A perception ability evaluation method, system and medium based on roadside sensors for vehicle-road cooperation

Abstract

The present invention relates to a method, system and medium for evaluating the perception ability of vehicle-road collaborative roadside sensors. The method includes Q1: Based on the camera in the vehicle-road collaborative roadside sensors, with the sensor deployment position as the coordinate origin and the horizontal rotation angle being zero, obtaining the characteristic coordinate data of the camera perception area, projecting the camera perception area onto the target road surface, and obtaining the coordinate points of the camera perception area on the target road surface through two-dimensional coordinate transformation; Q2: Based on the lidar in the vehicle-road collaborative roadside sensors, with the sensor deployment position as the coordinate origin, obtaining the coordinate points of the lidar perception area on the target road surface through the rotation coordinate transformation algorithm. The present invention can not only provide the perception and coverage models of vehicle-road collaborative roadside sensors, simplify the design of sensor layout schemes and the calculation of coverage rates, but also provide an analysis and control method for the main factors affecting the accuracy of vehicle-road collaborative roadside sensors.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle-road collaborative roadside sensors, and in particular, to a method, system and medium for evaluating the perception ability of vehicle-road collaborative roadside sensors. Background Art

[0002] In a vehicle-road collaborative system, the use of roadside perception technology can effectively make up for the perception blind spots of vehicles, provide real-time information on the road environment for drivers, and at the same time provide monitoring and prediction of the road traffic environment for relevant traffic departments. The roadside perception system includes various sensors such as vision and radar, and with the cooperation of edge computing devices, it can achieve real-time acquisition of current road traffic participants and road condition information. As an important part of the system, the perception ability of roadside sensors greatly affects the efficiency and accuracy of roadside information acquisition. Summary of the Invention

[0003] In view of the above deficiencies of the prior art, the present invention provides a method, system and medium for evaluating the perception ability of vehicle-road collaborative roadside sensors, which can not only provide the perception and coverage models of vehicle-road collaborative roadside sensors, simplify the design of sensor layout schemes and the calculation of coverage rates, but also provide analysis and control methods for the main factors affecting the accuracy of vehicle-road collaborative roadside sensors.

[0004] To achieve the above and other related objectives, the technical solutions provided by the present invention are as follows:

[0005] A method for evaluating the perception ability of vehicle-road collaborative roadside sensors, wherein the vehicle-road collaborative roadside sensors include a camera and a lidar, and the method includes the following steps:

[0006] Q1: Based on the camera in the vehicle-road collaborative roadside sensors, with the sensor deployment position as the coordinate origin and the horizontal rotation angle being zero, obtain the characteristic coordinate data of the camera perception area, project the camera perception area onto the target road surface, and obtain the coordinate points of the camera perception area on the target road surface through two-dimensional coordinate transformation;

[0007] Q2: Based on the lidar in the vehicle-road collaborative roadside sensors, with the sensor deployment position as the coordinate origin, obtain the characteristic coordinate data of the lidar perception area, project the lidar perception area onto the target road surface, and obtain the coordinate points of the lidar perception area on the target road surface through a rotation coordinate transformation algorithm;

[0008] Q3: According to the coordinate points of the camera perception area on the target road surface and the coordinate points of the lidar perception area on the target road surface, obtain the effective coverage area of the camera and the effective coverage area of the lidar, and input the effective coverage area of the camera and the effective coverage area of the lidar into the sensor perception ability evaluation algorithm to output the sensor perception ability evaluation result one;

[0009] Q4: Based on the camera in the vehicle-road collaborative roadside sensor, process the image of the target road surface obtained by the camera according to the Gaussian pyramid algorithm to obtain an image set composed of multiple sub-images with different resolutions. Based on the lidar in the vehicle-road collaborative roadside sensor, obtain the point cloud data set of the target road surface, and input the image set and the point cloud data set into the sensor perception ability evaluation algorithm to output the second sensor perception ability evaluation result;

[0010] Q5: Based on the first sensor perception ability evaluation result and the second sensor perception ability evaluation result, output the comprehensive sensor perception ability evaluation result according to the multi-source data fusion algorithm.

[0011] Further, in step Q1, the obtaining of the coordinate points of the camera sensing area on the target road surface through two-dimensional coordinate transformation includes the following steps:

[0012] Q11: Based on the characteristic coordinate data of the camera sensing area, obtain the minimum sensing distance and the maximum sensing distance of the camera,

[0013] where D min is the minimum sensing distance of the camera, D max is the maximum sensing distance of the camera, H is the distance from the camera to the coordinate origin, γ is the pitch angle of the camera, and VFOV is the vertical field of view;

[0014] Q12: Based on the minimum sensing distance D min and the maximum sensing distance D max of the camera, obtain the vertex coordinates of the camera sensing area,

[0015]

[0016]

[0017] where X A 、X B 、X C 、X D are the four vertices of the camera sensing area respectively, and HFOV is the horizontal field of view of the camera;

[0018] Q13: Based on the four vertices X A 、X B 、X C 、X D of the camera sensing area, obtain the coordinate points of the camera sensing area on the target road surface through the sensing area discrimination condition.

[0019] Further, the discrimination condition of the sensing area is: point P is any coordinate point, satisfying:

[0020] Then point P is the coordinate point of the camera sensing area; otherwise, it is not.

[0021] Further, in step Q2, obtaining the coordinate points of the lidar sensing area on the target road surface through the rotation coordinate transformation algorithm includes:

[0022] Q21: Based on the characteristic coordinate data of the lidar sensing area, determine the minimum blind zone radius and the maximum detection distance of the fan-shaped area,

[0023] R min =GM, R max =GN, where R min is the minimum blind zone radius of the fan-shaped area, R max is the maximum detection distance of the fan-shaped area, GM is the distance from the coordinate origin to the fan-shaped area, and GN is the distance from the coordinate origin to the fan-shaped area;

[0024] Q22: Based on the minimum blind zone radius R min and the maximum detection distance R max of the fan-shaped area, combined with the central angle θ of the fan-shaped area, output the coordinate points of the lidar sensing area on the target road surface.

[0025] Further, in step Q4, the Gaussian pyramid algorithm includes:

[0026] Q41: Based on the image of the target road surface obtained by the camera, perform Gaussian smoothing processing on the image and output the preprocessed image data information;

[0027] Q42: Based on the preprocessed image data information, perform downsampling processing on the preprocessed image data information and output an image set composed of multiple sub-images with different resolutions.

[0028] Further, the sensor sensing ability evaluation algorithm includes:

[0029] Determine the sensor sensing ability evaluation factors, including the effective coverage area and resolution of the camera, the effective coverage area and point cloud data of the lidar;

[0030] Establish a multivariate comparison function between the evaluation factors and the evaluation results, and obtain the sensor sensing ability evaluation result according to the multivariate comparison function.

[0031] Further, the multivariate comparison function is the corresponding function of the sensor sensing ability evaluation factors according to the preset evaluation criteria.

[0032] Further, the sensor sensing ability evaluation results include the resolution and effective coverage area of the camera, the effective coverage area and precision of the lidar.

[0033] To achieve the above and other related objectives, the present invention also provides a perception ability evaluation system based on vehicle-road collaborative roadside sensors, including a computer device, which is programmed or configured to execute the steps of any one of the above-mentioned perception ability evaluation methods based on vehicle-road collaborative roadside sensors.

[0034] To achieve the above and other related objectives, the present invention also provides a computer-readable storage medium, characterized in that a computer program programmed or configured to execute any one of the above-mentioned perception ability evaluation methods based on vehicle-road collaborative roadside sensors is stored on the computer-readable storage medium.

[0035] The present invention has the following positive effects:

[0036] 1. The present invention provides a perception and coverage model for vehicle-road collaborative roadside sensors, which simplifies the design of sensor layout schemes and the calculation of coverage rates.

[0037] 2. The present invention provides an analysis and control method for the main factors affecting the accuracy of vehicle-road collaborative roadside sensors.

[0038] 3. The present invention processes the image data acquired by the camera through the Gaussian pyramid algorithm, and can analyze and study the image data from different scales and resolutions, improving the accuracy of image evaluation.

[0039] 4. The present invention analyzes the data of the camera and lidar respectively through the sensor perception ability evaluation algorithm to obtain the sensor perception ability evaluation results. The method is simple and efficient, with high accuracy, simplifies the complex process, and improves the data processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic flow chart of the method of the present invention;

[0041] Figure 2 It is a schematic flow chart of the Gaussian pyramid algorithm of the present invention;

[0042] Figure 3 It is a schematic diagram of the detection accuracy and image resolution of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0044] Embodiment 1: AsFigure 1 As shown in Figure 1 , a method for evaluating the perception ability of a vehicle-road collaborative roadside sensor, where the vehicle-road collaborative roadside sensor includes a camera and a lidar, includes the following steps:

[0045] Q1: Based on the camera in the vehicle-road collaborative roadside sensor, with the sensor deployment position as the coordinate origin and the horizontal rotation angle being zero, obtain the characteristic coordinate data of the camera's perception area, project the camera's perception area onto the target road surface, and obtain the coordinate points of the camera's perception area on the target road surface through two-dimensional coordinate transformation;

[0046] Q2: Based on the lidar in the vehicle-road collaborative roadside sensor, with the sensor deployment position as the coordinate origin, obtain the characteristic coordinate data of the lidar's perception area, project the lidar's perception area onto the target road surface, and obtain the coordinate points of the lidar's perception area on the target road surface through the rotation coordinate transformation algorithm;

[0047] Q3: According to the coordinate points of the camera's perception area on the target road surface and the coordinate points of the lidar's perception area on the target road surface, obtain the effective coverage area of the camera and the effective coverage area of the lidar, input the effective coverage area of the camera and the effective coverage area of the lidar into the sensor perception ability evaluation algorithm, and output the sensor perception ability evaluation result one;

[0048] Q4: Based on the camera in the vehicle-road collaborative roadside sensor, process the image of the target road surface obtained by the camera according to the Gaussian pyramid algorithm to obtain an image set composed of multiple sub-images with different resolutions. Based on the lidar in the vehicle-road collaborative roadside sensor, obtain the point cloud data set of the target road surface, input the image set and the point cloud data set into the sensor perception ability evaluation algorithm, and output the sensor perception ability evaluation result two;

[0049] Q5: Based on the sensor perception ability evaluation result one and the sensor perception ability evaluation result two, output the comprehensive evaluation result of the sensor perception ability according to the multi-source data fusion algorithm.

[0050] Among them, in step Q1, the obtaining of the coordinate points of the camera's perception area on the target road surface through two-dimensional coordinate transformation includes the following steps:

[0051] Q11: Based on the characteristic coordinate data of the camera's perception area, obtain the minimum perception distance and the maximum perception distance of the camera,

[0052] where D min is the minimum perception distance of the camera, D max is the maximum perception distance of the camera, H is the distance from the camera to the coordinate origin, γ is the pitch angle of the camera, and VFOV is the vertical field of view;

[0053] Q12: Based on the minimum sensing distance D of the camera min and the maximum sensing distance D max , obtain the vertex coordinates of the camera sensing area,

[0054]

[0055]

[0056] where X A , X B , X C , X D are the four vertices of the camera sensing area respectively, and HFOV is the horizontal field of view angle of the camera;

[0057] Q13: Based on the four vertices X A , X B , X C , X D of the camera sensing area, obtain the coordinate points of the camera sensing area on the target road surface through the discrimination conditions of the sensing area.

[0058] where the discrimination condition of the sensing area is: point P is any coordinate point, satisfying:

[0059] then point P is the coordinate point of the camera sensing area, otherwise it is not.

[0060] Embodiment 2: A method for evaluating the sensing ability of a vehicle-road collaborative roadside sensor based on Embodiment 1 is further described below.

[0061] The sensing detection effect is mainly affected by the target resolution. And the target resolution is affected by the size and distance factors of the target. Therefore, constructing a detection rate-resolution model and a resolution-distance model can provide an analysis method for evaluating the sensing accuracy. (If the ground truth boxes in the dataset are not large, the features will be very small during training. For small targets, data labeling errors are most likely to occur, and their recognition may be ignored or incorrect; the bounding boxes of very small objects may only contain a few pixels, which means that increasing the image resolution can increase the rich features that the detector can form from that small box).

[0062] where, in step Q2, the method for obtaining the coordinate points of the lidar sensing area on the target road surface through the rotation coordinate transformation algorithm includes:

[0063] Q21: Based on the characteristic coordinate data of the lidar sensing area, determine the minimum blind zone radius and the maximum detection distance of the fan-shaped area,

[0064] Rmin = GM, R max = GN, where R min is the minimum blind zone radius of the fan-shaped area, R max is the maximum detection distance of the fan-shaped area, GM is the distance from the coordinate origin to the fan-shaped area, and GN is the distance from the coordinate origin to the fan-shaped area;

[0065] Q22: Based on the minimum blind zone radius R min and the maximum detection distance R max of the fan-shaped area, combined with the central angle θ of the fan-shaped area, output the coordinate points of the lidar sensing area on the target road surface.

[0066] Among them, the Gaussian pyramid is a technology used in image processing, computer vision, and signal processing. Essentially, the Gaussian pyramid is a multi-scale representation method of signals, that is, the same signal or image is subjected to multiple Gaussian blurs and downsampled to generate multiple sets of signals or images at different scales for subsequent processing. The image pyramid is an image set composed of multiple sub-images of different resolutions of an image. This set of images is generated by continuously downsampling a single image, and the smallest image may have only one pixel point. As Figure 2 shown, in step Q4, the Gaussian pyramid algorithm includes:

[0067] Q41: Based on the image of the target road surface obtained by the camera, perform Gaussian smoothing processing on the image and output the preprocessed image data information;

[0068] Q42: Based on the preprocessed image data information, perform downsampling processing on the preprocessed image data information and output an image set composed of multiple sub-images of different resolutions.

[0069] Specifically, in traditional research, it is not easy to extract or detect significant features in an image at a single scale for analysis and research. However, often after changing the scale, the features that were originally ignored are more easily observed and discovered. Therefore, for the research point of efficient extraction of image features, introducing multi-scale analysis technology will become a main direction. In multi-scale and hierarchical image processing, clarify the scale levels of an image, and then correlate each level with each other. As Figure 3 shown, as the image resolution continues to increase, the detection accuracy also improves.

[0070] Among them, the sensor perception ability evaluation algorithm includes:

[0071] Determine the sensor perception ability evaluation factors, including the effective coverage area and resolution of the camera, the effective coverage area and point cloud data of the lidar;

[0072] A multivariate comparison function between the evaluation factors and the evaluation results is established, and according to the multivariate comparison function, the evaluation result of the sensor perception ability is obtained.

[0073] Wherein, the multivariate comparison function is a corresponding function made by the perception ability evaluation factor of the sensor according to a preset evaluation standard.

[0074] Wherein, the evaluation result of the sensor perception ability includes the resolution and effective coverage area of the camera, and the effective coverage area and precision of the lidar.

[0075] To achieve the above object and other related objects, the present invention also provides a perception ability evaluation system for a vehicle-road collaborative roadside sensor, including a computer device, which is programmed or configured to execute the steps of any one of the above-mentioned perception ability evaluation methods for a vehicle-road collaborative roadside sensor.

[0076] To achieve the above object and other related objects, the present invention also provides a computer-readable storage medium, characterized in that a computer program programmed or configured to execute any one of the above-mentioned perception ability evaluation methods for a vehicle-road collaborative roadside sensor is stored on the computer-readable storage medium.

[0077] In summary, the present invention can not only provide a perception and coverage model for vehicle-road collaborative roadside sensors, simplify the design of sensor layout schemes and the calculation of coverage rates, but also provide an analysis and control method for the main factors affecting the accuracy of vehicle-road collaborative roadside sensors.

[0078] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating the perception ability of roadside sensors based on vehicle-road cooperation, where the roadside sensors based on vehicle-road cooperation include cameras and lidar, Characterized in that, It includes the following steps: Q1: Based on the cameras in the roadside sensors based on vehicle-road cooperation, with the sensor deployment position as the coordinate origin and the horizontal rotation angle being zero, obtain the characteristic coordinate data of the camera perception area, project the camera perception area onto the target road surface, and obtain the coordinate points of the camera perception area on the target road surface through two-dimensional coordinate transformation; Q2: Based on the lidar in the roadside sensors based on vehicle-road cooperation, with the sensor deployment position as the coordinate origin, obtain the characteristic coordinate data of the lidar perception area, project the lidar perception area onto the target road surface, and obtain the coordinate points of the lidar perception area on the target road surface through the rotation coordinate transformation algorithm; Q3: According to the coordinate points of the camera perception area on the target road surface and the coordinate points of the lidar perception area on the target road surface, obtain the effective coverage area of the camera and the effective coverage area of the lidar, input the effective coverage area of the camera and the effective coverage area of the lidar into the sensor perception ability evaluation algorithm, and output the sensor perception ability evaluation result one; Q4: Based on the cameras in the roadside sensors based on vehicle-road cooperation, process the images of the target road surface obtained by the cameras according to the Gaussian pyramid algorithm to obtain an image set composed of multiple sub-images with different resolutions. Based on the lidar in the roadside sensors based on vehicle-road cooperation, obtain the point cloud data set of the target road surface, input the image set and the point cloud data set into the sensor perception ability evaluation algorithm, and output the sensor perception ability evaluation result two; Q5: Based on the sensor perception ability evaluation result one and the sensor perception ability evaluation result two, output the comprehensive sensor perception ability evaluation result according to the multi-source data fusion algorithm.

2. The method for evaluating the perception ability of roadside sensors based on vehicle-road cooperation according to claim 1, Characterized in that, In step Q1, the obtaining of the coordinate points of the camera perception area on the target road surface through two-dimensional coordinate transformation includes the following steps: Q11: Based on the characteristic coordinate data of the camera perception area, obtain the minimum perception distance and the maximum perception distance of the camera, Among them, D min is the minimum perception distance of the camera, D max is the maximum perception distance of the camera, H is the distance from the camera to the origin of coordinates, γ is the pitch angle of the camera, and VFOV is the vertical field of view; Q12: Based on the minimum sensing distance D of the camera min and the maximum sensing distance D max , obtain the vertex coordinates of the camera sensing area Among them, X A , X B , X C , X D are respectively the four vertices of the camera sensing area, and HFOV is the horizontal field of view angle of the camera; Q13: Based on the four vertices X A , X B , X C , X D of the camera sensing area, obtain the coordinate points of the camera sensing area on the target road surface through the sensing area discrimination condition.

3. The method for evaluating the perception ability of roadside sensors based on vehicle-road cooperation according to claim 2, Characterized in that, The discrimination condition of the perception area is: point P is any coordinate point, satisfying: Then the point P is the coordinate point of the camera sensing area, otherwise it is not.

4. The method for evaluating the perception ability of roadside sensors based on vehicle-road cooperation according to claim 1, Characterized in that, In step Q2, the obtaining of the coordinate points of the lidar perception area on the target road surface through the rotation coordinate transformation algorithm includes: Q21: Based on the characteristic coordinate data of the lidar perception area, determine the minimum blind area radius and the maximum detection distance of the fan-shaped area, R min = GM, R max = GN, where R min is the minimum blind zone radius of the fan-shaped area, R max is the maximum detection distance of the fan-shaped area, GM is the distance from the coordinate origin to the fan-shaped area, and GN is the distance from the coordinate origin to the fan-shaped area; Q22: Minimum blind area radius R of the fan-shaped area min and maximum detection distance R max , combined with the central angle θ of the fan-shaped area, output the coordinate points of the lidar sensing area on the target road surface.

5. The method for evaluating the perception ability of roadside sensors based on vehicle-road cooperation according to claim 1, Characterized in that, In step Q4, the Gaussian pyramid algorithm includes: Q41: Based on the image of the target road surface obtained by the camera, perform Gaussian smoothing processing on the image and output the preprocessed image data information; Q42: Based on the preprocessed image data information, perform downsampling processing on the preprocessed image data information and output an image set composed of multiple sub-images with different resolutions.

6. The perception ability evaluation method based on vehicle-road collaborative roadside sensors according to claim 1, characterized in that, the sensor perception ability evaluation algorithm includes: determine the perception ability evaluation factors of the sensor, including the effective coverage area and resolution of the camera, the effective coverage area and point cloud data of the lidar; establish a multivariate comparison function between the evaluation factors and the evaluation results, and obtain the sensor perception ability evaluation result according to the multivariate comparison function.

7. The perception ability evaluation method based on vehicle-road collaborative roadside sensors according to claim 6, characterized in that: the multivariate comparison function is a corresponding function made by the perception ability evaluation factors of the sensor according to the preset evaluation criteria.

8. The perception ability evaluation method based on vehicle-road collaborative roadside sensors according to claim 6, characterized in that: the sensor perception ability evaluation results include the resolution and effective coverage area of the camera, the effective coverage area and precision of the lidar.

9. A perception ability evaluation system based on vehicle-road collaborative roadside sensors, including a computer device, characterized in that, the computer device is programmed or configured to execute the steps of the perception ability evaluation method based on vehicle-road collaborative roadside sensors according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, a computer program programmed or configured to execute the perception ability evaluation method based on vehicle-road collaborative roadside sensors according to any one of claims 1 to 8 is stored on the computer-readable storage medium.

Images