Evaluation method and device of vehicle drivable area detection algorithm and computer readable storage medium

Abstract

The invention discloses an evaluation method and device for a vehicle travelable area detection algorithm and a computer readable storage medium, and the method comprises the steps: obtaining a vehicle travelable area through the application of the vehicle travelable area detection algorithm; the vehicle driving area is converted into a self-vehicle coordinate system, and the vehicle driving area in the self-vehicle coordinate system is converted into grids for binaryzation; the binarized grids are mapped to a polar coordinate system, the distance between the travelable area of each angle resolution in the polar coordinate system and the vehicle serves as a detection value, the error between the detection value and a true value is calculated, and the error of the travelable area in the polar coordinate system is obtained; and evaluating the error according to a preset evaluation dimension. According to the method, the detection result under the self-vehicle coordinate system is converted into the polar coordinate value with actual physical significance, the actual physical distance difference under the self-vehicle coordinate system is directly evaluated, and more objective end-to-end evaluation of the travelable area is completed.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent driving technology, and specifically relates to an evaluation method, device and computer-readable storage medium for a vehicle drivable area detection algorithm. Background Art

[0002] Perception is an important part of the automatic parking algorithm. The perception tasks include parking space detection, drivable area detection, etc. Among them, drivable area detection provides important information for downstream planning and control, and is the focus and difficulty of automatic parking perception. The drivable area detection algorithm is an algorithm for detecting the drivable area and the impassable area when parking. This algorithm is an important part of the automatic parking perception task.

[0003] Different vehicle hardware, computing resources, etc. in automatic parking scenarios result in different technical routes of drivable area detection algorithms. According to the different sensors they rely on, they can be divided into vision-based, ultrasonic, and multi-sensor fusion algorithms. According to whether deep learning is used, they can be divided into traditional methods and deep learning-based methods. Different algorithms have their own focus. How to evaluate the pros and cons of an algorithm is the key to promoting algorithm iteration, and it is also an important reference for control and planning algorithms.

[0004] The drivable area detection algorithm needs to consider the recognition of different semantic objects in the perception area and the spatial relationship between the object and the vehicle. Different hardware and algorithms produce different types of detection algorithms. Generally, the algorithm is evaluated using a computer vision-based evaluation method, that is, the intersection over union (IoU) with the true value is used to calculate the quality of the recognition. However, this method is based on the pixel coordinate system and treats the task as a graphic segmentation task. The evaluation is the quality of the image segmentation in the pixel coordinate system. The obtained value can only evaluate the quality of the image segmentation, and cannot evaluate the error of the drivable area in the physical space. In addition, in automatic parking, the distance between the edge of the drivable area and the vehicle, the distance between the vehicle and the corner point of the parking space, etc., will have different effects on the drivable area. A single evaluation index cannot evaluate the above factors. How to multi-dimensionally and quantitatively evaluate the drivable area in the vehicle coordinate system becomes extremely difficult. Therefore, how to evaluate and quantify the quality of the detection algorithm is the key to improving the algorithm. Summary of the invention

[0005] The technical problem to be solved by the embodiments of the present invention is to provide a method, device and computer-readable storage medium for evaluating a vehicle drivable area detection algorithm, so as to achieve quantitative evaluation of the drivable area detection algorithm.

[0006] In order to solve the above technical problems, the present invention provides an evaluation method for a vehicle drivable area detection algorithm, comprising the following steps:

[0007] By applying a vehicle drivable area detection algorithm, a vehicle drivable area is obtained;

[0008] The vehicle drivable area is converted into the vehicle coordinate system, and the vehicle drivable area in the vehicle coordinate system is converted into a grid for binarization;

[0009] The binarized grid is mapped to a polar coordinate system, and the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system is used as a detection value, and the error between the detection value and the true value is calculated to obtain the error of the drivable area in the polar coordinate system;

[0010] The error is evaluated according to a preset evaluation dimension.

[0011] Preferably, the converting the vehicle drivable area in the vehicle coordinate system into a grid for binarization specifically includes:

[0012] According to the set rasterization parameters, the vehicle's drivable area in the vehicle coordinate system is converted into a grid;

[0013] The grid contains pixel values ​​of multiple drivable areas with semantic information. If all pixel values ​​in a grid are drivable, the grid is assigned a value of 1; if a pixel value is not drivable, the grid is assigned a value of 0, thereby obtaining a binary grid; the gridding parameters include detection distance and grid size.

[0014] Preferably, mapping the binarized grid to a polar coordinate system specifically includes:

[0015] According to the set polar coordinate parameters, the binary grid in the vehicle coordinate system is divided;

[0016] Take all grid semantic information in each angular resolution, extend the area from near to far according to the distance, until an un-drivable grid appears, and take the position where the grid appears as the value of the angle in the polar coordinate system;

[0017] Traverse the 360° area with the vehicle as the origin and obtain the polar coordinate values ​​respectively, so as to obtain a complete drivable area represented by polar coordinates.

[0018] Preferably, the true value of each angular resolution in the polar coordinate system is obtained by: converting the true value to the vehicle coordinate system, and according to the set rasterization parameters, converting the true value in the vehicle coordinate system into a grid for binarization; mapping the binarized grid to the polar coordinate system to obtain the true value of each angular resolution in the polar coordinate system.

[0019] Preferably, the calculating the error between the detected value and the true value specifically includes:

[0020] For each angle in the polar coordinate system, the same drivable distance between the detected value and the true value within the angle is selected according to the different distances from the vehicle.

[0021] Calculate the difference between two identical drivable distances to obtain the error between the detected value and the true value.

[0022] Preferably, the evaluating the error according to a preset evaluation dimension specifically includes: according to the preset evaluation dimension, setting corresponding indicators or thresholds to classify and evaluate the error, wherein the preset evaluation dimension includes at least one of the following: error size, error direction, error distribution, and error stability.

[0023] Preferably, when the method is applied to an automatic parking scenario, it includes:

[0024] Get the distance between the vehicle and the corner point of the parking space;

[0025] The distances between the vehicle and the corner points of the parking spaces are grouped according to preset distance intervals;

[0026] For each distance interval, a vehicle drivable area detection algorithm is applied to obtain a vehicle drivable area;

[0027] The vehicle drivable area is converted into the vehicle coordinate system, and the vehicle drivable area in the vehicle coordinate system is converted into a grid for binarization;

[0028] The binarized grid is mapped to a polar coordinate system, and the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system is used as a detection value, and the error between the detection value and the true value is calculated to obtain the error of the drivable area in the polar coordinate system;

[0029] The error is evaluated according to a preset evaluation dimension.

[0030] The present invention also provides an evaluation device for a vehicle drivable area detection algorithm, comprising:

[0031] An acquisition module, used for obtaining a vehicle drivable area by applying a vehicle drivable area detection algorithm;

[0032] A rasterization and binarization module, used for converting the vehicle drivable area into the vehicle coordinate system, and converting the vehicle drivable area in the vehicle coordinate system into a grid for binarization;

[0033] An error calculation module is used to map the binarized grid to a polar coordinate system, and use the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system as a detection value, calculate the error between the detection value and the true value, and obtain the error of the drivable area in the polar coordinate system;

[0034] An evaluation module is used to evaluate the error according to a preset evaluation dimension.

[0035] Preferably, the converting the vehicle drivable area in the vehicle coordinate system into a grid for binarization specifically includes:

[0036] According to the set rasterization parameters, the vehicle's drivable area in the vehicle coordinate system is converted into a grid;

[0037] The grid contains pixel values ​​of multiple drivable areas with semantic information. If all pixel values ​​in a grid are drivable, the grid is assigned a value of 1; if a pixel value is not drivable, the grid is assigned a value of 0, thereby obtaining a binary grid; the gridding parameters include detection distance and grid size.

[0038] Preferably, mapping the binarized grid to a polar coordinate system specifically includes:

[0039] According to the set polar coordinate parameters, the binary grid in the vehicle coordinate system is divided;

[0040] Take all grid semantic information in each angular resolution, extend the area from near to far according to the distance, until an un-drivable grid appears, and take the position where the grid appears as the value of the angle in the polar coordinate system;

[0041] Traverse the 360° area with the vehicle as the origin and obtain the polar coordinate values ​​respectively, so as to obtain a complete drivable area represented by polar coordinates.

[0042] Preferably, the true value of each angular resolution in the polar coordinate system is obtained by: converting the true value to the vehicle coordinate system, and according to the set rasterization parameters, converting the true value in the vehicle coordinate system into a grid for binarization; mapping the binarized grid to the polar coordinate system to obtain the true value of each angular resolution in the polar coordinate system.

[0043] Preferably, the calculating the error between the detected value and the true value specifically includes:

[0044] For each angle in the polar coordinate system, the same drivable distance between the detected value and the true value within the angle is selected according to the different distances from the vehicle.

[0045] Calculate the difference between two identical drivable distances to obtain the error between the detected value and the true value.

[0046] Preferably, the evaluating the error according to a preset evaluation dimension specifically includes: according to the preset evaluation dimension, setting corresponding indicators or thresholds to classify and evaluate the error, wherein the preset evaluation dimension includes at least one of the following: error size, error direction, error distribution, and error stability.

[0047] The present invention also provides an evaluation device for a vehicle drivable area detection algorithm, comprising:

[0048] An acquisition module is used to acquire the distance between the vehicle and the corner point of the parking space, group the distance between the vehicle and the corner point of the parking space according to a preset distance interval, and apply a vehicle drivable area detection algorithm to each distance interval to obtain the vehicle drivable area;

[0049] A rasterization and binarization module, used for converting the vehicle drivable area into the vehicle coordinate system, and converting the vehicle drivable area in the vehicle coordinate system into a grid for binarization;

[0050] An error calculation module is used to map the binarized grid to a polar coordinate system, and use the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system as a detection value, calculate the error between the detection value and the true value, and obtain the error of the drivable area in the polar coordinate system;

[0051] An evaluation module is used to evaluate the error according to a preset evaluation dimension.

[0052] The present invention also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the evaluation method of the vehicle drivable area detection algorithm.

[0053] The implementation of the present invention has the following beneficial effects: the present invention proposes a quantitative evaluation method based on binary polar coordinates, which converts the detection results in the self-vehicle coordinate system into polar coordinate values ​​with actual physical meanings, and directly evaluates the actual physical distance difference in the self-vehicle coordinate system. Compared with the traditional pixel coordinate system IoU evaluation method, it can output the quantization error in the physical space in the self-vehicle coordinate system, complete a more objective end-to-end evaluation of the drivable area, and provide the downstream algorithm with an error value in a unified coordinate system; the present invention also adopts the constraints of the spatial relationship of parking spaces according to the parking scene, combines the parking space corner points and the self-vehicle distance factors, so that the evaluation of the drivable area in the automatic parking scene is more reasonable, and effectively combines the characteristics of the parking scene, so that the quantization of automatic parking in the present invention is more multi-dimensional, avoiding the single evaluation index in the whole scene. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 It is a flowchart of an evaluation method of a vehicle drivable area detection algorithm according to an embodiment of the present invention.

[0056] Figure 2 It is a specific flow chart of an evaluation method of a vehicle drivable area detection algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following descriptions of the embodiments refer to the accompanying drawings to illustrate specific embodiments in which the present invention may be implemented.

[0058] Please refer to Figure 1 As shown, the first embodiment of the present invention provides an evaluation method for a vehicle drivable area detection algorithm, comprising the following steps:

[0059] By applying a vehicle drivable area detection algorithm, a vehicle drivable area is obtained;

[0060] The vehicle drivable area is converted into the vehicle coordinate system, and the vehicle drivable area in the vehicle coordinate system is converted into a grid for binarization;

[0061] The binarized grid is mapped to a polar coordinate system, and the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system is used as a detection value, and the error between the detection value and the true value is calculated to obtain the error of the drivable area in the polar coordinate system;

[0062] The error is evaluated according to a preset evaluation dimension.

[0063] Through the above steps, it can be seen that the embodiment of the present invention proposes a quantitative evaluation method based on binary polar coordinates. Compared with the traditional IoU evaluation method of the pixel coordinate system, it can output the quantized error in the physical space under the vehicle coordinate system, complete a more objective end-to-end evaluation of the drivable area, and provide the downstream algorithm with an error value in a unified coordinate system.

[0064] Specifically, please combine Figure 2As shown, the vehicle drivable area detection algorithm analyzes and processes sensor data, identifies elements such as road surfaces, obstacles, lane lines, and determines the area where the vehicle can be safely driven in the current environment based on this information. Different vehicle drivable area detection algorithms can be applied to obtain the vehicle drivable area detected by the detection algorithm. In this embodiment, the vehicle drivable area detected by different vehicle drivable area detection algorithms can be used as input values ​​for subsequent rasterization, binarization, polar coordinate conversion, etc., and compared with the true value converted to the same coordinate system to obtain the error of the drivable area in the polar coordinate system, thereby obtaining a quantitative evaluation result.

[0065] Taking the vehicle drivable area obtained by applying a certain vehicle drivable area detection algorithm as an example, first, according to the calibration parameters, the detected drivable area is converted into the vehicle coordinate system. It can be understood that in this embodiment, the calibration parameters refer to camera calibration parameters or sensor calibration parameters, which are used to ensure that the information obtained from the camera or sensor can be accurately mapped to the real world. Camera calibration parameters usually include intrinsic parameters (used to describe the characteristics of the camera itself, including focal length, image center, radial and tangential distortion coefficients, etc.) and extrinsic parameters (used to describe the position and orientation of the camera, usually including rotation and translation matrices); sensor calibration parameters usually include offsets, rotation matrices, and scale factors, etc.

[0066] Then, according to the set rasterization parameters, the drivable area in the vehicle coordinate system is converted into a grid, that is, the original drivable area data is converted into grid data. Rasterization is a common data processing method that can convert continuous spatial data into discrete grid data. The rasterization parameters include detection distance and grid size. The detection distance defines how far the area to be detected is, and the grid size defines the size of each grid. For example, if the detection distance is set to 100 meters and the grid size is 1 meter, a 100×100 grid map will be obtained. The grid contains multiple pixel values ​​of drivable areas with semantic information. If all the pixel values ​​in a grid are drivable, then the grid is marked as drivable and the grid is assigned a value of 1; if there are pixel values ​​that are not drivable, the grid is marked as not drivable and the grid is assigned a value of 0. Thus, the binary grid is obtained, which is convenient for subsequent calculations and processing.

[0067] The following further introduces a specific method of converting the binary grid in the vehicle coordinate system into polar coordinate representation in an embodiment of the present invention.

[0068] First, according to the set polar coordinate parameters, the binary grid in the self-vehicle coordinate system is divided to realize the conversion from the self-vehicle coordinate system to the polar coordinate system. In this embodiment, the polar coordinate parameters include the detection distance parameter and the angular resolution parameter of the polar coordinates. The detection distance parameter defines the maximum distance that the polar coordinates can represent, and the angular resolution defines the accuracy of the polar coordinates in terms of angle. For example, if the detection distance is set to 100 meters and the angular resolution is 1°, the polar coordinates obtained will contain 360 angles, each angle corresponding to a distance value.

[0069] Then, take all the grid semantic information in each angular resolution and extend the area from near to far according to the distance. In each angular resolution, start from the starting point of the vehicle and extend outward along this angle. If a grid is drivable, continue to extend outward; if a grid that is not drivable appears, stop extending and take the position where the grid that is not drivable appears as the value of this angle in the polar coordinate system, that is, the polar coordinate value.

[0070] Traverse the 360° area with the vehicle as the origin and obtain the polar coordinate values ​​respectively, so as to obtain a complete drivable area represented by polar coordinates.

[0071] Through the above process, a simplified and easy-to-handle polar coordinate map representing the drivable area can be obtained, which has good directionality and visibility.

[0072] Next, the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system is taken as the detection value and compared with the true value. The true value can be the vehicle drivable area determined based on the ground real environment or reference information such as high-precision maps, or it can be manually marked; the true value represents the optimal vehicle drivable area theoretically expected under the same circumstances. It should be noted that in this embodiment, the true value is also subjected to the aforementioned rasterization, binarization and polar coordinate processing, that is, the true value is converted to the vehicle coordinate system, and according to the set rasterization parameters, the true value in the vehicle coordinate system is converted into a grid for binarization; the binarized grid is mapped to the polar coordinate system to obtain the true value of each angular resolution in the polar coordinate system.

[0073] The specific way to calculate the error between the detected value and the true value is:

[0074] For each angle in the polar coordinate system, the same drivable distance of the detected value and the true value within the angle is selected according to the difference in distance from the vehicle (e.g., the closest drivable distance, the farthest drivable distance). It should be noted that for the edge point of the drivable distance, if the polar coordinate of the point is still in the drivable area, the point is not calculated as an error.

[0075] Next, the difference between two identical drivable distances is calculated to obtain the error between the detected value and the true value, which is used as the error of the drivable area in the polar coordinate system.

[0076] Taking the nearest drivable distance as an example, the nearest drivable distance is the shortest distance from the starting point (i.e. the current position of the vehicle) to the drivable area in this angle direction. The difference between the two nearest drivable distances is calculated to obtain the error between the detected value and the true value.

[0077] After obtaining the error of the drivable area in the polar coordinate system, this embodiment evaluates the error according to the preset evaluation dimension. The evaluation dimension may include error size, error direction, error distribution, error stability, etc., and corresponding indicators or thresholds may be set to classify and evaluate the errors, for example, errors less than a certain threshold are divided into accurate areas, errors within a certain range are divided into acceptable areas, and large errors are divided into inaccurate areas.

[0078] Furthermore, this embodiment also evaluates the error of the drivable area in the automatic parking scenario in combination with the spatial relationship of the parking space. Specifically, first, the distance between the vehicle and the corner point of the parking space is obtained by calculating the geometric distance between the vehicle position and the corner point of the parking space; then, the distance between the vehicle and the corner point of the parking space is grouped according to the preset distance intervals, such as short distance, medium distance and long distance; for each distance interval, the vehicle drivable area detection algorithm is applied, and the detected vehicle drivable area is converted to the vehicle coordinate system according to the method described in the previous steps, and converted into a grid for binarization; then, the binarized grid is mapped to the polar coordinate system, and the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system is calculated as the detection value; the error between the detection value and the true value is calculated. The error can be evaluated according to the aforementioned evaluation dimensions, and the error conditions in different distance intervals can also be compared, and it can be determined which intervals have large errors, as well as possible error patterns or trends.

[0079] By setting different intervals according to the distance between the vehicle and the corner point of the parking space and evaluating the error of the vehicle's drivable area detection in each interval, we can have a more comprehensive understanding of the algorithm's performance in different distance ranges, help discover and improve problems within a specific distance range, and provide optimization strategies and algorithm improvement suggestions for automatic parking scenarios, avoiding the single evaluation indicator in all scenarios.

[0080] Corresponding to the evaluation method of the vehicle drivable area detection algorithm described in the first embodiment of the present invention, the second embodiment of the present invention provides an evaluation device for the vehicle drivable area detection algorithm, including:

[0081] An acquisition module, used for obtaining a vehicle drivable area by applying a vehicle drivable area detection algorithm;

[0082] A rasterization and binarization module, used for converting the vehicle drivable area into the vehicle coordinate system, and converting the vehicle drivable area in the vehicle coordinate system into a grid for binarization;

[0083] An error calculation module is used to map the binarized grid to a polar coordinate system, and use the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system as a detection value, calculate the error between the detection value and the true value, and obtain the error of the drivable area in the polar coordinate system;

[0084] An evaluation module is used to evaluate the error according to a preset evaluation dimension.

[0085] Preferably, the converting the vehicle drivable area in the vehicle coordinate system into a grid for binarization specifically includes:

[0086] According to the set rasterization parameters, the vehicle's drivable area in the vehicle coordinate system is converted into a grid;

[0087] The grid contains pixel values ​​of multiple drivable areas with semantic information. If all pixel values ​​in a grid are drivable, the grid is assigned a value of 1; if a pixel value is not drivable, the grid is assigned a value of 0, thereby obtaining a binary grid; the gridding parameters include detection distance and grid size.

[0088] Preferably, mapping the binarized grid to a polar coordinate system specifically includes:

[0089] According to the set polar coordinate parameters, the binary grid in the vehicle coordinate system is divided;

[0090] Take all grid semantic information in each angular resolution, extend the area from near to far according to the distance, until an un-drivable grid appears, and take the position where the grid appears as the value of the angle in the polar coordinate system;

[0091] Traverse the 360° area with the vehicle as the origin and obtain the polar coordinate values ​​respectively, so as to obtain a complete drivable area represented by polar coordinates.

[0092] Preferably, the true value of each angular resolution in the polar coordinate system is obtained by: converting the true value to the vehicle coordinate system, and according to the set rasterization parameters, converting the true value in the vehicle coordinate system into a grid for binarization; mapping the binarized grid to the polar coordinate system to obtain the true value of each angular resolution in the polar coordinate system.

[0093] Preferably, the calculating the error between the detected value and the true value specifically includes:

[0094] For each angle in the polar coordinate system, the same drivable distance between the detected value and the true value within the angle is selected according to the different distances from the vehicle.

[0095] Calculate the difference between two identical drivable distances to obtain the error between the detected value and the true value.

[0096] Preferably, the evaluating the error according to a preset evaluation dimension specifically includes: according to the preset evaluation dimension, setting corresponding indicators or thresholds to classify and evaluate the error, wherein the preset evaluation dimension includes at least one of the following: error size, error direction, error distribution, and error stability.

[0097] Corresponding to the evaluation method of the vehicle drivable area detection algorithm described in the first embodiment of the present invention, the third embodiment of the present invention provides an evaluation device for the vehicle drivable area detection algorithm, including:

[0098] An acquisition module is used to acquire the distance between the vehicle and the corner point of the parking space, group the distance between the vehicle and the corner point of the parking space according to a preset distance interval, and apply a vehicle drivable area detection algorithm to each distance interval to obtain the vehicle drivable area;

[0099] A rasterization and binarization module, used for converting the vehicle drivable area into the vehicle coordinate system, and converting the vehicle drivable area in the vehicle coordinate system into a grid for binarization;

[0100] An error calculation module is used to map the binarized grid to a polar coordinate system, and use the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system as a detection value, calculate the error between the detection value and the true value, and obtain the error of the drivable area in the polar coordinate system;

[0101] An evaluation module is used to evaluate the error according to a preset evaluation dimension.

[0102] Corresponding to the evaluation method of the vehicle drivable area detection algorithm described in the aforementioned embodiment 1 of the present invention, embodiment 4 of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located controls the evaluation method of the vehicle drivable area detection algorithm as described in embodiment 1 of the present invention.

[0103] Preferably, the computer program can be divided into one or more modules / units (such as computer program 1, computer program 2, ...), and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the device.

[0104] The processor may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may be any conventional processor. The processor is the control center of the device, and various parts of the device are connected using various interfaces and lines.

[0105] The memory mainly includes a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function, etc., and the data storage area can store related data, etc. In addition, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, and a flash card (Flash Card), etc., or the memory can also be other volatile solid-state storage devices.

[0106] It should be noted that the above-mentioned device may include but is not limited to a processor and a memory, which can be understood by those skilled in the art.

[0107] For the working principle and process of the above embodiment, please refer to the description of the above embodiment of the present invention, which will not be repeated here.

[0108] It can be seen from the above description that compared with the prior art, the beneficial effects of the present invention are as follows: the present invention proposes a quantitative evaluation method based on binary polar coordinates, which converts the detection results in the self-vehicle coordinate system into polar coordinate values ​​with actual physical meanings, and directly evaluates the actual physical distance difference in the self-vehicle coordinate system. Compared with the traditional pixel coordinate system IoU evaluation method, the quantization error in the physical space in the self-vehicle coordinate system can be output, and a more objective end-to-end evaluation of the drivable area is completed, providing the downstream algorithm with an error value in a unified coordinate system; the present invention also adopts the constraints of the spatial relationship of parking spaces according to the parking scene, combined with the parking space corner points and the self-vehicle distance factors, so that the evaluation of the drivable area in the automatic parking scene is more reasonable, and the characteristics of the parking scene are effectively combined, so that the quantization of automatic parking in the present invention is more multi-dimensional, and the single evaluation index in the whole scene is avoided.

[0109] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for evaluating a vehicle drivable area detection algorithm, characterized in that: The following steps are involved: By applying a vehicle drivable area detection algorithm, a vehicle drivable area is obtained; The vehicle drivable area is converted into the vehicle coordinate system, and the vehicle drivable area in the vehicle coordinate system is converted into a grid for binarization; The binarized grid is mapped to a polar coordinate system, and the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system is used as a detection value, and the error between the detection value and the true value is calculated to obtain the error of the drivable area in the polar coordinate system; The error is evaluated according to a preset evaluation dimension.

2. The method according to claim 1, characterized in that The converting of the vehicle drivable area in the vehicle coordinate system into a grid for binarization specifically includes: According to the set rasterization parameters, the vehicle's drivable area in the vehicle coordinate system is converted into a grid; The grid contains pixel values ​​of multiple drivable areas with semantic information. If all pixel values ​​in a grid are drivable, the grid is assigned a value of 1; if a pixel value is not drivable, the grid is assigned a value of 0, thereby obtaining a binary grid; the gridding parameters include detection distance and grid size.

3. The method according to claim 1, characterized in that Mapping the binarized grid to the polar coordinate system specifically includes: According to the set polar coordinate parameters, the binary grid in the vehicle coordinate system is divided; Take all grid semantic information in each angular resolution, extend the area from near to far according to the distance, until an un-drivable grid appears, and take the position where the grid appears as the value of the angle in the polar coordinate system; Traverse the 360° area with the vehicle as the origin and obtain the polar coordinate values ​​respectively, so as to obtain a complete drivable area represented by polar coordinates.

4. The method according to claim 1, characterized in that The true value of each angular resolution in the polar coordinate system is obtained by converting the true value to the ego-vehicle coordinate system, and according to the set rasterization parameters, converting the true value in the ego-vehicle coordinate system into a grid for binarization; The binarized grid is mapped to the polar coordinate system to obtain the true value of each angular resolution in the polar coordinate system.

5. The method according to claim 1, characterized in that The calculating the error between the detected value and the true value specifically includes: For each angle in the polar coordinate system, the same drivable distance between the detected value and the true value within the angle is selected according to the different distances from the vehicle. Calculate the difference between two identical drivable distances to obtain the error between the detected value and the true value.

6. The method according to claim 5, characterized in that The evaluating the error according to the preset evaluation dimension specifically includes: according to the preset evaluation dimension, setting corresponding indicators or thresholds to classify and evaluate the error, wherein the preset evaluation dimension includes at least one of the following: error size, error direction, error distribution, and error stability.

7. The method according to claim 1, characterized in that When the method is applied to an automatic parking scenario, it includes: Get the distance between the vehicle and the corner point of the parking space; The distances between the vehicle and the corner points of the parking spaces are grouped according to preset distance intervals; For each distance interval, a vehicle drivable area detection algorithm is applied to obtain a vehicle drivable area; The vehicle drivable area is converted into the vehicle coordinate system, and the vehicle drivable area in the vehicle coordinate system is converted into a grid for binarization; The binarized grid is mapped to a polar coordinate system, and the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system is used as a detection value, and the error between the detection value and the true value is calculated to obtain the error of the drivable area in the polar coordinate system; The error is evaluated according to a preset evaluation dimension.

8. An evaluation device for a vehicle drivable area detection algorithm, characterized in that: include: An acquisition module, used for obtaining a vehicle drivable area by applying a vehicle drivable area detection algorithm; A rasterization and binarization module, used for converting the vehicle drivable area into the vehicle coordinate system, and converting the vehicle drivable area in the vehicle coordinate system into a raster for binarization; An error calculation module is used to map the binarized grid to a polar coordinate system, and use the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system as a detection value, calculate the error between the detection value and the true value, and obtain the error of the drivable area in the polar coordinate system; An evaluation module is used to evaluate the error according to a preset evaluation dimension.

9. An evaluation device for a vehicle drivable area detection algorithm, characterized in that: include: An acquisition module is used to acquire the distance between the vehicle and the corner point of the parking space, group the distance between the vehicle and the corner point of the parking space according to a preset distance interval, and apply a vehicle drivable area detection algorithm to each distance interval to obtain the vehicle drivable area; A rasterization and binarization module, used for converting the vehicle drivable area into the vehicle coordinate system, and converting the vehicle drivable area in the vehicle coordinate system into a raster for binarization; An error calculation module is used to map the binarized grid to a polar coordinate system, and use the distance between the drivable area and the vehicle at each angular resolution in the polar coordinate system as a detection value, calculate the error between the detection value and the true value, and obtain the error of the drivable area in the polar coordinate system; An evaluation module is used to evaluate the error according to a preset evaluation dimension.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program; wherein, when the computer program is run, it controls the device where the computer-readable storage medium is located to execute the evaluation method of the vehicle drivable area detection algorithm according to any one of claims 1 to 7.

Images