A tunnel traffic behavior intelligent monitoring and analysis method and electronic equipment

Abstract

The present application relates to the field of tunnel intelligent traffic technology, and specifically to a method and electronic equipment for intelligent monitoring and analysis of tunnel traffic behavior, which specifically includes: obtaining a clutter interference degree based on the difference between each scanning point and a neighboring scanning point in radar data, obtaining a far-field correction degree based on the position information of the scanning point and the difference in the transmission and reception energy, obtaining a rejection value of each scanning point in combination with the clutter interference degree, and screening the scanning points using the rejection value; constructing radar point cloud data from the retained scanning points; obtaining visual point cloud data through visual images using monocular estimation, fusing point cloud data with radar point cloud data, obtaining driving information of vehicles in the tunnel using a 3D multi-target detection algorithm, and reporting sudden abnormal situations to a traffic command center; screening and judging clutter interference points, eliminating the interference of clutter interference points on radar data, improving the accuracy of the fused point cloud data, and realizing accurate monitoring of traffic behavior in the tunnel.

Description

Technical Field

[0001] The present application relates to the field of tunnel intelligent traffic technology, and specifically to a tunnel traffic behavior intelligent monitoring and analysis method and electronic equipment. Background Art

[0002] With the continuous increase in the number of cars, the number of vehicles on the road is increasing, resulting in an increasing contradiction between the high rate of traffic accidents and the inefficiency of emergency rescue. Especially for blind areas such as highway tunnels, accidents are prone to slow detection, long processing time, and easy to cause secondary accidents. Therefore, a method and equipment for intelligent monitoring and analysis of tunnel traffic behavior is needed.

[0003] At present, radar-visual integrated devices are used for traffic behavior monitoring in highway tunnels because they have the good penetration of radar equipment, are less affected by light, and provide continuous image information for video equipment. The radar-visual integrated device mainly uses the technical means of radar-visual fusion to monitor vehicles running in the tunnel. However, the current radar-visual fusion solutions mostly rely on expert experience design, have cumbersome fusion steps, and low data utilization. In addition, during radar-visual fusion, radar data is easily interfered by clutter from reflected signals from tunnel walls, internal facilities, etc., while video data is easily affected by ambient light, resulting in low accuracy of radar-visual fusion, which is easy to cause misjudgment in the monitoring and analysis of traffic behavior in tunnels. In response to the above problems, this application proposes a method and electronic equipment for intelligent monitoring and analysis of tunnel traffic behavior. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a tunnel traffic behavior intelligent monitoring and analysis method and electronic equipment, and the technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a method for intelligent monitoring and analysis of tunnel traffic behavior, the method comprising the following steps:

[0006] Collect radar scanning depth map, radar reflection intensity map and visual image in the tunnel at each time;

[0007] Based on the uniformity of the depth value distribution of all scanning points in the neighborhood of each scanning point in the radar scanning depth map, the depth highlight value of each scanning point is constructed; based on the difference in radar reflection intensity of the scanning point in the radar reflection intensity map combined with the difference in depth value, the intensity highlight value of each scanning point is constructed; the fusion value of the depth highlight value and the intensity highlight value of each scanning point is used as the clutter interference degree of each scanning point;

[0008] Distinguish near-field area points and far-field area points based on the depth value of each scanning point; construct the far-field correction degree of each far-field area point based on the difference between the transmitted radar energy and the reflected radar energy of each far-field area point and the position of each far-field area point; set the far-field correction degree of the near-field area point; use the fusion value of the clutter interference degree and the far-field correction degree of each scanning point as the rejection value of each scanning point;

[0009] Scanning points are eliminated based on the elimination values ​​of all scanning points; radar point cloud data at each moment is obtained based on the depth values ​​of the remaining scanning points at each moment; visual point cloud data at each moment is obtained based on the visual images at each moment; and tunnel traffic behavior detection is performed in combination with radar point cloud data.

[0010] In one embodiment, the process of obtaining the depth highlight value of each scanning point is as follows:

[0011] Construct the neighborhood of each scanning point with each scanning point as the center;

[0012] Calculate the average value of the depth values ​​of all scanning points in the neighborhood; count the number of scanning points in the neighborhood whose depth values ​​are greater than the average value, recorded as a first number; count the number of scanning points in the neighborhood whose depth values ​​are less than or equal to the first average value, recorded as a second number;

[0013] The depth highlight value of the current scanning point is recorded as B, and the expression of B is: , where Represents the range of depth values ​​of all scan points in the local neighborhood of the current scan point; It is the ratio of the maximum value to the minimum value in the first quantity and the second quantity of the current scanning point.

[0014] In one embodiment, the expression of the intensity highlight value of each scanning point is:

[0015] , where C represents the intensity highlight value of the current scanning point, and They represent the average depth value and average radar reflection intensity of all scanning points in the local neighborhood of the current scanning point, respectively. and They respectively represent the depth value and radar reflection intensity of the current scanning point.

[0016] In one embodiment, the process of distinguishing near-field area points from far-field area points is:

[0017] The depth values ​​of all scanning points in the radar scanning depth map are used as the input of the threshold segmentation algorithm, and the distance threshold is output. The scanning points with depth values ​​less than or equal to the distance threshold are marked as near-field area points, and the scanning points with depth values ​​greater than the distance threshold are marked as far-field area points.

[0018] In one embodiment, the expression of the far field correction degree of each far field area point is:

[0019] , where F represents the far-field correction degree of the current far-field area point, Indicates the Euclidean distance from the current far-field point to the upper boundary of the radar scanning depth map. and They respectively represent the energy of the transmitted millimeter-wave radar and the reflected radar energy received by the current far-field area point.

[0020] In one embodiment, the process of removing scanning points is as follows:

[0021] The rejection values ​​of all scanning points at the current sampling moment are used as the input of the threshold segmentation algorithm, and the segmentation threshold is output; the scanning points with rejection values ​​greater than the segmentation threshold are removed from the radar scanning depth map.

[0022] In one embodiment, the process of acquiring the radar point cloud data at each moment is as follows:

[0023] For the remaining scanning points, the radar point cloud data is determined by the depth value of the scanning point combined with the corresponding internal parameters of the millimeter-wave radar and the coordinates of the main point.

[0024] In one embodiment, the process of acquiring the visual point cloud data at each moment is as follows:

[0025] The visual image at the current sampling moment is used as the input of the uneven illumination image correction algorithm, and the output is the corrected visual image; the corrected visual image is used as the input of the depth estimation model, and the output is the visual point cloud data.

[0026] In one embodiment, the acquisition process of tunnel traffic behavior detection is:

[0027] All radar point cloud data and all visual point cloud data at the current sampling moment are used as the input of the point cloud registration algorithm, and the output is the point cloud data after radar and visual fusion at the current sampling moment;

[0028] The point cloud data after the fusion of the radar and vision is passed through the 3D point cloud target tracking algorithm to obtain the spatial position information of each vehicle in the tunnel at the current sampling time, and assign different label data to each vehicle;

[0029] Based on the label data and spatial position information of each vehicle at continuous sampling moments, the driving status of each vehicle is obtained.

[0030] In a second aspect, an embodiment of the present application also provides an electronic device for intelligent monitoring and analysis of tunnel traffic behavior, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0031] The embodiments of the present application have at least the following beneficial effects:

[0032] The present application deploys intelligent monitoring and analysis electronic equipment for tunnel traffic behavior in the tunnel, and uses a data acquisition device to obtain radar data and visual data; the clutter interference degree is obtained based on the difference between each scanning point and the neighboring scanning point in the radar data, thereby reflecting the interference of each scanning point by the tunnel clutter, and the far-field correction degree is obtained by using the position information of the scanning point and the difference in the emission and reception energy, so as to avoid the possibility of misjudging the far-field point as a clutter interference point due to spatial discontinuity, and finally obtain the rejection value of each scanning point, and use the rejection value to screen the scanning points; radar point cloud data is constructed from the retained scanning points, and visual point cloud data is obtained by monocular estimation for the visual image, thereby realizing point cloud data fusion, and finally the driving information of the vehicle in the tunnel is obtained by using a 3D multi-target detection algorithm, and the sudden abnormal situation is reported to the traffic command center. In the traditional radar-visual fusion algorithm, the fusion of radar point cloud and visual point cloud is directly realized based on the radar scanning data. This application focuses on the clutter interference caused by the scattering of radar transmission signals by static objects such as tunnel walls in the complex environment of the tunnel, screens and judges the clutter interference points, eliminates the interference of clutter interference points on radar data, improves the accuracy of the fused point cloud data, and realizes accurate monitoring of traffic behavior in the tunnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 A flowchart of a method for intelligent monitoring and analysis of tunnel traffic behavior provided by an embodiment of the present application;

[0035] Figure 2 Schematic diagram of the process of obtaining the depth highlight value of each scanning point. DETAILED DESCRIPTION

[0036] In order to further explain the technical means and effects adopted by this application to achieve the predetermined invention purpose, the specific implementation method, structure, features and effects of a tunnel traffic behavior intelligent monitoring and analysis method and electronic equipment proposed in this application are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0037] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0038] The following is a detailed description of a tunnel traffic behavior intelligent monitoring and analysis method and a specific solution of an electronic device provided by the present application in conjunction with the accompanying drawings.

[0039] See also Figure 1 , which shows a flowchart of a method for intelligent monitoring and analysis of tunnel traffic behavior provided by an embodiment of the present application, the method comprising the following steps:

[0040] Step S1, collecting radar scanning depth map, radar reflection intensity map and visual image in the tunnel at each time.

[0041] Taking a single tunnel as an example, the intelligent monitoring and analysis electronic equipment for tunnel traffic behavior is deployed at the top of the tunnel at the midpoint of the tunnel, and the equipment is oriented in the direction of the traffic flow and tilted downward by 30°, so that a sufficiently large viewing angle can be obtained to clearly collect the tunnel traffic situation, wherein the radar signal is transmitted by the millimeter wave radar and the reflected radar signal of each scanning point in the tunnel is collected, and the time difference between the transmitted radar signal and the reflected radar signal received by each scanning point is used as the depth value of each scanning point, and the image composed of the depth values ​​of all scanning points is used as the radar scanning depth map; the radar reflection intensity map is constructed according to the intensity of the reflected radar signal of all scanning points; and the visual image in the tunnel is collected by the visual sensor. Preferably, in one embodiment of the present application, when collecting images in the tunnel, the present embodiment sets the image collection time interval to 5s, that is, the radar scanning depth map, radar reflection intensity map and visual image are collected every 5s in the embodiment of the present application. As other embodiments, the implementer can decide the image collection time interval according to the actual situation.

[0042] It should be noted that the implementer can set the downward tilt angle of the smart monitoring and analysis electronic equipment according to actual conditions, and this application does not impose any specific restrictions.

[0043] Step S2, constructing a depth highlight value for each scanning point based on the uniformity of the depth value distribution of all scanning points in the neighborhood of each scanning point in the radar scanning depth map; constructing an intensity highlight value for each scanning point based on the difference in radar reflection intensity of the scanning point in the radar reflection intensity map combined with the depth value difference; and obtaining the clutter interference degree of each scanning point by combining the intensity highlight values.

[0044] Due to the complex tunnel environment, when the millimeter-wave radar transmits the radar scanning signal, it is easily affected by the reflection and scattering of the walls and ground in the tunnel, resulting in multipath effect and clutter interference. As a result, the depth values ​​of some scanning points in the radar scanning depth map are inaccurate, which will affect the accuracy of subsequent radar-vision fusion. Therefore, it is necessary to combine the radar reflection information to perform depth correction on the radar depth data.

[0045] In the millimeter wave radar scanning scenario, the normal situation should be the surface situation in the tunnel scene, that is, each scanning point in the radar scanning depth map should form a continuously changing surface, and for adjacent points with close depth values, the values ​​of the radar signal reflection intensity of the two are relatively close. However, if the current scanning point is an interference point affected by clutter, the reflection signal of the current scanning point may be a clutter signal that has been reflected multiple times during the previous scan. At this time, the time difference between the transmitted signal and the received signal is small, which is manifested as a small distance in the depth value, forming an isolated point. At the same time, since the reflection signal corresponding to this point is the previous scanning signal after multiple rounds of reflection, the reflection intensity of this scanning point is significantly smaller than the reflection intensity of other scanning points.

[0046] (1) In the radar scanning depth map, the depth highlight value of each scanning point is constructed based on the uniformity of the depth value distribution of all scanning points in the neighborhood of each scanning point to analyze the isolation and continuity of each scanning point with the surrounding scanning points. Specifically:

[0047] (1.1) Determine the local neighborhood of each scanning point in the radar scanning depth map: construct a 3*3 neighborhood of each scanning point with each scanning point as the center, and use the 3*3 neighborhood of each scanning point as the local neighborhood of each scanning point. It should be noted that, in other embodiments of the present application, the implementer can set the size of the local neighborhood according to actual conditions.

[0048] (1.2) Calculate the average value of the depth values ​​of all scanning points in the local neighborhood of each scanning point, recorded as the first average value; count the number of scanning points in the local neighborhood whose depth values ​​are greater than the first average value, recorded as the first number; count the number of scanning points in the local neighborhood whose depth values ​​are less than or equal to the first average value, recorded as the second number;

[0049] (1.3) For a local neighborhood of any scanning point, a depth highlight value of the scanning point is constructed based on the difference in depth values ​​of the scanning points in the local neighborhood and the difference between the first quantity and the second quantity. The expression is:

[0050] , where B represents the depth highlight value of the current scanning point; Represents the range of depth values ​​of all scan points in the local neighborhood of the current scan point; It is the ratio of the maximum value to the minimum value in the first quantity and the second quantity of the current scanning point.

[0051] If the scanning point is a normal scanning point, it must be connected to the surrounding scanning points in space, that is, the depth values ​​in the local neighborhood are not much different, and the range value is small. At the same time, due to the continuity of the spatial surface, the scanning point data greater than or less than the average depth value in the local neighborhood is not much different, so the obtained value of n is small, and finally the depth highlight value of the current scanning point is small. On the contrary, if the scanning point is a clutter interference point, it is relatively isolated in space, which makes the depth highlight value larger.

[0052] (2) Construct the intensity highlight value of each scanning point based on the difference in depth value and radar reflection intensity of the scanning point in the local neighborhood:

[0053] , where C represents the intensity highlight value of the current scanning point, and They represent the average depth value and average radar reflection intensity of all scanning points in the local neighborhood of the current scanning point, respectively. and They respectively represent the depth value and radar reflection intensity of the current scanning point.

[0054] For normal scanning points, they are spatially continuous with the surrounding scanning points, so the average depth reflection intensity ratio and its own depth reflection intensity ratio are slightly different, resulting in a smaller intensity highlight value. However, for clutter interference points, since the reflected signal comes from the clutter reflected by multipath during the last sampling, the radar reflection energy is severely attenuated, resulting in a smaller corresponding radar reflection intensity value. Therefore, the difference between the average depth reflection intensity ratio and its own depth reflection intensity ratio is large, resulting in a larger intensity highlight value.

[0055] (3) The fusion value of the depth highlight value and the intensity highlight value of the current scanning point is used as the clutter interference degree of the current scanning point. It should be noted that the fusion described in this application is to combine multiple variables. The specific fusion method can be determined according to the actual situation during the application process, and this application does not make any special restrictions.

[0056] Preferably, in one embodiment of the present application, the clutter interference degree of the current scanning point may be the sum of the depth highlight value and the intensity highlight value of the current scanning point;

[0057] In other embodiments of the present application, the clutter interference degree of the current scanning point may be the product of the depth highlight value and the intensity highlight value of the current scanning point.

[0058] If the current scanning point is a clutter interference point, it will be an isolated point on the radar scanning depth map, so the depth value of the scanning point in the local neighborhood of the scanning point will be greatly different, and the scanning point will be obviously different from the local neighborhood points, resulting in a larger value of quantity comparison and a larger depth highlight value. In the radar reflection intensity map, the energy intensity is obviously lower, resulting in a larger difference in the depth reflection intensity ratio between the scanning point and the local neighborhood, and finally a larger clutter interference value of the scanning point.

[0059] Step S3, distinguishing near-field area points and far-field area points based on the depth value of each scanning point; constructing the far-field correction degree of each far-field area point based on the difference between the transmitted radar energy and the reflected radar energy of each far-field area point, combined with the position of each far-field area point; setting the far-field correction degree of the near-field area point; and obtaining the rejection value of each scanning point based on the clutter interference degree and the far-field correction degree.

[0060] In the millimeter-wave radar scanning scene, due to the physical and signal processing characteristics of the radar system, the depth map obtained by the radar scan presents different resolutions at different distances. Specifically, the radar system's scanning points in the near field area are denser, while the scanning points in the far field area are sparser. Therefore, the value of the clutter interference calculated for the scanning points in the far field area may be large, which may lead to misjudgment of the clutter interference scanning points. Therefore, further analysis is required in combination with the radar acquisition data.

[0061] (1) In order to distinguish the near-field area and the far-field area in the tunnel scene, the depth values ​​of all scanning points in the radar scanning depth map are used as the input of the Otsu threshold method. The algorithm outputs the distance threshold, and the scanning points with depth values ​​less than or equal to the distance threshold are marked as near-field area points, and the scanning points with depth values ​​greater than the distance threshold are marked as far-field area points. The points in the far-field area are further analyzed. The Otsu threshold method is a well-known technology and will not be described in detail in this application.

[0062] It should be noted that for the calculation of the distance threshold, this application only provides a threshold segmentation method. There are many existing threshold segmentation algorithms, and implementers can also use other threshold segmentation algorithms to calculate the distance threshold. This application does not make specific restrictions.

[0063] (2) For a normal scanning point in the far field, due to the influence of the tilt angle and resolution, the distance between the scanning point and the corresponding scanning point in the local neighborhood in the actual space is far away, and it cannot be a relatively continuous plane. Therefore, it is not accurate to judge the interference of the scanning point by clutter only by the clutter interference degree.

[0064] Under normal circumstances, in the radar scanning depth map, the closer the distance to the upper boundary of the image, the farther the scanning point is from the acquisition module. At this time, it mainly corresponds to the road surface information and vehicle information of the tunnel, and the overall possibility of being interfered by the reflection inside the tunnel is small. At the same time, the normal scanning point is the signal transmitted by the millimeter-wave radar device at the current sampling moment and is reflected by the ground or vehicle and received, so the energy intensity attenuation is smaller than that of the transmission. For the scanning point with clutter interference, the energy intensity attenuation is large because it is received by the refraction and scattering of static objects passing through the tunnel.

[0065] Based on the above analysis, the far-field correction degree of each far-field area point is constructed, and the expression is:

[0066] , where F represents the far-field correction degree of the current far-field area point, Indicates the Euclidean distance from the current far-field point to the upper boundary of the radar scanning depth map. and They respectively represent the energy of the transmitted millimeter-wave radar and the reflected radar energy received by the current far-field area point.

[0067] If the current far-field point is closer to the upper boundary, the smaller the portion of the tunnel wall corresponding to the depth map, the less likely it is to be interfered with. At the same time, the scanning point in the far-field area can directly reflect the radar signal, so the attenuation of the received radar reflected energy is smaller than that of the radar reflected energy, and the final value of the far-field correction is smaller. On the contrary, if the scanning point is interfered by clutter, the corresponding received signal is an interference signal that is scattered and reflected multiple times, so the energy intensity is more severely attenuated, and the value of the far-field correction is larger.

[0068] (3) Setting the far field correction degree of the near field area point. Preferably, in the embodiment of the present application, the far field correction degree of the near field area point is set to 1. In other embodiments of the present application, the implementer can set the far field correction degree of the near field area point according to actual conditions, and the present application does not make any specific restrictions.

[0069] (4) The fusion value of the clutter interference degree and the far-field correction degree of each scanning point is used as the rejection value of each scanning point.

[0070] Preferably, in one embodiment of the present application, the rejection value of each scanning point may be a normalized value of the product of the clutter interference degree and the far-field correction degree of each scanning point;

[0071] In other embodiments of the present application, the rejection value of each scanning point may be the sum of the normalized value of the clutter interference degree and the normalized value of the far-field correction degree of each scanning point.

[0072] As a result, the rejection value is calculated for all scanning points in the radar scanning depth map at the current sampling moment. For normal scanning points, the corresponding clutter interference degree and far-field correction degree are small, and the corresponding rejection value is small. For real clutter interference scanning points, the calculated clutter interference degree and far-field correction value are large, so the rejection value is large.

[0073] Step S4, performing scanning point elimination based on the elimination values ​​of all scanning points; obtaining radar point cloud data at each moment based on the depth values ​​of the remaining scanning points at each moment; obtaining visual point cloud data at each moment based on the visual images at each moment; and performing tunnel traffic behavior detection in combination with radar point cloud data.

[0074] (1) The rejection values ​​of all scanning points at the current sampling moment are used as the input of the cross-validation algorithm, and the segmentation threshold is output; the scanning points whose rejection values ​​are greater than the segmentation threshold are removed from the radar scanning depth map; for the remaining scanning points, the radar point cloud data is determined by the depth value of the scanning point, combined with the corresponding internal parameters of the millimeter-wave radar and the coordinates of the main point. Among them, the cross-validation algorithm is a well-known technology, and the specific process is not repeated here.

[0075] It should be noted that for obtaining the segmentation threshold of the rejection value of all scanning points, this application only provides a threshold segmentation method. There are many existing threshold segmentation methods. Implementers can also use other threshold segmentation algorithms to calculate the segmentation threshold of the rejection value of all scanning points. This application does not make specific restrictions.

[0076] (2) In order to reduce the impact of uneven light in the tunnel, the visual image at the current sampling moment is used as the input of the two-dimensional gamma function illumination image adaptive correction algorithm, and the output is the corrected visual image; the corrected visual image is used as the input of the DenseNet monocular depth estimation model, and the output is visual point cloud data. Among them, the two-dimensional gamma function illumination image adaptive correction algorithm and the DenseNet monocular depth estimation model are both well-known technologies, and the specific process will not be repeated here.

[0077] It should be noted that, for the illumination correction of visual images, this application only provides a method for uneven illumination image correction. There are many existing methods for uneven illumination image correction. Implementers may also use other uneven illumination image correction algorithms to correct the illumination of visual images. This application does not make any specific restrictions.

[0078] (3) All radar point cloud data and all visual point cloud data at the current sampling moment are used as inputs of the ICP iterative closest point algorithm, and the output is the point cloud data after radar and visual fusion at the current sampling moment, thereby realizing the fusion of radar data and visual data at the current sampling moment. Among them, the ICP iterative closest point algorithm is a well-known technology, and the specific process is not repeated here.

[0079] It should be noted that for the fusion of radar point cloud data and visual point cloud data, this application only provides a point cloud registration algorithm. There are many existing point cloud registration algorithms, and implementers can also use other point cloud registration algorithms to fuse radar point cloud data and visual point cloud data. This application does not make specific restrictions.

[0080] (4) The point cloud data after the fusion of the radar and vision is passed through the SimTrack point cloud 3D target tracking algorithm to obtain the spatial position information of each vehicle in the tunnel at the current sampling time, and assign different label data to each vehicle. Among them, the SimTrack point cloud 3D target tracking algorithm is a well-known technology, and the specific process is not repeated here.

[0081] (5) Based on the continuous sampling moments, the driving status of each vehicle is obtained according to the label data and spatial position information of each vehicle in the tunnel at each sampling moment. Taking a single vehicle as an example, the mean of all point cloud data of the single vehicle is calculated as the center point of the single vehicle. If the center point of the vehicle does not move at five consecutive sampling moments, it indicates that the current vehicle is abnormally parked in the tunnel, or a safety accident occurs in the tunnel, forcing the vehicle to stop.

[0082] The event is then uploaded to the traffic management center using 5G wireless transmission to improve the traffic management department's rapid response to emergencies in the tunnel. It should be noted that the number of consecutive sampling moments can be set by the implementer, and this application does not impose specific restrictions.

[0083] The schematic diagram of the process of obtaining the depth highlight value of each scanning point is as follows: Figure 2 shown.

[0084] Based on the same inventive concept as the above method, an embodiment of the present application also provides an electronic device for intelligent monitoring and analysis of tunnel traffic behavior, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-mentioned methods for intelligent monitoring and analysis of tunnel traffic behavior.

[0085] In summary, the embodiment of the present application provides a method for intelligent monitoring and analysis of tunnel traffic behavior, by deploying electronic equipment for intelligent monitoring and analysis of tunnel traffic behavior in the tunnel, and using a data acquisition device to obtain radar data and visual data; the clutter interference degree is obtained based on the difference between each scanning point and the neighboring scanning point in the radar data, thereby reflecting the interference of each scanning point by tunnel clutter, and the far-field correction degree is obtained by using the position information of the scanning point and the difference in the transmitting and receiving energies, so as to avoid the possibility of misjudging the far-field point as a clutter interference point due to its discontinuity in space, and finally the rejection value of each scanning point is obtained, and the scanning points are screened by using the rejection value; radar point cloud data is constructed from the retained scanning points, and visual point cloud data is obtained by monocular estimation for the visual image, thereby realizing point cloud data fusion, and finally the driving information of the vehicle in the tunnel is obtained by using a 3D multi-target detection algorithm, and the sudden abnormal situation is reported to the traffic command center. In the traditional radar-visual fusion algorithm, the fusion of radar point cloud and visual point cloud is directly realized based on the radar scanning data. This application focuses on the clutter interference caused by the scattering of radar transmission signals by static objects such as tunnel walls in the complex environment of the tunnel, screens and judges the clutter interference points, eliminates the interference of clutter interference points on radar data, improves the accuracy of the fused point cloud data, and realizes accurate monitoring of traffic behavior in the tunnel.

[0086] It should be noted that the above sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of the present application. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0087] The various embodiments in the present application are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0088] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for intelligent monitoring and analysis of tunnel traffic behavior, characterized in that: The method comprises the following steps: Collect radar scanning depth map, radar reflection intensity map and visual image in the tunnel at each time; Based on the uniformity of the depth value distribution of all scanning points in the neighborhood of each scanning point in the radar scanning depth map, the depth highlight value of each scanning point is constructed; based on the difference in radar reflection intensity of the scanning point in the radar reflection intensity map combined with the difference in depth value, the intensity highlight value of each scanning point is constructed; the fusion value of the depth highlight value and the intensity highlight value of each scanning point is used as the clutter interference degree of each scanning point; Distinguish near-field area points and far-field area points based on the depth value of each scanning point; construct the far-field correction degree of each far-field area point based on the difference between the transmitted radar energy and the reflected radar energy of each far-field area point and the position of each far-field area point; set the far-field correction degree of the near-field area point; use the fusion value of the clutter interference degree and the far-field correction degree of each scanning point as the rejection value of each scanning point; Scanning points are eliminated based on the elimination values ​​of all scanning points; radar point cloud data at each moment is obtained based on the depth values ​​of the remaining scanning points at each moment; visual point cloud data at each moment is obtained based on the visual images at each moment; and tunnel traffic behavior detection is performed in combination with radar point cloud data.

2. A tunnel traffic behavior intelligent monitoring and analysis method as claimed in claim 1, characterized in that: The process of obtaining the depth highlight value of each scanning point is as follows: Construct the neighborhood of each scanning point with each scanning point as the center; Calculate the average value of the depth values ​​of all the scanning points in the neighborhood; count the number of scanning points in the neighborhood whose depth values ​​are greater than the average value, and record it as a first number; Count the number of scanning points in the neighborhood whose depth values ​​are less than or equal to the first mean value, and record it as the second number; The depth highlight value of the current scanning point is recorded as B, and the expression of B is: , where Represents the range of depth values ​​of all scan points in the local neighborhood of the current scan point; It is the ratio of the maximum value to the minimum value in the first quantity and the second quantity of the current scanning point.

3. The intelligent monitoring and analysis method for tunnel traffic behavior according to claim 1, characterized in that: The expression of the intensity highlight value of each scanning point is: , where C represents the intensity highlight value of the current scanning point, and They represent the average depth value and average radar reflection intensity of all scanning points in the local neighborhood of the current scanning point, respectively. and They respectively represent the depth value and radar reflection intensity of the current scanning point.

4. The intelligent monitoring and analysis method for tunnel traffic behavior according to claim 1, characterized in that: The process of distinguishing near-field points from far-field points is as follows: The depth values ​​of all scanning points in the radar scanning depth map are used as the input of the threshold segmentation algorithm, and the distance threshold is output. The scanning points with depth values ​​less than or equal to the distance threshold are marked as near-field area points, and the scanning points with depth values ​​greater than the distance threshold are marked as far-field area points.

5. The intelligent monitoring and analysis method for tunnel traffic behavior according to claim 1, characterized in that: The expression of the far field correction degree of each far field area point is: , where F represents the far-field correction degree of the current far-field area point, Indicates the Euclidean distance from the current far-field point to the upper boundary of the radar scanning depth map. and They respectively represent the energy of the transmitted millimeter-wave radar and the reflected radar energy received by the current far-field area point.

6. The intelligent monitoring and analysis method for tunnel traffic behavior according to claim 1, characterized in that: The process of removing scanning points is as follows: The rejection values ​​of all scanning points at the current sampling moment are used as the input of the threshold segmentation algorithm, and the segmentation threshold is output; the scanning points with rejection values ​​greater than the segmentation threshold are removed from the radar scanning depth map.

7. The intelligent monitoring and analysis method for tunnel traffic behavior according to claim 1, characterized in that: The acquisition process of the radar point cloud data at each moment is as follows: For the remaining scanning points, the radar point cloud data is determined by the depth value of the scanning point combined with the corresponding internal parameters of the millimeter-wave radar and the coordinates of the main point.

8. The intelligent monitoring and analysis method for tunnel traffic behavior according to claim 1, characterized in that: The process of acquiring the visual point cloud data at each moment is as follows: The visual image at the current sampling moment is used as the input of the uneven illumination image correction algorithm, and the output is the corrected visual image; the corrected visual image is used as the input of the depth estimation model, and the output is the visual point cloud data.

9. The intelligent monitoring and analysis method for tunnel traffic behavior according to claim 1, characterized in that: The acquisition process of tunnel traffic behavior detection is as follows: All radar point cloud data and all visual point cloud data at the current sampling moment are used as the input of the point cloud registration algorithm, and the output is the point cloud data after radar and visual fusion at the current sampling moment; The point cloud data after the fusion of the radar and vision is passed through the 3D point cloud target tracking algorithm to obtain the spatial position information of each vehicle in the tunnel at the current sampling time, and assign different label data to each vehicle; Based on the label data and spatial position information of each vehicle at continuous sampling moments, the driving status of each vehicle is obtained.

10. An electronic device for intelligent monitoring and analysis of tunnel traffic behavior, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images