A method and system for detecting the retroreflective brightness coefficient of vehicle-mounted signs

Abstract

The present invention discloses a vehicle-mounted sign retroreflection brightness coefficient detection method and system, which obtain image data and point cloud data of a road area to be detected taken by an observation module on the top of a detection vehicle, and obtain the relative distance between the sign to be detected in the road area to be detected and the detection vehicle according to a pre-trained first deep learning detection model; when the relative distance meets the preset shooting distance, the detection module is triggered to take continuous photos of the sign to be detected with and without light source illumination, and two continuous frame images obtained by the continuous photos are respectively input into a deep learning classification and recognition model to obtain a target sign image, and the difference between the average grayscale values ​​of each target sign image is calculated to obtain an interference-free grayscale value, and the retroreflection brightness coefficient of the sign to be detected is obtained according to a preset retroreflection conversion formula, and the continuous frame images are used to eliminate the interference caused by the external environment when measuring the retroreflection brightness coefficient of the sign, thereby improving the measurement accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of road detection based on laser radar detection and image processing, and in particular to a method and system for detecting the retroreflective brightness coefficient of a vehicle-mounted sign. Background Art

[0002] Road traffic signs are important infrastructure to promote road traffic safety and improve road operation efficiency. They can provide drivers with information such as traffic prohibitions, warnings, instructions, and road traffic conditions to prevent or reduce traffic injuries and ensure driving safety. However, due to the erosion of sunlight, rain, pollution, dust or human damage in the natural environment, the reflective film of traffic signs fades, breaks, cracks, and the retroreflective effect decreases. When the reflective effect of the sign is insufficient, it seriously affects driving safety and may even endanger the life and property of road users. The retroreflective brightness coefficient of traffic signs is a key evaluation criterion for measuring the safety performance and reading performance of traffic signs. The effective reading distance of road traffic signs is limited. Therefore, the stronger the retroreflective ability of the retroreflective material on the surface of the sign, the more it can improve the brightness of the sign, optimize sign recognition, and improve the sight distance. Measuring the retroreflective brightness coefficient of traffic signs can provide an accurate basis for highway maintenance, and can timely discover, repair, and replace signs with insufficient reflective effects.

[0003] At present, there are two main methods for measuring the retroreflective brightness coefficient of in-service road traffic signs: a handheld retroreflective brightness coefficient measuring instrument and a portable retroreflective brightness coefficient measuring instrument. The technology of the handheld retroreflective brightness coefficient measuring instrument is relatively mature. When measuring in-service road traffic signs, it is generally below the sign, with the help of a ladder or a lifting vehicle to raise the measurement personnel to the position of the sign, and use the handheld measuring instrument to perform direct contact measurement of the sign. This method is time-consuming and labor-intensive. It is limited by the manual single-point operation mode, has low measurement efficiency, high working costs, and can only be used for random inspections. The portable retroreflective brightness coefficient measuring instrument uses a fixed bracket or a ground trolley to raise the measuring instrument installed on the telescopic rod to the sign surface, and the inspection personnel operate the measuring device on the ground to approach the sign for direct contact measurement. This device reduces the safety risks of measurement personnel when climbing to a certain extent, but there are still problems such as difficulty in carrying the device, long measurement time, low efficiency, only random inspections, and incomplete sign inspection.

[0004] Therefore, how to improve the comprehensiveness and efficiency of traffic sign retroreflective brightness coefficient detection and reduce time and money costs is an urgent problem to be solved. Summary of the invention

[0005] The purpose of the present invention is to provide a vehicle-mounted sign retro-reflective brightness coefficient detection method and system, which measures the distance and position information of traffic signs by combining laser radar and cameras, and uses continuous frame images to eliminate the interference caused by the external environment when measuring the retro-reflective brightness coefficient of the sign, thereby improving the measurement accuracy.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign, which specifically comprises the following steps:

[0008] S1. Obtain image data and point cloud data of the road area to be detected taken by the observation module on the top of the detection vehicle, input the image data of the road area to be detected into a pre-trained first deep learning detection model, and combine the point cloud data to obtain the relative distance between the sign to be detected in the road area to be detected and the detection vehicle;

[0009] S2, determining whether the relative distance between the sign to be detected and the detection vehicle meets the preset shooting distance. If so, triggering the detection module to take continuous photos of the sign to be detected with and without light source illumination to obtain two continuous frame images;

[0010] S3, inputting the two continuous frame images into the deep learning classification and recognition model respectively, and obtaining two target sign images with a plurality of color regions, color labels corresponding to each color region, grayscale values ​​and color region coordinates;

[0011] S4, calculating the average grayscale value corresponding to each color label in each target sign image according to the color label and grayscale value corresponding to each color area;

[0012] S5. Calculate the difference between the average grayscale values ​​corresponding to the same color label in the two target signboard images to obtain the interference-free grayscale value corresponding to each color label, input the interference-free grayscale value into the retroreflection conversion formula corresponding to each color label, and obtain the retroreflection brightness coefficient corresponding to each color label in the signboard to be detected.

[0013] In some specific implementations, the specific process of step S1 is:

[0014] S11, inputting the image data of the road area to be detected into the first deep learning detection model to obtain the first signboard data to be detected with the attribute category and the minimum circumscribed rectangular positioning frame label;

[0015] S12, preprocessing the point cloud data of the road area to be detected to obtain the point cloud data of the area to be detected of the sign to be detected;

[0016] S13, mapping the point cloud data of the to-be-detected area of ​​the to-be-detected sign to the pixel coordinate system through a preset radar-visual fusion conversion matrix to obtain data of a second to-be-detected sign;

[0017] S14. Filter the data of the second signboard to be detected according to the data of the first signboard to be detected to obtain the category of the signboard to be detected, the reflection intensity, and the relative distance and relative position from the detection vehicle.

[0018] In some specific implementations, step S2 further includes the steps of:

[0019] If the relative distance between the signboard to be detected and the inspection vehicle meets the preset shooting distance, the detection module at the corresponding position is triggered to control the light source to take continuous photos of the signboard to be detected with and without light illumination according to the relative position of the signboard to be detected and the inspection vehicle, and two continuous frame images are obtained. The relative position includes the detection signboard being above or to the right of the inspection vehicle, and the detection module includes an upper detection module and a right detection module.

[0020] In some specific implementations, the specific process of step S2 triggering the detection module to take a photo of the sign to be detected is:

[0021] S21, setting the length L1 of the interval to be photographed, and obtaining the distance L4 to be photographed according to the sum of the length L1 of the interval to be photographed and the preset shooting distance L;

[0022] S22, determine whether the relative distance L2 between the sign to be detected and the detection vehicle satisfies the requirement of being less than the distance to be photographed L4 and greater than the preset photographing distance L. If so, record the road mileage x1 of the detection vehicle traveled obtained from the mileage measurement module at this time.

[0023] S23, calculating the difference between the relative distance L2 in step S22 and the preset shooting distance L to obtain the delayed shooting distance L3;

[0024] S24, obtaining the road mileage x2 of the inspection vehicle measured by the mileage measurement module in real time, calculating the difference between the road mileage x2 and x1, and obtaining the mileage △x;

[0025] S25. When the driving mileage △x reaches the delayed shooting distance L3, the detection module is triggered to take continuous photos of the sign to be detected with and without light source illumination to obtain two continuous frame images.

[0026] In some specific implementations, the deep learning classification and recognition model includes a sign recognition model and a multi-classification model, and the specific process of step S3 is:

[0027] S31, inputting two consecutive frame images into a sign recognition model respectively, identifying the sign category and the minimum bounding rectangle positioning frame in each consecutive frame image, and cutting out the sign area from the consecutive frame images according to the minimum bounding rectangle positioning frame to obtain two corresponding first target images with sign category identification;

[0028] S32, inputting the two first target images into the multi-classification model respectively, and for each first target image, segmenting the first target image into a plurality of corresponding color regions according to the color label, to obtain two second target images with the plurality of color regions, the color labels corresponding to the color regions, and the color region coordinates;

[0029] S33. Preprocess the two continuous frame images respectively, and use the preprocessed continuous frame images to calculate the average grayscale value corresponding to each color area in the second target image corresponding to each continuous frame image, and finally obtain two target signboard images corresponding to the two continuous frame images with a number of color areas, color labels corresponding to each color area, grayscale values ​​and color area coordinates.

[0030] In some specific implementations, the specific process of step S4 is:

[0031] S41, for each target sign image, dividing the target sign image into a plurality of even target areas;

[0032] S42, counting the color areas contained in each target area, and classifying the contained color areas according to color labels, and calculating the average grayscale value corresponding to the color areas with the same color label in each target area according to the grayscale value corresponding to each color area contained in each target area.

[0033] In some specific implementations, the specific process of step S5 is:

[0034] The difference between the average grayscale values ​​of the corresponding color labels of the corresponding target areas in the two target signboard images is calculated to obtain the non-interference grayscale value corresponding to the same color label in each target area;

[0035] The non-interference grayscale value corresponding to each color label in each target area is respectively input into the retroreflection conversion formula corresponding to each color label to obtain the retroreflection brightness coefficient corresponding to each color label in each target area.

[0036] In some specific embodiments, the retroreflective conversion formula is:

[0037] ax+by= 0;

[0038] Among them, x represents the non-interference grayscale value corresponding to the color label to be calculated in each target area, represents the retroreflection brightness coefficient corresponding to the color label in the target area, and a and b represent the pre-fitted coefficients corresponding to each color label.

[0039] In some specific implementations, the radar-visual fusion conversion matrix in step S13 is:

[0040]

[0041] in, represents the radar coordinates, ( u , v ) represents the pixel coordinates, Indicates the distance between the camera and the shooting point in the camera coordinate system. Represents the camera intrinsic parameter matrix, R is the rotation matrix from radar to camera, and T is the translation matrix from radar to camera.

[0042] In some specific implementations, the first deep learning detection model uses a target detection model based on yolov8, and the process of obtaining the point cloud data of the area to be detected of the sign to be detected is:

[0043] The point cloud data of the road area to be detected is subjected to noise removal and screening processing to obtain pre-processed point cloud data, which includes the position and reflection intensity of the interfering objects and traffic signs;

[0044] According to the difference in reflection intensity between the interfering object and the signboard, the point cloud data of the area to be detected of the signboard to be detected is extracted from the point cloud data of the road area to be detected.

[0045] In a second aspect, the present application proposes a vehicle-mounted sign retroreflective brightness coefficient detection system, comprising an observation module, a detection module and a mileage measurement module respectively connected to a control module for communication, wherein:

[0046] The observation module includes a camera and a radar. The optical axes of the camera and the radar are installed on the roof of the inspection vehicle parallel to the length direction of the inspection vehicle to capture images of the road area to be inspected.

[0047] The detection module includes a detection camera and a light source for lighting, and the detection camera and the light source for lighting are installed on the roof of the detection vehicle at a set angle to shoot the sign to be detected on the road area to be detected;

[0048] The mileage measurement module is installed on the left rear wheel hub of the inspection vehicle to measure the road mileage traveled by the inspection vehicle;

[0049] The control module is used to execute a method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign as described in the first aspect.

[0050] The inventive concept of this application is:

[0051] There are two main existing methods for measuring the retroreflective brightness coefficient of road traffic signs: a handheld retroreflective brightness coefficient measuring instrument and a portable retroreflective brightness coefficient measuring instrument. Both methods have the problem of low efficiency in collection, measurement and maintenance.

[0052] Based on the above problems, the present application proposes a method of first obtaining the image and point cloud data of the road area to be detected by the radar and camera of the observation module, combining the image and point cloud data of the road area to be detected to measure the distance and position information of the traffic sign, and judging the relative distance between the detection vehicle and the sign based on the distance and position information of the sign. When the relative distance reaches the preset shooting distance, the detection module is triggered to take continuous photos of the sign to be detected. When taking photos continuously, two images of the continuous frames of the current sign to be detected are obtained. The first image is a bright image when there is light source illumination, and the second image is a dark image when there is no light source illumination. Using the images of continuous frames, the interference caused by the external environment when measuring the retroreflection brightness coefficient of the sign is eliminated, and the measurement accuracy is improved. And based on the two continuous frame images, the influence of illumination and reflection on the measurement can be avoided, and the color grayscale value calculated by obtaining the continuous frame images is obtained by taking the difference, and the grayscale value without interference under the condition of only light source is obtained. Based on the retroreflection conversion formula between the preset grayscale value and the retroreflection brightness coefficient, the retroreflection brightness coefficient of the sign to be detected can be finally obtained according to the obtained grayscale value.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] 1. The present invention adopts a detection vehicle equipped with a detection system, without closing the road. The detection vehicle detects the signboard at a normal road speed. The speed of the detection vehicle does not affect the detection result, which can greatly improve the detection efficiency;

[0055] 2. A method combining laser radar and camera is proposed to measure the distance and position information of traffic signs. The image of continuous frames is used while the image grayscale value is used to estimate the retroreflective brightness coefficient of the sign, eliminating the interference caused by the external environment when measuring the retroreflective brightness coefficient of the sign, and improving the measurement accuracy;

[0056] 3. In order to ensure the real-time shooting and timely detection when the vehicle speed is fast, the present invention sets a waiting shooting interval for the detection module, adopts a trigger mechanism for extended-distance shooting, and improves the efficiency and accuracy of the measurement of the retroreflective brightness coefficient of the sign. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A structural block diagram of a vehicle-mounted sign retroreflective brightness coefficient detection system provided by an embodiment of the present invention;

[0058] Figure 2A schematic diagram of the installation positions of various modules in the system provided by an embodiment of the present invention;

[0059] Figure 3 A flow chart of a method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign provided by an embodiment of the present invention;

[0060] Figure 4 A schematic diagram of dividing a target sign image into nine target areas provided by an embodiment of the present invention;

[0061] Figure 5 A schematic diagram of photography by a trigger detection module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] The relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0064] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0065] Additionally, descriptions of well-known structures, functions, and configurations may be omitted for clarity and conciseness.One of ordinary skill in the art will recognize that various changes and modifications may be made to the examples described herein without departing from the spirit and scope of the present disclosure.

[0066] Technologies, methods, and apparatus known to ordinary technicians in the relevant field may not be discussed in detail, but where appropriate, such technologies, methods, and apparatus should be considered part of the authorization specification.

[0067] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0068] Example 1

[0069] like Figure 1As shown, this embodiment provides a vehicle-mounted sign retroreflective brightness coefficient detection system, including an observation module, a detection module and a mileage measurement module respectively connected to the control module for communication, wherein:

[0070] The observation module includes a camera and a radar. The optical axes of the camera and the radar are installed on the roof of the inspection vehicle parallel to the length direction of the inspection vehicle to capture images of the road area to be inspected.

[0071] The detection module includes a detection camera and a light source for lighting, which are installed on the roof of the detection vehicle at a set angle to shoot the signs to be detected on the road area to be detected; the detection module includes an upper sign detection module and a right sign detection module for shooting the signs above and on the right side of the road area to be detected;

[0072] The mileage measurement module is installed on the left rear wheel hub of the inspection vehicle to measure the road mileage traveled by the inspection vehicle;

[0073] The control module executes a vehicle-mounted sign retro-reflective brightness coefficient detection method according to the road area image and point cloud data to be detected, the sign to be detected, the road mileage of the detected vehicle and other data received from each module.

[0074] like Figure 2 As shown, each module in the traffic sign retroreflective brightness coefficient measurement system of this embodiment is mounted on a detection vehicle, and the length direction of the detection vehicle is parallel to the driving direction of the road. The detection vehicle is equipped with an observation module, an upper sign detection module, a right sign detection module, a control module, a mileage measurement module, and other parts.

[0075] The present invention discloses a method for measuring the retroreflective brightness coefficient of a vehicle-mounted traffic sign. The measuring method adopts a detection vehicle, on which an observation module for photographing an image of a road area to be detected is installed; an upper sign detection module and a right sign detection module for photographing the sign to be detected on the road area to be detected; a control module for controlling the operation of a camera, a light source and a radar and performing inference calculation on the collected images, and a mileage measurement module for measuring the road mileage traveled by the detection vehicle. The observation module includes an observation camera and a laser radar, and it is necessary to ensure that the optical axes of the observation camera and the laser radar are parallel to the length direction of the detection vehicle. In this embodiment, the height of the radar and camera installation position from the ground is 2.01 meters. The upper sign detection module and the right sign detection module each include a detection camera and a light source for illumination. The control module is controlled by a single-chip microcomputer as the control core. The core of the mileage measurement module is a photoelectric encoder, which is installed on the left rear wheel hub of the detection vehicle. The cameras used in this embodiment are all monocular cameras, and the light source is a xenon lamp. And the sign to be detected in this embodiment refers to the upper and right sign of other grade roads such as expressways, national and provincial roads.

[0076] Example 2

[0077] like Figure 3 , based on Example 1 Figure 1 The present embodiment provides a method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign, and the method specifically comprises the following steps:

[0078] S1. Obtain image data and point cloud data of the road area to be detected taken by the observation module on the top of the detection vehicle, input the image data of the road area to be detected into a pre-trained first deep learning detection model, and combine the point cloud data to obtain the relative distance between the sign to be detected in the road area to be detected and the detection vehicle;

[0079] When the inspection vehicle is traveling on the road to be inspected, the radar and camera of the observation module shoot the area to be inspected at a fixed frequency. The fixed frequency can be freely set by the control module to trigger the distance or time interval. In this embodiment, the fixed time interval is used to trigger the acquisition, and the acquisition interval is 200ms. When the inspection vehicle is traveling on the road to be inspected, the control module triggers the camera and radar in the observation module at a fixed frequency to simultaneously capture the two-dimensional image data and point cloud data of the road area to be inspected.

[0080] Specifically, the specific process of step S1 is:

[0081] S11, inputting the image data of the road area to be detected into the first deep learning detection model to obtain the first signboard data to be detected with the attribute category and the minimum circumscribed rectangular positioning frame label;

[0082] Specifically, the first deep learning detection model adopts a target detection model based on yolov8, whose lightweight and real-time performance can ensure good detection effect and detection speed under resource-constrained conditions.

[0083] The observation module can be used to collect image sample data. The two-dimensional image of the signboard obtained by the observation module in advance is used to create a deep learning target detection training set, and the target detection training set is input into the pre-built deep learning network to obtain the first deep learning detection model. In the process of collecting and processing data sets, the balance of the categories and numbers of various signboards must be ensured. There are three types of annotations: large, medium, and small, which are used to identify signboards of different sizes. The consistency and accuracy of the annotations must be ensured. During training, it is necessary to set network parameters, including the number of training rounds, batch size, learning rate, confidence loss coefficient and other parameters. Finally, refer to the recall rate Recall, accuracy Precision, map0.5 and map0.5:0.95 to comprehensively evaluate the detection effect of the model.

[0084] When the detection system is working, for each image of the road area to be detected, the first deep learning detection model is used to infer the attribute category of the sign to be detected in the area to be detected and the minimum circumscribed rectangular positioning frame of the sign to be detected.

[0085] S12, preprocessing the point cloud data of the road area to be detected to obtain the point cloud data of the area to be detected of the sign to be detected;

[0086] The process of obtaining the point cloud data of the area to be detected of the sign to be detected is:

[0087] The point cloud data of the road area to be detected is subjected to noise removal and screening processing to obtain pre-processed point cloud data, which includes the position and reflection intensity of the interfering objects and traffic signs;

[0088] According to the difference in reflection intensity between the interfering object and the signboard, the point cloud data of the area to be detected of the signboard to be detected is extracted from the point cloud data of the road area to be detected.

[0089] After the radar scans the road to be detected, the point cloud data is preprocessed to remove noise and point clouds that are too far or too close to obtain preprocessed point cloud data. The preprocessed point cloud data contains the position and reflection intensity information of the road surface, trees, buildings, other vehicles and traffic signs. The area to be detected of the sign to be detected can be extracted based on the difference in reflection intensity between objects such as the road surface and trees and the sign. The area to be detected of the sign to be detected is mapped to the pixel coordinate system by the obtained radar-vision fusion conversion matrix, and filtered based on the results obtained by the first deep learning detection model, so as to accurately obtain the category, relative distance, relative position and reflection intensity information of the sign to be detected. The relative distance uses the minimum or average value of the distance information in the sign or the distance value of the center point of the sign as the final distance information.

[0090] S13, mapping the point cloud data of the to-be-detected area of ​​the to-be-detected sign to the pixel coordinate system through a preset radar-visual fusion conversion matrix to obtain data of a second to-be-detected sign;

[0091] The laser-visual fusion conversion matrix is:

[0092]

[0093] in, represents the radar coordinates, ( u , v ) represents the pixel coordinates, Indicates the distance between the camera and the shooting point in the camera coordinate system. represents the camera intrinsic parameter matrix, R is the rotation matrix from radar to camera, and T is the translation matrix from radar to camera. The detailed derivation process of obtaining the radar-vision fusion transformation matrix is ​​shown in step 1-step 2.

[0094] S14. Filter the data of the second signboard to be detected according to the data of the first signboard to be detected to obtain the category of the signboard to be detected, the reflection intensity, and the relative distance and relative position from the detection vehicle.

[0095] S2, determining whether the relative distance between the sign to be detected and the detection vehicle meets the preset shooting distance. If so, triggering the detection module to take continuous photos of the sign to be detected with and without light source illumination to obtain two continuous frame images;

[0096] Step S2 also includes the steps of:

[0097] If the relative distance between the signboard to be detected and the inspection vehicle meets the preset shooting distance, the detection module at the corresponding position is triggered to control the light source to take continuous photos of the signboard to be detected with and without light illumination according to the relative position of the signboard to be detected and the inspection vehicle, and two continuous frame images are obtained. The relative position includes the detection signboard being above or to the right of the inspection vehicle, and the detection module includes an upper detection module and a right detection module.

[0098] When the preset shooting distance in this embodiment meets the distance requirement of 60 meters for measuring the retroreflective brightness coefficient of the sign, the detection area (the upper detection area or the right detection area) to which the sign belongs is determined according to the relative position of the sign, and the driving control module triggers the corresponding detection module to continuously take pictures of the current sign to be detected. Each time the detection module continuously obtains two images of the continuous frames of the current sign to be detected, the first image is a bright image when illuminated by a light source, and the second image is a dark image when there is no light source.

[0099] When the detection module is triggered to take continuous photos of the sign to be detected, according to the "Road Traffic Signs and Markings Manual: 2023 Edition", "Highway Network Traffic Sign Setting Technology" and other materials, it is clear that the lower edge of the cantilever, gantry, and elevated attached sign for highways, first- and second-class highways is not less than 5 meters from the road surface, and the third- and fourth-class highways are not less than 4.5 meters. The light source and camera of the detection module are installed on the detection vehicle at a specific angle. When the detection vehicle is traveling on the road to be detected, when the distance between the detection vehicle and the sign meets the shooting distance requirement (60 meters in this embodiment), the incident angle between the optical axis of the light source and the axis of the retroreflector of the sign is -4 degrees, and the observation angle between the optical axis of the light source and the optical axis of the camera is 0.2 degrees. Meet the measurement requirements for road traffic reflective film in national standards.

[0100] S3, inputting the two continuous frame images into the deep learning classification and recognition model respectively, and obtaining two target sign images with a plurality of color regions, color labels corresponding to each color region, grayscale values ​​and color region coordinates;

[0101] The deep learning classification and recognition model includes a sign recognition model and a multi-classification model. The specific process of step S3 is as follows:

[0102] S31, inputting two consecutive frame images into a sign recognition model respectively, identifying the sign category and the minimum bounding rectangle positioning frame in each consecutive frame image, and cutting out the sign area from the consecutive frame images according to the minimum bounding rectangle positioning frame to obtain two corresponding first target images with sign category identification;

[0103] The detection module can be used for image data collection. After preprocessing, the massive amount of data collected is used to construct a training data set, which is then input into a pre-built deep learning network to train a sign recognition model. There are three types of annotated signs: large, medium, and small. They are used to identify signs of different sizes, and the consistency and accuracy of the annotations must be ensured.

[0104] The pre-trained sign recognition model is used to infer and predict the two target images to obtain more accurate sign categories, minimum bounding rectangle positioning boxes and other information. The sign recognition model is the yolov8 model.

[0105] S32, inputting the two first target images into the multi-classification model respectively, and for each first target image, segmenting the first target image into a plurality of corresponding color regions according to the color label, to obtain two second target images with the plurality of color regions, the color labels corresponding to the color regions, and the color region coordinates;

[0106] The multi-classification model is a softmax multi-classification model based on a multi-layer perceptron (MLP). The massive signboards obtained by the detection module are automatically or manually cut out into different color areas through algorithms and unified into image data sets with consistent width and height. These image data sets are pre-processed by grayscale conversion, contrast adjustment, image enhancement, etc. to make them into training data sets. The training set, test set, and validation set are divided into a ratio of 8:1:1 to train the classification model. The third classification model can segment different color areas on the signboard and obtain color labels for the corresponding areas.

[0107] S33. Preprocess the two continuous frame images respectively, and use the preprocessed continuous frame images to calculate the average grayscale value corresponding to each color area in the second target image corresponding to each continuous frame image, and finally obtain two target signboard images corresponding to the two continuous frame images with a number of color areas, color labels corresponding to each color area, grayscale values ​​and color area coordinates.

[0108] S4, calculating the average grayscale value corresponding to each color label in each target sign image according to the color label and grayscale value corresponding to each color area;

[0109] The specific process of step S4 is:

[0110] S41, for each target sign image, dividing the target sign image into a plurality of even target areas;

[0111] S42, counting the color areas contained in each target area, and classifying the contained color areas according to color labels, and calculating the average grayscale value corresponding to the color areas with the same color label in each target area according to the grayscale value corresponding to each color area contained in each target area.

[0112] Specifically, the target sign image is evenly divided into 9 target areas. In each target area, each color label and color area and the corresponding grayscale value in the target area are traversed to obtain the average grayscale value of different color areas in each target area. After the sign is divided into multiple areas, the color label of each area is more accurate.

[0113] For example, Figure 4 As shown, for the first target area 1 of the target sign image, the target area 1 may include color area 1 with a red color label, color area 2 and color area 3 with blue color labels. In this way, when calculating the average grayscale value of different color areas in the target area, the color areas corresponding to the same color label are first counted. The red label has one color area, and the blue label has two color areas. The grayscale values ​​corresponding to each color area are obtained, and the average grayscale value corresponding to the red label and the average grayscale value corresponding to the blue label are calculated.

[0114] S5. Calculate the difference between the average grayscale values ​​corresponding to the same color label in the two target signboard images to obtain the interference-free grayscale value corresponding to each color label, input the interference-free grayscale value into the retroreflection conversion formula corresponding to each color label, and obtain the retroreflection brightness coefficient corresponding to each color label in the signboard to be detected.

[0115] The difference between the average grayscale values ​​of the corresponding color labels of the corresponding target areas in the two target signboard images is calculated to obtain the interference grayscale value corresponding to the same color label in each target area;

[0116] The non-interference grayscale value corresponding to each color label in each target area is respectively input into the retroreflection conversion formula corresponding to each color label to obtain the retroreflection brightness coefficient corresponding to each color label in each target area.

[0117] In order to avoid the influence of illumination and reflection on the measurement, the grayscale values ​​of the color areas calculated for each target area in the two target sign images are differentiated to obtain the grayscale value without interference under the condition of light source only. The grayscale value of each target area is converted into the retroreflection brightness coefficient by the retroreflection conversion formula obtained above. At this point, the retroreflection brightness coefficient corresponding to the color area in each target area is obtained after the sign is divided into 9 target areas. Finally, the control module stores the detection image and the detection result.

[0118] Among them, the retroreflection conversion formula is:

[0119] ax+by= 0

[0120] Among them, x represents the non-interference grayscale value corresponding to the color label to be calculated in each target area, represents the retroreflection brightness coefficient corresponding to the color label in the target area, and a and b represent the pre-fitted coefficients corresponding to each color label.

[0121] When the detection module captures the image of the sign, it will record the mileage information and GPS positioning information of the current detection vehicle, which can be converted into the pile number information of the current sign on the road. These location information and the detection image and detection results of the sign are recorded in groups. According to the mileage information, GPS information, and pile number information of the sign, the location, current status, and retroreflective brightness coefficient of each in-service sign can be determined. It provides more basic data for the management and maintenance of road traffic signs and indicates the maintenance direction.

[0122] Specifically, in the above process, the process of obtaining the radar-visual fusion conversion matrix is ​​as follows:

[0123] Step 1: Establish radar coordinate system and pixel coordinate system.

[0124] The origin of the radar coordinate system is the center point of the radar illumination window. The Z axis of the radar coordinate system is parallel to the length direction of the inspection vehicle, the X axis of the radar coordinate system is parallel to the width direction of the inspection vehicle, and the Y axis of the radar coordinate system is perpendicular to the length direction of the inspection vehicle. The pixel coordinate system refers to the upper left corner of the image as the origin, the horizontal right is the u axis, and the vertical downward is the v axis. Each point in the radar coordinate system corresponds to each point in the pixel coordinate system one by one, and the conversion between the radar coordinate system and the pixel coordinate system can be realized through a unique relationship.

[0125] Step 2: Radar and vision fusion calibration of observation module.

[0126] Solve the transformation relationship matrix between the pixel coordinate system and the radar coordinate system, and use the calibration plate to build the corresponding relationship between the camera and the radar. The solution steps include:

[0127] Step 2.1, solve the camera intrinsic parameter matrix.

[0128] The camera internal parameters are parameters related to the camera's own characteristics and are only related to the camera itself. Prepare a black and white checkerboard calibration plate with known spacing, fix the camera installation position, take complete images of the calibration plate at different angles, distances, and tilt angles, take at least 20 photos, and use Zhang Zhengyou's calibration method to determine the internal parameter matrix. The formula is as follows:

[0129]

[0130] in, is the distance between the camera and the shooting point in the camera coordinate system, and is the focal length of the camera on the x-axis and y-axis, is the pixel value of the horizontal translation between the center point of the image coordinate system and the pixel coordinate system, is the pixel value of the vertical translation between the center point of the image coordinate system and the pixel coordinate system, is the camera coordinate; (u, v) is the pixel coordinate, is the camera intrinsic parameter matrix;

[0131] Step 2.2: Radar-vision fusion calibration, solve the transformation matrix from pixel coordinate system to radar coordinate system.

[0132] Before calibration, the installation positions of the camera and radar must be fixed first, and the positions between them cannot be changed after the calibration begins. In this application, the corner points of the calibration plate are used as calibration targets. By taking advantage of the non-repetitive scanning of the radar and the high density of the point cloud, it is easy to find the exact position of the corner points of the calibration plate in the radar point cloud. Use the camera and radar to obtain image data and point cloud data of the calibration plate in different postures multiple times. Specifically:

[0133] When shooting, choose an open environment, fix the position of the camera and radar, and ensure that the field of view of the camera and radar point cloud are consistent. Constantly change the angle, distance, and position of the calibration plate (the distance should not be too close when shooting, and the distance between the radar and the calibration plate should be more than 3 meters). For each posture of the calibration plate, an image data and a set of point cloud data need to be captured. Select at least 10 positions with different angles and distances to capture images of the calibration plate. The positions should be distributed as evenly as possible, and different tilt angles of left and right, up and down should be collected.

[0134] Obtain the pixel coordinates of the four corner points in each pose calibration plate from the image data, and obtain the radar coordinates of the four corner points in the corresponding pose calibration plate from the point cloud data. Ensure that the order of the corner point positions obtained from the image data and the radar data is consistent. Group and record the corner point pixel coordinates and radar point cloud coordinates of the corresponding corner points in each image, and obtain the original data for the mutual conversion between pixel coordinates and radar coordinates.

[0135] Solve the rotation matrix and translation matrix from the camera coordinate system to the radar coordinate system. The conversion formula is:

[0136]

[0137] in, represents the radar coordinates, represents pixel coordinates, is the distance between the camera and the shooting point in the camera coordinate system, is the camera internal parameter matrix, R is the rotation matrix from radar to camera, T is the translation matrix from radar to camera, is the camera extrinsic parameter matrix, is the transformation matrix from the pixel coordinate system to the radar coordinate system. Based on the obtained rotation and translation matrix, the residual and reprojection error are calculated, the abnormal data that introduces large errors are removed, and the rotation matrix and translation matrix after the corrected data are recalculated.

[0138] According to the camera intrinsic parameter matrix and the extrinsic parameter matrix Get the matrix , which is the transformation matrix between the pixel coordinate system and the radar coordinate system.

[0139] After the camera, lens and radar parameters of the observation module are determined, only one calibration is required. If the aforementioned device selection remains unchanged, the above calibration results are pre-stored in the control module and called when needed.

[0140] It should be understood that this embodiment uses the grayscale value of the image to estimate the retroreflection brightness coefficient of the sign, and it is necessary to calibrate the retroreflection conversion formula. Specifically, in order to obtain the retroreflection conversion formula, a corresponding relationship between the grayscale value of the sign in the two-dimensional image and the actual retroreflection brightness coefficient of the sign is established. After the grayscale value of the sign in the image is obtained, it can be converted into the retroreflection brightness coefficient of the sign through the corresponding relationship. Retroreflection calibration is performed in the following steps:

[0141] (1) Prepare signboards with various colors and known retroreflective brightness coefficients as calibration plates. There are two ways to obtain the retroreflective brightness coefficient of a signboard. One is that the signboard will have a rated reference value when it leaves the factory. The other is that it can be measured using a handheld measuring device. According to industry standards, the retroreflective brightness coefficient is generally measured at three points in different areas of the same color on the signboard and the average of the three points is calculated as the retroreflective brightness coefficient of the color on the signboard.

[0142] (2) Build the signboard according to the relevant specifications for the installation of road signs, ensure that the inspection vehicle and the signboard are at a specific distance. The specific distance used in the present invention is 60 meters. Fine-tune the vertical height and horizontal distance of the signboard, take pictures of the signboard in different positions, and calculate the average grayscale value of different colors on the signboard in the image. Record the average grayscale value of different colors and the known retroreflection brightness coefficient, and obtain the conversion expression.

[0143] (3) Use the least squares method to fit the expression.

[0144] The retroreflection brightness coefficient is regarded as the dependent variable, and the corresponding average gray value is regarded as the independent variable. The problem is transformed into solving a set of two-variable linear equations:

[0145]

[0146] The following solution can be obtained:

[0147]

[0148] Here, a represents the slope and b represents the intercept.

[0149] After fitting the linear regression model using the least squares method, the coefficient of determination was used To measure the degree of fit between the predicted result data and the actual data. ‌ is a statistic used to measure the degree of fit of the model to the data, indicating the proportion of the variation explained by the model to the total variation, and its value is between 0 and 1. A higher A high value means that the model is able to better explain the variation in the data, while a low A value of indicates that the model does not fit the data well.

[0150] when = 1, it means that the model fits the data perfectly, that is, the model prediction value is completely consistent with the actual observed value. When <1, it means that the model fits the data to a certain extent, but there is still unexplained variation. = 0, indicating that the model cannot explain any variation in the data, that is, there is no correlation between the model predictions and the actual observed values.

[0151] The calculation formula is:

[0152]

[0153] in, represents the model prediction value, Represents the true value of the data, represents the average value of the real data, and n represents n sets of data. In this embodiment, the model obtained is ≈1, that is, the predicted value given by the model fitted using the least squares method is completely consistent with the true value.

[0154] When the camera and lens parameters of the measurement module are determined, only one calibration is required. If the aforementioned device selection remains unchanged, the above calibration results are stored in the control module and called when needed. And a set of coefficients a and b corresponding to each color label are calculated and stored in the system and called when calculating.

[0155] Example 3

[0156] The difference between this embodiment and embodiment 2 is that in step S2, when the detection module is triggered to continuously take photos of the sign to be detected with and without light source illumination, the delayed shooting is used to improve the efficiency and accuracy of the measurement of the retroreflective brightness coefficient of the sign. The triggering process is:

[0157] S21, setting the length L1 of the interval to be photographed, and obtaining the distance L4 to be photographed according to the sum of the length L1 of the interval to be photographed and the preset shooting distance L;

[0158] S22, determine whether the relative distance L2 between the sign to be detected and the detection vehicle satisfies the requirement of being less than the distance to be photographed L4 and greater than the preset photographing distance L. If so, record the road mileage x1 of the detection vehicle traveled obtained from the mileage measurement module at this time.

[0159] S23, calculating the difference between the relative distance L2 in step S22 and the preset shooting distance L to obtain the delayed shooting distance L3;

[0160] S24, obtaining the road mileage x2 of the inspection vehicle measured by the mileage measurement module in real time, calculating the difference between the road mileage x2 and x1, and obtaining the mileage △x;

[0161] S25. When the driving mileage △x reaches the delayed shooting distance L3, the detection module is triggered to take continuous photos of the sign to be detected with and without light source illumination to obtain two continuous frame images.

[0162] In this embodiment, the detection module is installed at a fixed height on the detection vehicle (the installation position is 2.01 meters above the ground), and the shooting distance L that meets the requirements for taking pictures is 60 meters. When the vehicle speed is fast, in order to ensure the real-time shooting and the timeliness of the detection, the waiting shooting interval of the detection module is set. The length of the waiting shooting interval is related to the vehicle speed. The faster the vehicle speed, the longer the distance to be shot should be. For example, when the vehicle speed is about 100km / h, it is better to set the length of the waiting shooting interval L1 to 10 meters. Figure 5As shown, the distance to be photographed L4 = the length of the interval to be photographed L1 + the shooting distance L. When the inspection vehicle travels in the interval to be photographed, the relative straight-line distance between the inspection vehicle and the sign S to be inspected can be calculated according to the observation module, and the delayed shooting distance L3 can be obtained by subtracting the set shooting distance L. The control module monitors through the mileage measurement module that the inspection vehicle has traveled the delayed shooting distance L3 since the delayed trigger task was received and arrived at the shooting point P, and then triggers the inspection module to take a photo.

[0163] For example, at a certain moment, the inspection vehicle is traveling in the area to be photographed. The control module calculates that the signboard of the road area to be inspected is located above the road and the distance between the inspection vehicle and the signboard is less than the distance to be photographed L4. The inspection vehicle is now located at the task trigger point P1. The delayed photographing distance is calculated by the formula L3=L2-L. The control module then sets the control task to control the upper inspection module to photograph the signboard when the inspection vehicle arrives at the shooting point P.

[0164] Each shot of the detection module will obtain two images of the continuous frame of the current sign to be detected. The first image is a bright image when illuminated by a light source, and the second image is a dark image when there is no light source.

[0165] It is understandable that this application has promoted the detection of retroreflective brightness coefficient of road traffic signs into the stage of automated detection. This system completely realizes non-contact measurement of traffic signs. When measuring, you only need to turn on the detection system and follow simple driving requirements to achieve real-time detection of road signs in service. Compared with manual direct contact measurement, it greatly improves the detection efficiency and saves a lot of time and labor costs.

[0166] First, the deep learning yolov8 algorithm is used to identify the signboards in the area to be inspected. Then, combined with the radar point cloud data, the type, relative position and other information of the signboards can be extracted more efficiently and accurately. The radar-vision fusion detection method has strong adaptability to complex detection environments and can adapt to situations such as signboard occlusion, defects, and dirt.

[0167] Furthermore, the type of signboards that need to be detected can be set or excluded based on the type and location information of the signboards, making the detection more flexible, efficient and user-friendly. In addition, shooting continuous frame images of the signboards and numerically suppressing them can effectively avoid interference with the detection of the reverse brightness coefficient caused by direct sunlight or various reflected light. The deep learning classification model is used to divide the color of the signboards, which improves the detection precision and accuracy compared to the traditional color extraction algorithm. The system is easy to use, simple to operate, and has a fast detection speed.

[0168] Those skilled in the art will appreciate that the present invention may be implemented in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0169] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0170] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0171] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0172] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. According to the technical essence of the present invention, within the spirit and principles of the present invention, any simple modification, equivalent replacement and improvement made to the above embodiment still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign, characterized in that: The specific steps include: S1. Obtain image data and point cloud data of the road area to be detected taken by the observation module on the top of the detection vehicle, input the image data of the road area to be detected into a pre-trained first deep learning detection model, and combine the point cloud data to obtain the relative distance between the sign to be detected in the road area to be detected and the detection vehicle; S2, determining whether the relative distance between the sign to be detected and the detection vehicle meets the preset shooting distance. If so, triggering the detection module to take continuous photos of the sign to be detected with and without light source illumination to obtain two continuous frame images; S3, inputting the two continuous frame images into the deep learning classification and recognition model respectively, and obtaining two target sign images with a plurality of color regions, color labels corresponding to each color region, grayscale values ​​and color region coordinates; The deep learning classification and recognition model includes a sign recognition model and a multi-classification model. The specific process of step S3 is as follows: S31, inputting two consecutive frame images into a sign recognition model respectively, identifying the sign category and the minimum bounding rectangle positioning frame in each consecutive frame image, and cutting out the sign area from the consecutive frame images according to the minimum bounding rectangle positioning frame to obtain two corresponding first target images with sign category identification; S32, inputting the two first target images into the multi-classification model respectively, and for each first target image, segmenting the first target image into a plurality of corresponding color regions according to the color label, to obtain two second target images with the plurality of color regions, the color labels corresponding to the color regions, and the color region coordinates; S33, preprocessing the two continuous frame images respectively, and using the preprocessed continuous frame images, calculating the average grayscale value corresponding to each color area in the second target image corresponding to each continuous frame image, and finally obtaining two target sign images corresponding to the two continuous frame images with a plurality of color areas, color labels corresponding to each color area, grayscale values ​​and color area coordinates; S4, calculating the average grayscale value corresponding to each color label in each target sign image according to the color label and grayscale value corresponding to each color area; S5. Calculate the difference between the average grayscale values ​​corresponding to the same color label in the two target signboard images to obtain the interference-free grayscale value corresponding to each color label, input the interference-free grayscale value into the retroreflection conversion formula corresponding to each color label, and obtain the retroreflection brightness coefficient corresponding to each color label in the signboard to be detected.

2. A method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign according to claim 1, characterized in that: The specific process of step S1 is: S11, inputting the image data of the road area to be detected into the first deep learning detection model to obtain the first signboard data to be detected with the attribute category and the minimum circumscribed rectangular positioning frame label; S12, preprocessing the point cloud data of the road area to be detected to obtain the point cloud data of the area to be detected of the sign to be detected; S13, mapping the point cloud data of the to-be-detected area of ​​the to-be-detected sign to the pixel coordinate system through a preset radar-visual fusion conversion matrix to obtain data of a second to-be-detected sign; S14. Filter the data of the second signboard to be detected according to the data of the first signboard to be detected to obtain the category of the signboard to be detected, the reflection intensity, and the relative distance and relative position from the detection vehicle.

3. A method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign according to claim 2, characterized in that: Step S2 also includes the steps of: If the relative distance between the signboard to be detected and the inspection vehicle meets the preset shooting distance, the detection module at the corresponding position is triggered to control the light source to take continuous photos of the signboard to be detected with and without light illumination according to the relative position of the signboard to be detected and the inspection vehicle, and two continuous frame images are obtained. The relative position includes the detection signboard being above or to the right of the inspection vehicle, and the detection module includes an upper detection module and a right detection module.

4. A method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign according to claim 2, characterized in that: The specific process of step S2 triggering the detection module to take a photo of the sign to be detected is as follows: S21, setting the length L1 of the interval to be photographed, and obtaining the distance L4 to be photographed according to the sum of the length L1 of the interval to be photographed and the preset shooting distance L; S22, determine whether the relative distance L2 between the sign to be detected and the detection vehicle satisfies the requirement of being less than the distance to be photographed L4 and greater than the preset photographing distance L. If so, record the road mileage x1 of the detection vehicle traveled obtained from the mileage measurement module at this time. S23, calculating the difference between the relative distance L2 in step S22 and the preset shooting distance L to obtain the delayed shooting distance L3; S24, obtaining the road mileage x2 of the inspection vehicle measured by the mileage measurement module in real time, calculating the difference between the road mileage x2 and x1, and obtaining the mileage △x; S25. When the driving mileage △x reaches the delayed shooting distance L3, the detection module is triggered to take continuous photos of the sign to be detected with and without light source illumination to obtain two continuous frame images.

5. A method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign according to claim 1, characterized in that: The specific process of step S4 is: S41, for each target sign image, dividing the target sign image into a plurality of even target areas; S42, counting the color areas contained in each target area, and classifying the contained color areas according to color labels, and calculating the average grayscale value corresponding to the color areas with the same color label in each target area according to the grayscale value corresponding to each color area contained in each target area.

6. A method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign according to claim 5, characterized in that: The specific process of step S5 is: The difference between the average grayscale values ​​of the corresponding color labels of the corresponding target areas in the two target signboard images is calculated to obtain the interference grayscale value corresponding to the same color label in each target area; The non-interference grayscale value corresponding to each color label in each target area is respectively input into the retroreflection conversion formula corresponding to each color label to obtain the retroreflection brightness coefficient corresponding to each color label in each target area.

7. A method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign according to claim 5, characterized in that: The retroreflection conversion formula is: ax+by= 0； Among them, x represents the non-interference grayscale value corresponding to the color label to be calculated in each target area, represents the retroreflection brightness coefficient corresponding to the color label in the target area, and a and b represent the pre-fitted coefficients corresponding to each color label.

8. A method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign according to claim 2, characterized in that: The radar-vision fusion conversion matrix in step S13 is: (X L , Y L , Z L ) represents radar coordinates, (u, v) represents pixel coordinates, Z C It represents the distance between the camera and the shooting point in the camera coordinate system, M1 represents the camera intrinsic parameter matrix, R is the rotation matrix from the radar to the camera, and T is the translation matrix from the radar to the camera.

9. A vehicle-mounted sign retroreflective brightness coefficient detection system, characterized in that: It includes an observation module, a detection module and a mileage measurement module which are respectively connected to the control module in communication, wherein: The observation module includes a camera and a radar. The optical axes of the camera and the radar are installed on the roof of the inspection vehicle parallel to the length direction of the inspection vehicle to capture images of the road area to be inspected. The detection module includes a detection camera and a light source for lighting, and the detection camera and the light source for lighting are installed on the roof of the detection vehicle at a set angle to shoot the sign to be detected on the road area to be detected; The mileage measurement module is installed on the left rear wheel hub of the inspection vehicle to measure the road mileage traveled by the inspection vehicle; The control module is used to execute a method for detecting the retroreflective brightness coefficient of a vehicle-mounted sign as described in any one of claims 1-8.

Images