Vehicle-mounted highway guardrail beam plate bolt missing rapid detection device

Abstract

The invention provides a vehicle-mounted highway guardrail beam plate bolt missing rapid detection device, which comprises a mounting mechanism, a camera system, a light supplementing lamp system and a control system, and is characterized in that the mounting mechanism is used for being connected with a vehicle, and the camera system is connected to the mounting mechanism and performs data transmission with the control system; the camera system is used for collecting guardrail images, the light supplementing lamp system is connected to the mounting mechanism and used for supplementing light for the camera system, and the control system is used for analyzing and processing guardrail image data collected by the camera system and meanwhile used for controlling the camera system and the light supplementing lamp system. The device is used for detecting the bolt missing condition of the road guardrail, the detection efficiency and speed can be greatly improved, road traffic is prevented from being blocked, meanwhile, comprehensive detection of the guardrail is achieved, and the situation that the error of a sampling detection result is large is avoided.

Description

Technical Field

[0001] The invention relates to the technical field of traffic safety facility detection, and in particular to a vehicle-mounted highway guardrail beam plate bolt missing rapid detection device. Background Art

[0002] Corrugated steel guardrails are widely used in highway projects. They are composed of corrugated beam plates, columns, ends, splicing bolts, connecting bolts, barrier blocks, brackets, cross beams and other components. Corrugated beam plates and columns are two important components of corrugated steel guardrails, which have a very important impact on the mechanical properties of the guardrails and the normal performance of their protective functions. Since the guardrails are set along the longitudinal direction of the highway, their length is often long, so the corrugated beam plates need to be connected by splicing bolts to form a complete corrugated beam plate system. At the same time, the corrugated beam plates and guardrails also need to be connected. The connection between the corrugated beam plates and the guardrails mainly depends on the connecting bolts. Whether it is a splicing bolt or a connecting bolt, if it becomes loose or falls off during use, it will seriously affect the connection between the corrugated beam plates of the guardrail and between them and the columns, thereby affecting the anti-collision performance of the entire guardrail, posing a serious threat to traffic safety.

[0003] Therefore, during the operation of the highway, it is necessary to check and count the missing guardrail bolts so as to supplement and improve them in time and ensure the mechanical properties of the guardrail. However, due to the large number of splicing bolts and connecting bolts in the guardrail, it is often time-consuming and laborious to check them one by one. Therefore, in the actual inspection and road maintenance process, the inspectors generally randomly select 5 beams with bolts per kilometer for inspection, and obtain the missing bolts at the selected locations by visual inspection. The biggest drawback of the sampling mode is that the number of selected samples is small and not representative, while selecting more samples will make the inspection work time-consuming, labor-intensive and inefficient. Moreover, the bolts are relatively small in size compared to the entire guardrail system. In addition, due to the limitations of the special structure of the corrugated beam plate, it is also difficult to obtain the missing bolts of the entire highway guardrail by other means.

[0004] It can be seen that the current detection of missing bolts of highway corrugated steel guardrails mainly has the following problems: 1. Manual detection is inefficient, the detection time is long and the number of samples is small. Traffic needs to be closed during detection, which affects the normal traffic of the road; 2. It is impossible to achieve fast and accurate detection, especially in the process of highway detection, the emergency lane needs to be occupied, which will pose a threat to the personal safety of the inspectors; 3. There are no tools and methods specifically for the rapid detection of missing guardrail bolts in the existing technology. Most of the means of bolt missing detection are based on drones and are mainly used in the field of power inspection, which is not applicable in the field of highway engineering. Therefore, it is urgent to propose a device and method that can efficiently and safely detect the missing bolts of corrugated steel guardrails. Summary of the invention

[0005] In view of the low efficiency and potential safety hazards of traditional manual inspection of missing bolts of highway corrugated steel guardrails, the present invention proposes a vehicle-mounted rapid detection device for missing guardrail beam bolts. The device can quickly detect missing guardrail bolts based on vehicle-mounted equipment without affecting the normal passage of vehicles on the road, and can effectively ensure the detection efficiency and the safety of the inspectors.

[0006] In order to solve the above technical problems, the present invention is implemented by the following technical solutions: a vehicle-mounted highway guardrail beam plate bolt missing rapid detection device, including a mounting mechanism, a camera system, a fill light system, and a control system, wherein the mounting mechanism is used to connect with the vehicle, the camera system is connected to the mounting mechanism and transmits data between the control system, the camera system is used to realize the collection of guardrail images, the fill light system is connected to the mounting mechanism to fill light for the camera system, and the control system is used to analyze and process the guardrail image data collected by the camera system, and is also used to control the camera system and the fill light system. In the actual application process, the vehicle-mounted highway guardrail beam plate bolt missing rapid detection device is first fixedly installed on the side of the vehicle close to the guardrail through the mounting mechanism, and then the vehicle is driven normally on the road section to be detected, the control system controls the camera system to start and start collecting guardrail image data, the camera system transmits the collected guardrail image data to the control system, and the control system controls the start and shut down of the fill light system according to the guardrail image information. After completing the detection of all guardrails, the control system outputs the missing status of the corrugated steel guardrail bolts in the detected road section.

[0007] Furthermore, the mounting mechanism also includes a height adjustment mechanism and a mounting platform, one end of the height adjustment mechanism is fixedly connected to the mounting platform, and the other end of the height adjustment mechanism is fixedly connected to the vehicle. The height adjustment mechanism can adjust the installation height of the mounting platform on the vehicle, so that the vehicle-mounted highway guardrail beam and plate bolt missing rapid detection device can adapt to vehicles of different heights.

[0008] Furthermore, the camera system includes a camera angle adjustment mechanism and a high dynamic camera. The camera angle adjustment mechanism is fixedly installed above the mounting platform, and the high dynamic camera is rotatably connected to the camera angle adjustment mechanism. By setting the camera angle adjustment mechanism, the camera angle can be adjusted so that the camera can adapt to different road conditions and image acquisition requirements of different beam and slab type guardrails. The sampling frequency of the high dynamic camera is 25Hz, and the sampling interval is 2m. ， The high dynamic camera captures images with a field of view of 1.4×0.7m.

[0009] Furthermore, the fill light system includes a fill light angle adjustment mechanism and a fill light. The fill light angle adjustment mechanism is fixedly installed under the mounting platform, and the fill light is rotatably connected to the fill light angle adjustment mechanism. By setting the fill light angle adjustment mechanism, the angle of the fill light can be adjusted to better match the camera, so that the fill light can better meet the fill light requirements of high dynamic cameras.

[0010] Furthermore, the control system includes a host computer, a slave computer, a GPS, a relay, and a distance sensor. The host computer is composed of data acquisition software and data processing software. The high-dynamic camera acquires data in real time and sends the image to the host computer through Ethernet; the GPS sends the positioning information to the slave computer through the serial port, and the distance sensor sends the distance data to the slave computer through the I / O port. After the data is shaped by the slave computer, the slave computer sends the positioning information and distance information to the host computer through Ethernet; the host computer combines the brightness information of the image and sends the light-on command to the slave computer, and the slave computer controls the opening and closing of the fill light through the relay.

[0011] The host computer combines traditional vision and deep learning vision in terms of visual algorithms. The traditional vision part is responsible for image preprocessing, removing noise from high-dynamic cameras, water mist in the environment, suppressing backlight, and adjusting image brightness; the deep learning part is responsible for the positioning detection of guardrail beams and bolts.

[0012] The guardrail beam detection network and guardrail post detection network are trained using the shallow network of YOLOV5, and the 16-bit inference engine of TensorRT is used for inference. The bolt detection network is trained using the deep network of YOLOV5, and the 16-bit inference engine of TensorRT is used for inference. The guardrail posts detected are checked for duplicates, and the intercepted images of the same post are packaged into a data packet and put into the queue to be analyzed. Each data packet in the queue is detected using the guardrail bolt detection network, and the column interception image with the most guardrail bolt detection results is determined to be the bolt detection result of the column, including bolts, missing holes, and contour marks.

[0013] The detailed steps for the host computer to implement detection are:

[0014] A. Image preprocessing, i.e., removing noise, adjusting image length and width, defogging dark channels, adjusting image brightness, image normalization, and image channel adjustment;

[0015] Image B is copied from memory to video memory to infer the guardrail beam and column detection network models, and non-maximum suppression is performed on the detection results, while the detection results at the edge of the image are eliminated;

[0016] C removes duplicates from the detected pillars. The specific operation is as follows: record the car mileage D1 of the first detected pillar, located at column coordinate X1 in the image, and record the car mileage D of the subsequent detected pillars. M , located at column coordinate X in the image M , the value range of M is 2 to 100.

[0017] When the pillar on the left side of the car satisfies formula (C-1) and formula (C-2), they are considered to be the same pillar; when the pillar on the right side of the car satisfies formula (C-1) and formula (C-3), they are considered to be the same pillar.

[0018] D M -D1<2(C-1)

[0019] X M -X1≤100(C-2)

[0020] X M -X1≥-100(C-3)

[0021] The intercepted images of the same column are packaged into a data packet and put into a queue to be analyzed;

[0022] D copies the intercepted images in the data packets in the queue to be analyzed from the memory to the video memory for bolt detection model inference, and performs non-maximum suppression processing on the detection results, while eliminating the detection results at the edge of the image, and identifies the column intercepted image with the most guardrail bolt detection results as the bolt detection result of the column;

[0023] E obtains the bolt inspection result, which includes the inspection data of bolts, missing holes and contour marks;

[0024] F calculates the missing rate, where missing rate = number of missing holes / (number of bolts + number of missing holes + number of contour marks).

[0025] Among them, the upper computer adopts an industrial computer or a personal computer, and the lower computer adopts a PLC or an MCU.

[0026] The beneficial effects of the present invention are:

[0027] 1. Effectively improve the detection efficiency. This device adopts vehicle-mounted equipment. The installation mechanism, camera system, fill light system, and control system are all installed on the vehicle. During the detection process, it will not affect the normal traffic of the road. During the detection process, there is no need to touch the guardrail or perform manual on-site measurement. The detection speed can reach up to 90km / h;

[0028] 2. It can realize the full inspection of guardrails. The traditional method of detecting missing guardrail bolts is random inspection, which needs to balance the detection efficiency and accuracy. However, since the length of guardrails is often long, but the location of missing bolts is uncertain, it is impossible to realize the full coverage inspection of guardrail bolts, and some guardrails with missing bolts may be missed during the inspection process. However, this device is based on machine vision and image recognition, which can realize the sampling and analysis of all guardrails in the road section to be inspected without consuming much time.

[0029] 3. It can better adapt to highway guardrails. This device uses a vehicle as a carrier and installs the device on the side of the vehicle, and a certain distance is maintained between the vehicle and the guardrail. At the same time, the height and angle of the high-dynamic camera and the fill light can be adjusted through the height adjustment mechanism and the angle adjustment mechanism, so that the device as a whole can be suitable for various levels of highways. For example, when testing on a highway, since there is an emergency lane on the side of the road, the distance between the device and the guardrail is large. At this time, the adjustable range of the device installation height and the camera installation angle is large, which can better guarantee the size of the high-dynamic camera image acquisition; when testing on ordinary national highways, provincial highways, municipal roads or rural roads, since there is generally no emergency lane, the distance between the device and the guardrail is close. At this time, by reasonably adjusting the height of the device to keep it consistent with the height of the guardrail beam, and then reasonably adjusting the angle of the high-dynamic camera and the fill light, the size and quality of the high-dynamic camera image acquisition can be effectively guaranteed. Compared with the method of using drones for image acquisition, due to the special structure of the corrugated guardrail, the bolts are generally installed in the recessed part of the guardrail beam. The use of drones for shooting is blocked by trees, vehicles and the guardrail itself, which is not conducive to the acquisition of guardrail images. Since this device acquires images along the guardrail and is closer to the guardrail, the continuity of image acquisition is better, and the acquired images are also easier to analyze and process.

[0030] 3. Reduce the number of inspectors. Traditional inspection methods require the arrangement of multiple on-site inspectors, while this device only requires one inspector and one driver to carry out the inspection. The inspector can complete the on-site inspection by operating the supporting software on the vehicle, and the inspector can conduct the inspection on the vehicle, which effectively avoids the risk factors of on-site inspection and better protects the personal safety of the inspectors.

[0031] 4. Higher detection accuracy. The traditional manual sampling method can only estimate the missing bolts of the guardrail in the sampling section. This device can obtain the detection results of the missing bolts of the guardrail more accurately by comprehensively sampling the guardrail and including a deduplication algorithm. It has been verified that the accuracy of the missing bolt detection of this device can reach 95%.

[0032] 5. Easy to review. Since the device collects images of all guardrails on the road section to be tested, the abnormal detection results that appear in some detection processes can be manually reviewed again. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is an overall structural diagram of the vehicle-mounted highway guardrail beam plate bolt missing rapid detection device of the present invention;

[0034] Figure 2 It is a front view of the vehicle-mounted highway guardrail beam plate bolt missing rapid detection device of the present invention;

[0035] Figure 3 It is a side view of the vehicle-mounted highway guardrail beam plate bolt missing rapid detection device of the present invention;

[0036] Figure 4 is a flow chart of the visual algorithm of the present invention;

[0037] Figure 5 It is a system structure diagram of the present invention;

[0038] Figure 6 It is the overall detection result output by the host computer of the present invention;

[0039] Figure 7 It is the detection result of the host computer according to the pile number statistics of the present invention;

[0040] Figure 8 is the guardrail image after identification of the present invention;

[0041] Fig. 9 It is the bolt identification situation in the acquisition process of the present invention.

[0042] Among them: 1-installation mechanism; 2-camera system; 3-fill light system; 4-height adjustment mechanism; 5-installation platform; 6-camera angle adjustment mechanism; 7-high dynamic camera; 8-fill light angle adjustment mechanism; 9-fill light. DETAILED DESCRIPTION

[0043] In order to deepen the understanding of the present invention, the present invention will be further described in detail below in conjunction with examples. The examples are only used to explain the present invention and do not constitute a limitation on the protection scope of the present invention.

[0044] Example

[0045] like Figure 1-6 As shown, this embodiment proposes a vehicle-mounted highway guardrail beam plate bolt missing rapid detection device, including a mounting mechanism 1, a camera system 2, a fill light system 3, and a control system, wherein the mounting mechanism is used to connect with the vehicle, the camera system is connected to the mounting mechanism and transmits data to the control system, the camera system is used to realize the collection of guardrail images, the fill light system is connected to the mounting mechanism to fill light for the camera system, and the control system is used to analyze and process the guardrail image data collected by the camera system, and is also used to control the camera system and the fill light system. In the actual application process, the vehicle-mounted highway guardrail beam plate bolt missing rapid detection device is first fixedly installed on the side of the vehicle close to the guardrail through the mounting mechanism, and then the vehicle is driven normally on the road section to be detected, the control system controls the camera system to start and start collecting guardrail image data, the camera system transmits the collected guardrail image data to the control system, and the control system controls the start and shut down of the fill light system according to the guardrail image information. After completing the detection of all guardrails, the control system outputs the missing status of the corrugated steel guardrail bolts in the detected road section.

[0046] Furthermore, the mounting mechanism also includes a height adjustment mechanism 4 and a mounting platform 5, one end of the height adjustment mechanism is fixedly connected to the mounting platform, and the other end of the height adjustment mechanism is fixedly connected to the vehicle, and the height adjustment mechanism can adjust the mounting height of the mounting platform on the vehicle, so that the vehicle-mounted highway guardrail beam plate bolt missing rapid detection device can adapt to vehicles of different heights. The height adjustment mechanism can adopt an electric cylinder, a telescopic rod, etc.

[0047] Furthermore, the camera system includes a camera angle adjustment mechanism 6 and a high dynamic camera 7. The camera angle adjustment mechanism is fixedly installed above the mounting platform, and the high dynamic camera is rotatably connected to the camera angle adjustment mechanism. By setting the camera angle adjustment mechanism, the angle of the camera can be adjusted so that the camera can adapt to the image acquisition requirements of different road conditions and different beam and slab type guardrails. The camera angle adjustment mechanism 6 includes a U-shaped block, a rotating plate, a rotating rod, a guide rod, and a nut. Two vertical plates are symmetrically arranged on both sides of the U-shaped block. The vertical plates are provided with arc-shaped guide grooves and circular rotating grooves. The two rotating plates are fixedly connected to both sides of the high-dynamic camera 7, so that the high-dynamic camera 7 is clamped between the two vertical plates. One end of the two rotating rods is fixedly connected to the corresponding rotating plates, and the other ends of the two rotating rods are respectively inserted into the circular rotating grooves of the vertical plates and are rotatably connected to the vertical plates. One end of the two guide rods is fixedly connected to the corresponding rotating plates, and the other ends of the two guide rods are inserted into the arc-shaped guide grooves of the vertical plates and are slidably connected to the arc-shaped guide grooves. The guide rods are screw rods. After the angle of the high-dynamic camera 7 is adjusted to the right position, the position of the guide rods is fixed by nuts, thereby fixing the position and angle of the high-dynamic camera 7.

[0048] Furthermore, the fill light system includes a fill light angle adjustment mechanism 8 and a fill light 9. The fill light angle adjustment mechanism is fixedly installed under the mounting platform, and the fill light is rotatably connected to the fill light angle adjustment mechanism. By setting the fill light angle adjustment mechanism, the angle of the fill light can be adjusted to better match the camera, so that the fill light can better meet the fill light requirements of the high dynamic camera. The sampling frequency of the high dynamic camera is 25Hz, and the sampling interval is 2m. ， The image size collected by the high-dynamic camera is 1.4×0.7m. According to the requirements of the guardrail design specification, the spacing between adjacent columns of the corrugated beam guardrail is 1m, 1.5m, 2m, 3m or 4m. If the camera sampling spacing is 1m or 1.5m, a sampling spacing that is too short will result in too much guardrail image data collected, which will reduce the efficiency of image processing. If the camera sampling spacing is set to 3m or 4m, a sampling spacing that is too long will cause some guardrail images to be missed, and full coverage detection cannot be achieved. Therefore, the sampling spacing of the high-dynamic camera is determined to be the median of the column spacing, 2m, which can not only ensure full coverage of image acquisition, but also appropriately reduce the amount of image data, which is convenient for analysis and processing.

[0049] Furthermore, the control system includes a host computer, a slave computer, a GPS, a relay, and a distance sensor. The host computer is composed of data acquisition software and data processing software. The high-dynamic camera acquires data in real time and sends the image to the host computer through Ethernet; the GPS sends the positioning information to the slave computer through the serial port, and the distance sensor sends the distance data to the slave computer through the I / O port. After the data is shaped by the slave computer, the slave computer sends the positioning information and distance information to the host computer through Ethernet; the host computer combines the brightness information of the image and sends the light-on command to the slave computer, and the slave computer controls the opening and closing of the fill light through the relay.

[0050] Among them, the upper computer adopts an industrial computer or a personal computer, and the lower computer adopts a PLC or an MCU.

[0051] The host computer combines traditional vision and deep learning vision in terms of visual algorithms. The traditional vision part is responsible for image preprocessing, removing noise from high-dynamic cameras, water mist in the environment, suppressing backlight, and adjusting image brightness; the deep learning part is responsible for the positioning detection of guardrail beams and bolts.

[0052] The guardrail beam detection network and guardrail post detection network are trained using the shallow network of YOLOV5, and the 16-bit inference engine of TensorRT is used for inference. The bolt detection network is trained using the deep network of YOLOV5, and the 16-bit inference engine of TensorRT is used for inference. The guardrail posts detected are checked for duplicates, and the intercepted images of the same post are packaged into a data packet and put into the queue to be analyzed. Each data packet in the queue is detected using the guardrail bolt detection network, and the column interception image with the most guardrail bolt detection results is determined to be the bolt detection result of the column, including bolts, missing holes, and contour marks.

[0053] The detailed steps for the host computer to implement detection are:

[0054] A. Image preprocessing, i.e., removing noise, adjusting image length and width, defogging dark channels, adjusting image brightness, image normalization, and image channel adjustment;

[0055] Image B is copied from memory to video memory to infer the guardrail beam and column detection network models, and non-maximum suppression is performed on the detection results, while the detection results at the edge of the image are eliminated;

[0056] C removes duplicates from the detected pillars. The specific operation is as follows: record the car mileage D1 of the first detected pillar, located at column coordinate X1 in the image, and record the car mileage D of the subsequent detected pillars. M , located at column coordinate X in the image M , the value range of M is 2 to 100.

[0057] When the pillar on the left side of the car satisfies formula (C-1) and formula (C-2), they are considered to be the same pillar; when the pillar on the right side of the car satisfies formula (C-1) and formula (C-3), they are considered to be the same pillar.

[0058] D M -D1<2(C-1)

[0059] X M -X1≤100(C-2)

[0060] X M -X1≥-100(C-3)

[0061] The intercepted images of the same column are packaged into a data packet and put into a queue to be analyzed;

[0062] For the above-mentioned column deduplication step, the device samples by triggering a high-dynamic camera with a distance, and the sampling interval is 2m. Formula (C-1) is used to calculate the difference in vehicle mileage to determine whether the captured image is within the same sampling interval. If the difference in vehicle mileage of the detected columns is less than 2m, it indicates that the captured image is within the same sampling interval, and the columns in the image may be repeated. Furthermore, according to the specification, the minimum spacing of guardrail columns is 1m. Formula (C-2) is used to calculate the spacing between the columns in the captured image. The spacing is calculated using the difference in column coordinates of the columns in the image. If the absolute value of the difference in column coordinates of the two columns in the image is less than 100cm (i.e., 1m), it indicates that there must be repeated columns in the two images.

[0063] D copies the intercepted images in the data packets in the queue to be analyzed from the memory to the video memory for bolt detection model inference, and performs non-maximum suppression processing on the detection results, while eliminating the detection results at the edge of the image, and identifies the column intercepted image with the most guardrail bolt detection results as the bolt detection result of the column;

[0064] E obtains the bolt inspection result, which includes the inspection data of bolts, missing holes and contour marks;

[0065] F calculates the missing rate, where missing rate = number of missing holes / (number of bolts + number of missing holes + number of contour marks).

[0066] After field testing and verification, the accuracy of bolt missing detection of this device can reach more than 95% when the vehicle speed does not exceed 90km / h. After the speed exceeds 90km / h, the accuracy of bolt missing detection will decrease. When the vehicle speed is 90km / h, the corresponding speed converted to m / s is 25m / s. At the same time, according to the regulations, the minimum spacing of the guardrail steel columns is 1m. At this time, the sampling frequency of the high-dynamic camera is determined to be 25Hz, which can ensure that at least one image is collected at an average distance of 1m. While ensuring the number of guardrail images collected, it effectively avoids the excessive amount of data collected by the camera's high-frequency sampling, ensuring that the device can process data in real time and balance accuracy and efficiency.

[0067] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and the protection scope of the present invention is defined by the attached claims and their equivalents.

Claims

1. A vehicle-mounted rapid detection device for missing bolts of highway guardrail beams and plates, characterized in that: It includes a mounting mechanism, a camera system, a fill light system, and a control system, wherein the mounting mechanism is used to connect to the vehicle, the camera system is connected to the mounting mechanism and transmits data to the control system, the camera system is used to collect guardrail images, the fill light system is connected to the mounting mechanism to fill light for the camera system, and the control system is used to analyze and process the guardrail image data collected by the camera system, and is also used to control the camera system and the fill light system; The installation mechanism fixes the vehicle-mounted highway guardrail beam and slab bolt missing rapid detection device on the side of the vehicle close to the guardrail. The vehicle travels on the road section to be detected, and the control system controls the camera system to start and begin to collect guardrail image data. The camera system transmits the collected guardrail image data to the control system. The control system controls the start and stop of the fill light system according to the guardrail image information. After completing the detection of all guardrails, the control system outputs the missing status of the corrugated steel guardrail bolts in the detected road section.

2. A vehicle-mounted highway guardrail beam plate bolt missing rapid detection device according to claim 1, characterized in that: The mounting mechanism also includes a height adjustment mechanism and a mounting platform. One end of the height adjustment mechanism is fixedly connected to the mounting platform, and the other end of the height adjustment mechanism is fixedly connected to the vehicle.

3. A vehicle-mounted highway guardrail beam plate bolt missing rapid detection device according to claim 1, characterized in that: The camera system includes a camera angle adjustment mechanism and a high dynamic camera. The camera angle adjustment mechanism is fixedly installed above the mounting platform, and the high dynamic camera is rotatably connected to the camera angle adjustment mechanism. The sampling frequency of the high dynamic camera is 25Hz and the sampling interval is 2m. ， The high dynamic camera captures images with a field of view of 1.4×0.7m.

4. A vehicle-mounted highway guardrail beam plate bolt missing rapid detection device according to claim 1, characterized in that: The fill light system comprises a fill light angle adjustment mechanism and a fill light. The fill light angle adjustment mechanism is fixedly installed below the mounting platform. The fill light is rotatably connected to the fill light angle adjustment mechanism. The number of the fill light is one or two.

5. A vehicle-mounted highway guardrail beam plate bolt missing rapid detection device according to claim 3, characterized in that: The control system includes a host computer, a slave computer, GPS, relays, and distance sensors. The host computer includes data acquisition software and data processing software. The high-dynamic camera acquires data in real time and sends images to the host computer via Ethernet. The GPS sends positioning information to the slave computer via the serial port, and the distance sensor sends distance data to the slave computer via the I / O port. After data shaping by the slave computer, the slave computer sends the positioning information and distance information to the host computer via Ethernet. The host computer combines the brightness information of the image and sends the light-on command to the slave computer, and the slave computer controls the on and off of the fill light through a relay.

6. A vehicle-mounted highway guardrail beam plate bolt missing rapid detection device according to claim 5, characterized in that: The host computer combines traditional visual algorithms with deep learning visual algorithms in terms of visual algorithms. The traditional visual algorithm is responsible for image preprocessing, removing noise from high-dynamic cameras, water mist in the environment, suppressing backlight, and adjusting image brightness. The deep learning visual algorithm is responsible for guardrail beam detection, guardrail column detection, and bolt detection. The guardrail beam detection network and guardrail column detection network are trained using the shallow network of YOLOV5 and inferred using the 16-bit inference engine of TensorRT. The bolt detection network is trained using the deep network of YOLOV5 and inferred using the 16-bit inference engine of TensorRT. The detailed steps for the host computer to implement detection are: A. Image preprocessing, including noise removal, image length and width adjustment, image dark channel defogging, image brightness adjustment, image normalization, and image channel adjustment; Image B is copied from memory to video memory to infer the guardrail beam and column detection network models, and non-maximum suppression is performed on the detection results, while the detection results at the edge of the image are eliminated; C removes duplicates from the detected pillars. The specific operation is as follows: record the car mileage D1 of the first detected pillar, located at column coordinate X1 in the image, and record the car mileage D of the subsequent detected pillars. M , located at column coordinate X in the image M , the value range of M is from 2 to 100; When the pillar on the left side of the car satisfies formula (C-1) and formula (C-2), it is considered to be the same pillar; when the pillar on the right side of the car satisfies formula (C-1) and formula (C-3), it is considered to be the same pillar; D M -D1<2(C-1) X M -X1≤100(C-2) X M -X1≥-100(C-3) The intercepted images of the same column are packaged into a data packet and put into a queue to be analyzed; D copies the intercepted images in the data packets in the queue to be analyzed from the memory to the video memory for bolt detection model inference, and performs non-maximum suppression processing on the detection results, while eliminating the detection results at the edge of the image, and identifies the column intercepted image with the most guardrail bolt detection results as the bolt detection result of the column; E obtains the bolt inspection result, which includes the inspection data of bolts, missing holes and contour marks; F calculates the missing rate, where missing rate = number of missing holes / (number of bolts + number of missing holes + number of contour marks).

7. A vehicle-mounted highway guardrail beam plate bolt missing rapid detection device according to claim 5, characterized in that: The host computer is an industrial computer or a personal computer, and the slave computer is a PLC or an MCU.

Images