sign. The computing device may be configured to process the features (e.g., color, shape, etc.) or parameters representative of the features of the object through a classifier to determine whether the features of the object match typical features of the typical construction zone sign. The classifier can map input information (e.g., the features of the object) to a class (e.g., the object represents a construction zone sign). Examples of classifiers, training data, and classification algorithms are described above with regard to block 304 of the method 300 illustrated in FIG. 3.

In an example, the computing device may be configured to use information received from other sensors or units coupled to the vehicle 402, in addition to image-based information received from the image-capture device, to confirm or validate detection of a construction zone sign. For example, the computing device may be configured to assign or determine, based on the image-based information, a first likelihood that a candidate construction zone sign in the image portions relates to a construction zone. Further, the computing device may be configured to receive, from a LIDAR sensor (e.g., the LIDAR unit 132 in FIG. 1) coupled to the vehicle 402 and in communication with the computing device, LIDAR-based information that includes a 3D point cloud corresponding to the image portions (e.g., the image portion 608) depicting the candidate construction zone sign (e.g., the sign 412A). The 3D point cloud may comprise a set of points based on light emitted from the LIDAR and reflected from a surface of the candidate construction zone sign. The computing device may be configured to determine a second likelihood that the candidate construction zone sign relates to the construction zone, based on the LIDAR-based information, and confirm existence or detection of the construction zone sign based on the first likelihood and the second likelihood.

In another example, in addition to or alternative to receiving the LIDAR-based information, the computing device may be configured to receive, from a RADAR sensor (e.g., the RADAR unit 130 in FIG. 1) coupled to the computing device, RADAR-based information relating to location and characteristics of the candidate construction zone sign. The RADAR sensor may be configured to emit radio waves and receive back the emitted radio waves that bounced off the surface of the candidate construction zone sign. The received signals or RADAR-based information may be indicative, for example, of dimensional characteristics of the candidate construction zone sign, and may indicate that the candidate construction zone sign is stationary. The computing device may be configured to determine a third likelihood that the candidate construction zone sign relates to the construction zone, based on the RADAR-based information, e.g., based on a comparison of the characteristics of the candidate construction zone sign to standard characteristics of a typical construction zone sign. Further, the computing device may be configured to detect the construction zone sign based on the first likelihood, the second likelihood, and the third likelihood.

As an example, the computing device may be configured to determine an overall likelihood that is a function of the first likelihood, the second likelihood, and the third likelihood (e.g., a weighted combination of the first likelihood, the second likelihood, and the third likelihood), and the computing device may be configured to detect the construction zone sign based on the overall likelihood.

In one example, the computing device may be configured to detect the construction zone sign based on information received from multiple sources such as the image-capture device, the LIDAR sensor, and the RADAR sensor; but, in another example, the computing device may be configured to detect the construction zone sign based on a subset of information received from a subset of the multiple sources. For example, images captured by the image-capture device may be blurred due to a malfunction of the image-capture device. As another example, details of the road 404 may be obscured in the images because of fog. In these examples, the computing device may be configured to detect the construction zone sign based on information received from the LIDAR and/or RADAR units and may be configured to disregard the information received from the image-capture device.

In another example, the vehicle 402 may be travelling in a portion of the road 404 where some electric noise or jamming signals may exist, and thus the LIDAR and/or RADAR signals may not operate correctly. In this case, the computing device may be configured to detect the construction zone sign based on information received from the image-capture device, and may be configured to disregard the information received from the LIDAR and/or RADAR units.

In one example, the computing device may be configured to rank the plurality of sources of information based on a condition of the road 404 (e.g., fog, electronic jamming, etc.) and/or based on the respective reliability metric assigned to each source of the plurality of sources. The ranking may be indicative of which sensor(s) to rely on or give more weight to in detecting the construction zone sign. As an example, if fog is present in a portion of the road, then the LIDAR and RADAR sensors may be ranked higher than the image-based device, and information received from the LIDAR and/or RADAR sensor may be given more weight than respective information received from the image-capture device.

Referring back to FIG. 5, at block 508, the method 500 includes determining, using the computing device, a type of the construction zone sign in the one or more image portions. Various types of construction zone signs may exist. One construction zone sign type may be related to regulating speed limits when approaching and passing through a construction zone on a road. Another construction zone sign type may be related to lane changes, closure, reduction, merger, etc. Still another construction zone sign type may be related to temporary changes to direction of travel on the road. Example types of construction zone signs may include: “Right Lane Closed Ahead,” “Road Work Ahead,” “Be Prepared to Stop,” “Road Construction 1500 ft,” “One Lane Road Ahead,” “Reduced Speed Limit 30,” “Shoulder Work,” etc. Other example types are possible.

In an example, the computing device of the vehicle may be configured to determine a type of the detected construction zone based on shape, color, typeface of words, etc. of the construction zone sign. As an example, the computing device may be configured to use image recognition techniques to identify the type (e.g., shape of, or words written on, the construction zone sign) from an image of the detected construction zone sign.

As described above with respect to block 506, the computing device may be configured to utilize image recognition techniques to compare an object to a template of a typical construction zone sign to detect a construction zone sign. In an example, to determine the type of the detected construction zone sign, the computing device may be configured to compare portions of the detected construction zone sign to sub-templates of typical construction zone signs. In one example, the computing device may be configured to identify individual words or characters typed on the detected construction zone sign, and compare the identified words or characters to corresponding sub-templates of typical construction zone signs. In another example, the computing device may be configured to determine spacing between the characters or the words, and/or spacing between the words and edges of the detected construction zone sign. In still another example, the computing device may be configured to identify a font in which the words or characters are printed on the detected construction zone sign and compare the identified font to fonts sub-template associated with typical construction zone signs.

As an example, the computing device may be configured to detect a construction zone sign having the words “Road Work Ahead” typed on the detected construction zone sign. The computing device may be configured to extract individual characters or words “Road,” “Work,” “Ahead,” in the one or more image portions, and compare these words and characteristics of these words (e.g., fonts, letter sizes, etc.) to corresponding sub-templates typical construction zone signs. Also, the computing device may be configured to compare spacing between the three words and spacing between letters forming the words to corresponding sub-templates. Further, the computing device may be configured to compare spacing between the word “Road” and a left edge of the detected construction zone sign and the spacing between the word “Ahead” and a right edge of the detected construction zone sign to corresponding sub-templates. Based on these comparisons, the computing device may be configured to determine the type of the detected construction zone sign.

These features (e.g., characters, words, fonts, spacing, etc.) are examples for illustrations, and other features can be used and compared to typical features (i.e., sub-templates) of typical construction zone signs to determine the type of the detected construction zone sign.

In one example, in addition to or alternative to using image recognition, based on the RADAR-based information, the computing device may be configured to determine shape and dimensions of the construction zone sign and infer the type and associated road changes from the determined shape and dimensions. Other examples are possible.

At block 510, the method 500 includes modifying, using the computing device, a control strategy associated with a driving behavior of the vehicle, based on the type of the construction zone sign. The road changes due to existence of a construction zone on the road may be indicated by the type of the construction zone sign existing on ahead of the construction zone. The computing device may be configured to modify control strategy of the vehicle based on the determined type of the construction zone sign.

Examples of modifying the control strategy are described above with regard to block 306 of the method 300 illustrated in FIG. 3. As examples, the computing device may be configured to determine whether a lane shift and/or speed change are required as indicated by the type; utilize sensor information received from on-board or off-board sensors in making a navigation decision rather than preexisting map information; utilize the sensor information to estimate lane boundaries rather than the preexisting map information; determine locations of construction zone cones or barrels rather than lane markers on the road to estimate and follow the lane boundaries; and activate one or more sensors for detection of construction workers and making the navigation decision based on the detection. These examples and driving situations are for illustration only. Other examples and control strategies and driving behaviors are possible as well.

At block 512, the method 500 includes controlling, using the computing device, the vehicle based on the modified control strategy. Examples of controlling the vehicle based on the modified control strategy are described above with regard to block 308 of the method 300 illustrated in FIG. 3. As examples, the computing device may be configured to adjust translational velocity, or rotational velocity, or both, of the vehicle based on the modified driving behavior in order to follow another vehicle; maintain a predetermined safe distance with other vehicles; turn-on lights; reduce a speed of the vehicle; shift lanes; and stop the vehicle. These control actions and driving situations are for illustration only. Other actions and situations are possible as well.

As described with respect to block 506 of the method 500 illustrated in FIG. 5, the computing device may be configured to detect or confirm detection of the construction zone sign based on information received from a LIDAR sensor coupled to the vehicle and in communication with the computing device.

FIG. 7 is a flow chart of a method 700 for detection of the construction zone sign using LIDAR-based information, in accordance with an example embodiment. FIG. 8A illustrates LIDAR-based detection of the construction zone sign at a height greater than a threshold height from a surface of the road, in accordance with an example embodiment. FIG. 8B illustrates a LIDAR-based image depicting the area at the height greater than the threshold height from the surface of the road, in accordance with an example embodiment. FIGS. 7 and 8A-8B will be described together.

The method 700 may include one or more operations, functions, or actions as illustrated by one or more of blocks 702-712. Although the blocks are illustrated in a sequential order, these blocks may in some instances be performed in parallel, and/or in a different order than those described herein. Also, the various blocks may be combined into fewer blocks, divided into additional blocks, and/or removed based upon the desired implementation.

At block 702, the method 700 includes receiving, at a computing device configured to control a vehicle, from a light detection and ranging (LIDAR) sensor coupled to the computing device, LIDAR-based information comprising (i) a three-dimensional (3D) point cloud of a vicinity of a road on which the vehicle is travelling, where the 3D point cloud may comprise points corresponding to light emitted from the LIDAR and reflected from one or more objects in the vicinity of the road, and (ii) intensity values of the reflected light for the points. A LIDAR sensor or unit (e.g., the LIDAR unit 132 in FIG. 1) may be coupled to the vehicle and in communication with the computing device configured to control the vehicle. As described with respect to the LIDAR unit 132 in FIG. 1, LIDAR operation may involve an optical remote sensing technology that enables measuring properties of scattered light to find range and/or other information of a distant target. The LIDAR sensor/unit, for example, may be configured to emit laser pulses as a beam, and scan the beam to generate two dimensional or three dimensional range matrices. In an example, the range matrices may be used to determine distance to an object or surface by measuring time delay between transmission of a pulse and detection of a respective reflected signal.

In another example, the LIDAR sensor may be configured to rapidly scan an environment surrounding the vehicle in three dimensions. In some examples, more than one LIDAR sensor may be coupled to the vehicle to scan a complete 360° horizon of the vehicle. The LIDAR sensor may be configured to provide to the computing device a cloud of point data representing objects, which have been hit by the laser, on the road and the vicinity of the road. The points may be represented by the LIDAR sensor in terms of azimuth and elevation angles, in addition to range, which can be converted to (X, Y, Z) point data relative to a local coordinate frame attached to the vehicle. Additionally, the LIDAR sensor may be configured to provide to the computing device intensity values of the light or laser reflected off the objects.

At block 704, the method 700 includes determining, using the computing device, a set of points in the 3D point cloud representing an area at a height greater than a threshold height from a surface of the road. As described with respect to the method 500, construction zones on roads may be regulated by standard specifications and rules A minimum sign mounting height may be specified for a typical construction zone sign, for example. FIG. 8A illustrates the vehicle 402 travelling on the road 404 and approaching a construction zone indicated by the construction zone sign 412A. The LIDAR sensor coupled to the vehicle 402 may be scanning the horizon and providing the computing device with a 3D point cloud of the road 404 and a vicinity (e.g., sides) of the road 404. Further, the computing device may be configured to determine an area 802 at a height greater than a threshold height 804; the threshold height 804 may be the minimum sign mounting height specified for a typical construction zone sign according to the standard specifications of construction zones, for example. FIG. 8B illustrates a LIDAR-based image 806 including a set of points (e.g., a subset of the 3D point cloud) representing or corresponding to the determined area 802.

Referring back to FIG. 7, at block 706, the method 700 includes estimating, using the computing device, a shape associated with the set of points. The computing device may be configured to identify or estimate a shape depicted by the set of points representing the area at the height greater than the threshold height. For example, the computing device may be configured to estimate dimensional characteristics of the shape. In an example, the computing device may be configured to fit a predetermined shape to the shape depicted in the set of point to estimate the shape. As an example, in FIG. 8B, the computing device may be configured to estimate a diamond shape 808 in the set of point included in the LIDAR-based image 806.

Referring back to FIG. 7, at block 708, the method 700 includes determining, using the computing device, a likelihood that the set of points depicts a construction zone sign, based on the estimated shape and respective intensity values relating to the set of points. In an example, referring to FIG. 8B, the computing device may be configured to match or compare the estimated shape 808 to one or more shapes of typical construction zone signs; and the computing device may be configured to determine a match metric indicative of how similar the estimated shape 808 is to a given predetermined shape (e.g., a percentage of match between dimensional characteristics of the estimated shape 808 and a diamond shape of a typical construction zone sign). In one example, the computing device may be configured to identify edges of the estimated shape 808 and match a shape formed by the edges to a typical diamond shape of typical construction zone signs. The likelihood may be determined based on the match metric, for example.

Further, typical construction zone signs may be required by the standard specifications of construction zones to be made of a retroreflective sheeting materials such as glass beads or prisms, and the computing device may be configured to compare intensity values of points forming the estimated shaped 808 to a threshold intensity value of the retroreflective sheeting material. Based on the comparison, the computing device may be configured to confirm that the estimated shape 808 may represent a given construction zone sign. For example, if the intensity values are close to or within a predetermined value of the threshold intensity value, the computing device may be configured to determine a high likelihood that that the estimated shape 808 may represent a construction zone sign.

In an example, the computing device may be configured to determine a first likelihood based on a comparison of the estimated shape 808 to a predetermined shape of a typical construction zone sign, and may be configured to determine a second likelihood based on a comparison of the intensity values to the threshold intensity value. The computing device may be configured to combine the first likelihood and the second likelihood to determine a single likelihood that the set of points, which includes the points forming the estimated shape 808, depicts a construction zone sign.

In another example, the computing device may be configured to generate a probabilistic model (e.g., a Gaussian distribution), based on the estimated shape 808 (e.g., dimensional characteristics of the estimate shape 808) and the intensity values, to determine the likelihood that the set of points depicts a construction zone sign. For example, the likelihood may be determined as a function of a set of parameter values that are determined based on dimensions of the estimated shape 808 and the respective intensity values. In this example, the likelihood may be defined as equal to probability of an observed outcome (the estimated shape 808 represents a construction zone sign) given those parameter values.

In still another example, the computing device may be configured to cluster points (e.g., the points forming the estimated shape 808) depicted in the LIDAR-based image 806 together into a cluster, based on locations of the points or relative locations of the points to each other. The computing device further may be configured to extract from the cluster of points a set of features (e.g., dimensional characteristics of the estimated shape 808, and the intensity values of the points forming the estimated shape 808). The computing device may be configured to process this set of features through a classifier to determine the likelihood. The classifier can map input information (e.g., the set of features extracted from the cluster of points) to a class (e.g., the cluster represents a construction zone sign). Examples of classifiers, training data, and classification algorithms are described above with regard to block 304 of the method 300 illustrated in FIG. 3.

In one example, the likelihood may be qualitative such as “low,” “medium,” “high” or may be numerical such as a number on a scale, for example. Other examples are possible.

Referring back to FIG. 7, at block 710, the method 700 includes modifying, using the computing device, a control strategy associated with a driving behavior of the vehicle, based on the likelihood. Based on the likelihood (e.g., the likelihood exceeds a predetermined threshold), the computing device may be configured to determine existence of a construction zone sign indicative of an approaching construction zone. Further, the computing device may be configured to determine a type of the construction zone sign to determine severity of road changes due to existence of the construction zone on the road. For example, the computing device may be configured to modify control strategy of the vehicle based on the determined type of the construction zone sign.

As described above with respect to block 508 of the method 500 in FIG. 5, various types of construction zone signs may exist to regulate speed limits when approaching and passing through the construction zone, describe lane changes, closure, reduction, merger, etc., and describe temporary changes to direction of travel on the road, for example. The computing device of the vehicle may be configured to determine the type of the detected construction zone sign based on shape, color, typeface of words, etc. of the detected construction zone sign.

Examples of modifying the control strategy are described above with regard to block 306 of the method 300 illustrated in FIG. 3.

At block 712, the method 700 includes controlling, using the computing device, the vehicle based on the modified control strategy. Controlling the vehicle may include adjusting translational velocity, or rotational velocity, or both, of the vehicle based on the modified driving behavior. Examples of controlling the vehicle based on the modified control strategy are described above with regard to block 308 of the method 300 illustrated in FIG. 3.

In addition to or alternative to detection of the construction zone sign using the LIDAR-based information, the computing device may be configured to detect construction zone objects (e.g., cones, barrels, equipment, vests, chevrons, etc.) using the LIDAR-based information.

FIG. 9 is a flow chart of a method for detection of construction zone objects using LIDAR-based information, in accordance with an example embodiment. FIG. 10A illustrates LIDAR-based detection of construction zone cones in an area within a threshold distance from a surface of the road, in accordance with an example embodiment. FIG. 10B illustrates a LIDAR-based image depicting the area within the threshold distance from the surface of the road, in accordance with an example embodiment. FIG. 10C illustrates LIDAR-based detection of construction zone cones forming a lane boundary, in accordance with an example embodiment. FIG. 10D illustrates a LIDAR-based image depicting construction zone cones forming a lane boundary, in accordance with an example embodiment. FIGS. 9 and 10A-10D will be described together. Detection of construction zone cones is used herein to illustrate the method 900; however, other construction zone objects (e.g., construction zone barrels, equipment, vests, chevrons, etc.) can be detected using the method 900 as well.

The method 900 may include one or more operations, functions, or actions as illustrated by one or more of blocks 902-914. Although the blocks are illustrated in a sequential order, these blocks may in some instances be performed in parallel, and/or in a different order than those described herein. Also, the various blocks may be combined into fewer blocks, divided into additional blocks, and/or removed based upon the desired implementation.

At block 902, the method 900 includes receiving, at a so computing device configured to control a vehicle, from a light detection and ranging (LIDAR) sensor coupled to the computing device, LIDAR-based information relating to a three-dimensional (3D) point cloud of a road on which the vehicle is travelling, where the 3D point cloud may comprise points corresponding to light emitted from the LIDAR and reflected from one or more objects on the road. A LIDAR sensor or unit (e.g., the LIDAR unit 132 in FIG. 1) may be coupled to the vehicle and in communication with the computing device. As described above with respect to the LIDAR unit 132 in FIG. 1, and block 702 of the method 700 illustrated in FIG. 7, the LIDAR sensor may be configured to provide to the computing device a cloud of point data representing objects, on the road and the vicinity of the road. The points may be represented by the LIDAR sensor in terms of azimuth and elevation angles, in addition to range, which can be converted to (X, Y, Z) point data relative to a local coordinate frame attached to the vehicle.

At block 904, the method 900 includes determining, using the computing device, one or more sets of points in the 3D point cloud representing an area within a threshold distance from a surface of the road. As described above with respect to the methods 500 and 700, construction zones on roads may be regulated by standard specifications and rules. As an example, traffic safety cones may be used to separate and guide traffic past a construction zone work area. Cones may be specified to be about 18 inches tall, for example. In another example, for high speed and high volume of traffic, or nighttime operations, the cones may be specified to be 28 inches tall, and retro-reflectorized, or comprising bands made of retroreflective material. These examples are for illustration only, and other examples are possible.

FIG. 10A illustrates the vehicle 402 travelling on the road 404 and approaching a construction zone indicated by the construction zone cone 406. The LIDAR sensor coupled to the vehicle 402 may be configured to scan the horizon and provide the computing device with a 3D point cloud of the road 404 and a vicinity of the road 404. Further, the computing device may be configured to determine an area 1002 within a threshold distance 1004 of a surface of the road 404. For example, the threshold distance 1004 may be about 30 inches or more to include cones of standardized lengths (e.g., 18 inches or 28 inches). Other threshold distances are possible based on the standard specifications regulating a particular construction zone. FIG. 10B illustrates a LIDAR-based image 1006 including sets of points representing objects in the area 1002.

Referring back to FIG. 9, at block 906, the method 900 includes identifying one or more construction zone objects in the one or more sets of points. For example, the computing device may be configured to identify shapes of objects represented by the sets of points of the LIDAR-based 3D point cloud. For example, the computing device may be configured to estimate characteristics (e.g., dimensional characteristics) of a shape of an object depicted by a set of points, and may be configured to fit a predetermined shape to the shape to identify the object. As an example, in FIG. 10B, the computing device may be configured to identify a construction zone cone 1008 in the LIDAR-based image 1006.

In an example, to identify the construction zone objects in the sets of points, the computing device may be configured to determine, for each identified construction zone object, a respective likelihood of the identification. As an example, in FIG. 10B, the computing device may be configured to determine a shape of the cone 1008 defined by respective points of a set of points representing the cone 1008. Further, the computing device may be configured to match the shape to one or more shapes of standard construction zone cones. The computing device may be configured to determine a match metric indicative of how similar the shape is to a given standard shape of a typical construction zone cone (e.g., a percentage of match between dimensional characteristics of the shape and the given standard shape). The respective likelihood may be determined based on the match metric.

In another example, in addition to or alternative to identifying the cone 1008 based on shape, the computing device may be configured to cluster points (e.g., the points forming the cone 1008) depicted in the LIDAR-based image 1006 together into a cluster, based on locations of the points or relative locations of the points to each other. The computing device further may be configured to extract from the cluster of points a set of features (e.g., minimum height of the points, maximum height of the points, number of the points, width of the cluster of points, general statistics of the points at varying heights, etc.). The computing device may be configured to process this set of features through a classifier to determine whether the cluster of points represent a given construction zone cone. The classifier can map input information (e.g., the set of features extracted from the cluster of points) to a class (e.g., the cluster represents a construction zone cone). Examples of classifiers, training data, and classification algorithms are described above with regard to block 304 of the method 300 illustrated in FIG. 3.

Further, typical construction zone cones may be required by the standard specifications of construction zones to be made of a retroreflective sheeting materials such as glass beads or prisms, and the computing device may be configured to compare intensity values of points forming the cone 1008 to a threshold intensity value of the retroreflective sheeting material. Based on the comparison, the computing device may be configured to confirm identification of the cone 1008, for example.

In some examples, the computing device may be configured to exclude cones that are away from the road by a certain distance, since such cones may indicate a work zone that is away from the road and may not affect traffic. Also, in an example, the computing device may be configured to exclude sets of points that represent objects that clearly cannot be construction zone cones based on size as compared to a typical size of typical construction zone cones (e.g., too large or too small to be construction zone cones).

In an example, for reliable identification of a construction zone cone, the computing device may be configured to identify the construction zone cone based on LIDAR-based information received from two (or more) consecutive scans by the LIDAR to confirm the identification and filter out false identification caused by electronic or signal noise in a single scan.

Referring back to FIG. 9, at block 908, the method 900 includes determining, using the computing device, a number and locations of the one or more construction zone objects. As an example, in addition to specifying dimensional characteristics and reflective properties of typical construction zone cones, the standard specifications for construction zones also may specify requirements for number of and spacing between the cones. In an example, tighter spacing may be specified, under some conditions, to enhance guidance of vehicles and drivers. Table 1 illustrates an example of minimum spacing between construction zone cones based on speed limits.

TABLE 1
		Spacing for Speed A	Spacing for Speed B
	Speed A: 50 mph	40 fl	80 fl
	Speed B: 70 mph
	Speed A: 35 mph	30 fl	60 fl
	Speed B: 45 mph
	Speed A: 20 mph	20 fl	40 fl
	Speed B: 30 mph

These examples are for illustration only. Other examples of spacing requirements are possible as well.

In an example, if respective likelihoods for identification of the identified cones exceed a threshold likelihood, the computing device may be configured to further determine the number and locations of the identified cones. In FIG. 10C, the computing device may be configured to detect or identify, based on the LIDAR-based information, the cone(s) 406 and also determine number of the cone(s) 406 as well as locations or relative locations of the cone(s) 406 with respect to each other. For example, the computing device may be configured to determine a distance 1010 between the cone(s), and compare the distance 1010 with a predetermined distance (or spacing) specified in the standard specifications.

FIG. 10D illustrates a LIDAR-based image 1011 including sets of points representing construction zone cones 1012A-D. In addition to detecting or identifying the construction zone cones 1012A-D in the LIDAR-based image 1011, the computing device may be configured to estimate a respective distance between pairs of cones.

Referring back to FIG. 9, at block 910, the method 900 includes determining, using the computing device, a likelihood of existence of a construction zone, based on the number and locations of the one or more construction zone objects. As an example, a single cone on a side of the road may not be indicative of an active construction zone. Therefore, in addition to detecting presence of or identifying cones on the road, the computing device may be configured, for example, to determine, based on the number and locations (e.g., relative distance) of the cones, that the cones may form a lane boundary and are within a predetermined distance of each other, which may be indicative of an active construction zone causing road changes. The computing device thus may be configured to determine a likelihood or confirm that the cones are indicative of a construction zone, based on the determined number and locations of the cones.

In an example, the computing device may be configured to determine the number and locations of the construction zone cones based on the LIDAR-based information, and compare a pattern formed by the identified construction zone cones to a typical pattern formed by construction zone cones in a typical construction zone (e.g., pattern of cones forming a lane boundary). The computing device may be configured to determine that the detected construction zone cones are associated with a construction zone based on the comparison, and determine the likelihood accordingly.

In another example, the computing device may be configured to generate a probabilistic model (e.g., a Gaussian distribution), based on the determined number and locations of the cones to determine the likelihood of existence of the construction zone. For example, the likelihood may be determined as a function of a set of parameter values that are determined based on the number and locations of the identified cones. In this example, the likelihood may be defined as equal to probability of an observed outcome (the cones are indicative of a construction zone on the road) given those parameter values.

In still another example, the computing device may be configured to process information relating to the number and locations of the cones through a classifier to determine the likelihood. The classifier can map input information (e.g., the number and location of the cones) to a class (e.g., existence of the construction zone). Examples of classifiers and classification algorithms are described above with regard to block 304 of the method 300 illustrated in FIG. 3.

As an example, a training computing device may be configured to receive training data for a plurality of driving situations of a given vehicle. For example, respective training data may include, for each of the plurality of driving situations, respective LIDAR-based information relating to a respective 3D point cloud of a respective road. Based on the respective LIDAR-based information of the respective training data, the computing device may be configured to identify respective cones as well as determine respective number and locations of the respective cones. Also, the computing device may be configured to receive positive or negative indication of respective existence of a respective construction zone corresponding to the respective training data for each of the driving situations. Further the training computing device may be configured to correlate, for each driving situation, the positive or negative indication with the respective training data, and determine parameters (e.g., vector of weights for equation 1) of the classifier based on the correlations for the plurality of driving situations. These parameters may be provided to the computing device configured to control the vehicle such that as the computing device receives the LIDAR-based information, the computing device may be configured to process the LIDAR-based information through the classifier using the determined parameters of the classifier to determine the likelihood.

In one example, the likelihood may be qualitative such as “low,” “medium,” “high” or may be numerical such as a number on a scale, for example. Other examples are possible.

Referring back to FIG. 9, at block 912, the method 900 includes modifying, using the computing device, a control strategy associated with a driving behavior of the vehicle, based on the likelihood of the existence of the construction zone on the road. The computing device may be configured to modify or select, based on the determined likelihood of the existence of the construction zone, a control strategy comprising rules for actions that control the vehicle speed to safely maintain a distance with other objects and select a lane that is considered safest given road changes due to the existence of the construction zone. Examples of modifying the control strategy based on the likelihood are described above with regard to block 306 of the method 300 illustrated in FIG. 3.

At block 914, the method 900 includes controlling, using the computing device, the vehicle based on the modified control strategy. Controlling the vehicle may include adjusting translational velocity, or rotational velocity, or both, of the vehicle based on the modified driving behavior. Examples of controlling the vehicle based on the modified control strategy are described above with regard to block 308 of the method 300 illustrated in FIG. 3.

In some embodiments, the disclosed methods may be implemented as computer program instructions encoded on a computer-readable storage media in a machine-readable format, or on other non-transitory media or articles of manufacture. FIG. 11 is a schematic illustrating a conceptual partial view of an example computer program product 1100 that includes a computer program for executing a computer process on a computing device, arranged according to at least some embodiments presented herein. In one embodiment, the example computer program product 1100 is provided using a signal bearing medium 1101. The signal bearing medium 1101 may include one or more program instructions 1102 that, when executed by one or more processors may provide functionality or portions of the functionality described above with respect to FIGS. 1-10. Thus, for example, referring to the embodiments shown in FIGS. 3, 5, 7, and 9, one or more features of blocks 302-308, 502-512, 702-712, and 902-914 may be undertaken by one or more instructions associated with the signal bearing medium 1101. In addition, the program instructions 1102 in FIG. 11 describe example instructions as well.

In some examples, the signal bearing medium 1101 may encompass a computer-readable medium 1103, such as, but not limited to, a hard disk drive, a Compact Disc (CD), a Digital Video Disk (DVD), a digital tape, memory, etc. In some implementations, the signal bearing medium 1101 may encompass a computer recordable medium 1104, such as, but not limited to, memory, read/write (R/W) CDs, R/W DVDs, etc. In some implementations, the signal bearing medium 1101 may encompass a communications medium 1105, such as, but not limited to, a digital and/or an analog communication medium (e.g., a fiber optic cable, a waveguide, a wired communications link, a wireless communication link, etc.). Thus, for example, the signal bearing medium 1101 may be conveyed by a wireless form of the communications medium 1105 (e.g., a wireless communications medium conforming to the IEEE 802.11 standard or other transmission protocol).

The one or more programming instructions 1102 may be, for example, computer executable and/or logic implemented instructions. In some examples, a computing device such as the computing device described with respect to FIGS. 1-10 may be configured to provide various operations, functions, or actions in response to the programming instructions 1102 conveyed to the computing device by one or more of the computer readable medium 1103, the computer recordable medium 1104, and/or the communications medium 1105. It should be understood that arrangements described herein are for purposes of example only. As such, those skilled in the art will appreciate that other arrangements and other elements (e.g. machines, interfaces, functions, orders, and groupings of functions, etc.) can be used instead, and some elements may be omitted altogether according to the desired results. Further, many of the elements that are described are functional entities that may be implemented as discrete or distributed components or in conjunction with other components, in any suitable combination and location.

While various aspects and embodiments have been disclosed herein, other aspects and embodiments will be apparent to those skilled in the art. The various aspects and embodiments disclosed herein are for purposes of illustration and are not intended to be limiting, with the true scope being indicated by the following claims, along with the full scope of equivalents to which such claims are entitled. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting.

Claims

1. A method comprising:

receiving, at a computing device configured to control a vehicle, from an image-capture device coupled to the vehicle, an image of a vicinity of a road on which the vehicle is travelling;

selecting, from the image, using the computing device, an image portion corresponding to a side of the road at a predetermined height;

detecting, using the computing device, a construction zone sign in the image portion;

determining, using the computing device, a type of the construction zone sign;

detecting one or more construction zone objects on the road;

determining, based on a number and locations of the one or more construction zone objects, a severity of road changes; and

controlling, using the computing device, the vehicle based on (i) the type of the construction zone sign, and (ii) the construction zone sign and the severity of road changes.

2. The method of claim 1, wherein controlling the vehicle comprises controlling the vehicle in an autonomous operation mode.

3. The method of claim 1, wherein detecting the construction zone sign comprises:

detecting a candidate construction zone sign in the image portion; and

determining that the candidate construction zone sign relates to a construction zone based on one or more of shape, color, and pattern of the candidate construction zone sign.

4. The method of claim 1, further comprising:

detecting a candidate construction zone sign in the image portion;

determining a first likelihood that the candidate construction zone sign relates to a construction zone based on one or more of shape, color, and pattern of the candidate construction zone sign;

receiving, at the computing device, from a light detection and ranging (LIDAR) device coupled to the vehicle, LIDAR-based information comprising a three-dimensional (3D) point cloud corresponding to the image portion, wherein the 3D point cloud comprises a set of points based on light emitted from the LIDAR and reflected from a surface of the candidate construction zone sign; and

determining a second likelihood that the candidate construction zone sign relates to the construction zone based on the LIDAR-based information, wherein detecting the construction zone sign is based on the first likelihood and the second likelihood.

5. The method of claim 1, further comprising:

detecting a candidate construction zone sign in the image portion;

determining a first likelihood that the candidate construction zone sign relates to a construction zone, based on one or more of shape, color, and pattern of the candidate construction zone sign;

receiving, at the computing device, from a radio detection and ranging (RADAR) device coupled to the vehicle, RADAR-based information relating to location and one or more characteristics of the candidate construction zone sign; and

determining a second likelihood that the candidate construction zone sign relates to the construction zone based on the RADAR-based information, wherein detecting the construction zone sign is based on the first likelihood and the second likelihood.

6. The method of claim 5, wherein the RADAR-based information includes dimensional characteristics of the candidate construction zone sign and indicates that the candidate construction zone sign is stationary.

7. The method of claim 1, wherein determining the type of the construction zone sign comprises further comprising:

determining a type of the construction zone sign by determining a shape of the construction zone sign; and

matching the shape to one or more of shapes of typical construction zone signs, and wherein controlling the vehicle is further based on the determined type of construction sign.

8. The method of claim 1, wherein the one or more so construction zone objects comprise a cone, a barrel, a worker, a barrier, or construction equipment.

9. A non-transitory computer readable medium having stored thereon instructions that, when executed by a computing device of a vehicle, cause the computing device to perform operations comprising:

receiving, from an image-capture device coupled to the vehicle, an image of a vicinity of a road of travel of the vehicle;

selecting, from the image, an image portion corresponding to a side of the road;

detecting a construction zone sign in the image portion;

determining a type of the construction zone sign;

detecting one or more construction zone objects on the road;

determining, based on a number and locations of the one or more construction zone objects, a severity of road changes; and

controlling the vehicle based on (i) the type of the construction zone sign, and (ii) the construction zone sign and the severity of road changes.

10. The non-transitory computer readable medium of claim 9, wherein the one or more construction zone objects comprise a cone, a barrel, a worker, a barrier, or construction equipment.

11. The non-transitory computer readable medium of claim 9, wherein the operations further comprise:

identifying a lane boundary defined by the one or more constructions zone objects, wherein determining the severity of road changes is further based on the identified lane boundary.

12. The non-transitory computer readable medium of claim 9, wherein the operations further comprise:

detecting a candidate construction zone sign in the image portion;

determining a first likelihood that the candidate construction zone sign relates to a construction zone, based on one or more of shape, color, and pattern of the candidate construction zone sign;

receiving, from a light detection and ranging (LIDAR) device coupled to the vehicle, LIDAR-based information comprising a three-dimensional (3D) point cloud corresponding to the image portion, wherein the 3D point cloud comprises a set of points based on light emitted from the LIDAR and reflected from a surface of the candidate construction zone sign; and

determining a second likelihood that the candidate construction zone sign relates to the construction zone based on the LIDAR-based information, wherein detecting the construction zone sign is based on the first likelihood and the second likelihood.

13. The non-transitory computer readable medium of claim 9, wherein the operations further comprise:

detecting a candidate construction zone sign in the image portion;

determining a first likelihood that the candidate construction zone sign relates to a construction zone, based on one or more of shape, color, and pattern of the candidate construction zone sign;

receiving from a radio detection and ranging (RADAR) device coupled to the vehicle, RADAR-based information relating to location and one or more characteristics of the candidate construction zone sign; and

determining a second likelihood that the candidate construction zone sign relates to the construction zone based on the RADAR-based information, wherein detecting the construction zone sign is based on the first likelihood and the second likelihood.

14. The non-transitory computer readable medium of claim 13, wherein the RADAR-based information includes dimensional characteristics of the candidate construction zone sign and indicates that the candidate construction zone sign is stationary.

15. The non-transitory computer readable medium of claim 9, wherein determining the type of the construction zone sign comprises further comprising:

determining a type of the construction sign by determining a shape of the construction zone sign in the image portion; and

matching the shape to one or more of shapes of typical construction zone signs, and wherein controlling the vehicle is further based on the determined type of construction sign.

16. A control system for a vehicle, comprising:

an image-capture device;

a computing device in communication with the image-capture device; and

data storage comprising instructions that, when executed by the computing device, cause the control system to perform operations comprising:

receiving, from an image-capture device coupled to the vehicle, an image of a vicinity of a road of travel of the vehicle,

selecting, from the image, an image portion corresponding to a side of the road;

detecting a construction zone sign in the image portion,

determining a type of the construction zone sign,

detecting one or more construction zone objects on the road,

determining, based on a number and locations of the one or more construction zone objects, a severity of road changes, and

controlling the vehicle based on (i) the type of the construction zone sign, and (ii) the construction zone sign and the severity of road changes.

17. The control system of claim 16, wherein controlling the vehicle comprises controlling the vehicle in an autonomous operation mode.

18. The control system of claim 16, wherein detecting the construction zone sign comprises:

detecting a candidate construction zone sign in the image portion; and

determining that the candidate construction zone sign relates to a construction zone based on one or more of shape, color, and pattern of the candidate construction zone sign.

19. The control system of claim 16, wherein the one or more construction zone objects comprise a cone, a barrel, a worker, a barrier, or construction equipment.

20. The control system of claim 16, wherein the operations further comprise:

identifying a lane boundary defined by the one or more constructions zone objects, wherein determining the severity of road changes is further based on the identified lane boundary.

21. The method of claim 1, further comprising selecting a control strategy for controlling the vehicle based on the severity of road changes, and wherein controlling the vehicle is further based on the control strategy.

22. The method of claim 21, wherein the control strategy comprises default driving behavior.

23. The method of claim 21, wherein the control strategy comprises defensive driving behavior.

24. The method of claim 1, wherein the controlling the vehicle is further based on traffic information.

25. The method of claim 24, wherein the traffic information is received from one or more of (i) Global Positioning Satellite (GPS) device, (ii) vehicle-to-infrastructure communication, (iii) vehicle-to-vehicle communication, and (iv) a traffic report broadcast.

26. The method of claim 1, further comprising:

determining a type of the construction zone sign by determining a shape of the construction zone sign; and

matching the shape to one or more of shapes of typical construction zone signs, and wherein controlling the vehicle is further based on the determined type of construction sign.