Noisy point determination method, electronic equipment and storage medium

Abstract

The invention discloses a noisy point determination method, electronic equipment and a storage medium, and the method comprises the steps: carrying out the first visual angle projection and second visual angle projection of a target point cloud, and correspondingly obtaining a first projection image and a second projection image; clustering the target point clouds to obtain at least one clustered point cloud; determining a first judgment index of each clustering point cloud based on the first projection image; determining a second judgment index of each clustering point cloud based on the second projection image; and at least combining the first judgment index and the second judgment index of each clustering point cloud to determine whether each clustering point cloud belongs to a noisy point. In this way, the noisy point determination accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a noise determination method, electronic device and storage medium. Background Art

[0002] The point cloud data collected by the LiDAR generally contains noise points. Take the LiDAR on the vehicle as an example. During the driving process of the vehicle, the LiDAR will be used to provide safe and reliable distance perception. However, in extreme weather conditions such as rainy or snowy days, raindrops or snowflakes in the air will also return to the point cloud after being scanned by the laser beam emitted by the LiDAR. Especially when there is water on the ground, the tires running at high speed will form a dense water vapor group behind the car, forming a noise point cloud. These noise point clouds will form clustered targets and be mistakenly detected as obstacles, resulting in wrong decisions by the vehicle's planning and control modules, which will bring greater safety risks to the entire autonomous driving system. Therefore, how to accurately detect noise points is one of the problems to be solved. Summary of the invention

[0003] The main technical problem solved by the present application is to provide a noise point determination method, an electronic device and a storage medium, which can improve the accuracy of noise point determination.

[0004] To solve the above technical problems, the first aspect of the present application provides a noise determination method, which includes: performing first-perspective projection and second-perspective projection on the target point cloud respectively, and obtaining a first projection image and a second projection image correspondingly; clustering the target point cloud to obtain at least one cluster point cloud; based on the first projection image, determining a first judgment index for each cluster point cloud; and based on the second projection image, determining a second judgment index for each cluster point cloud; and determining whether each cluster point cloud is a noise point by at least combining the first judgment index and the second judgment index of each cluster point cloud.

[0005] The first projection image includes first projection points of at least some of the spatial points in the target point cloud.

[0006] Wherein, based on the first projection image, determining the first judgment index of each cluster point cloud includes: rasterizing the first projection image to obtain a first raster image; obtaining the initial raster index of each raster in the first raster image; wherein the initial raster index includes the number of first projection points in the raster, and at least one of the intensity mean and the distance mean of each first projection point in the raster, and the distance mean is the mean of each first distance between the spatial point corresponding to each first projection point and the preset origin; respectively taking each raster in the first raster image as the first target raster, and taking the initial raster index of the first target raster as the first judgment index of the spatial point corresponding to each first projection point in the first target raster; or, determining the final raster index of the first target raster based on the initial raster index of the first target raster and the initial raster index of the neighborhood raster of the first target raster; taking the final raster index of the first target raster as the first judgment index of the spatial point corresponding to each first projection point in the first target raster.

[0007] The final grid index includes target intensity and spatial dispersion, and the spatial dispersion includes at least one of spatial distance uniformity, spatial quantity uniformity, and density.

[0008] Wherein, based on the initial grid index of the first target grid and the initial grid index of the neighborhood grid of the first target grid, the final grid index of the target grid is determined, including: averaging the intensity means included in the initial grid index of the first target grid and the initial grid index of the neighborhood grid to obtain the target intensity; calculating the variance of the distance means included in the initial grid index of the first target grid and the initial grid index of the neighborhood grid to obtain the spatial distance uniformity; calculating the variance of the number of first projection points in the first target grid and the neighborhood grid to obtain the spatial number uniformity; dividing the total number of first projection points in the first target grid and the neighborhood grid by the number of the target grid and the neighborhood grid to obtain the density.

[0009] The second judgment index includes an average penetration rate, and the second projection image includes second projection points of at least part of the spatial points in the target point cloud.

[0010] Wherein, based on the second projection image, determining the second judgment index of each cluster point cloud includes: performing rasterization processing on the second projection image to obtain a second raster image; taking each grid covered by the cluster point cloud as a second target grid, and obtaining the distance difference between the maximum distance and the minimum distance corresponding to each second target grid; wherein the maximum distance and the minimum distance are respectively the maximum value and the minimum value in the second distance between the spatial point corresponding to each second projection point in the second target grid and the preset origin; averaging each distance difference to obtain an average penetration rate.

[0011] The target grid is the grid where the projection points exist in the grid image.

[0012] After clustering the target point cloud to obtain at least one clustered point cloud, the method further includes: determining a third judgment index of the clustered point cloud, wherein the third judgment index includes a linear correlation index and / or a spatial autocorrelation index.

[0013] Among them, at least the first judgment indicator and the second judgment indicator of each cluster point cloud are combined to determine whether each cluster point cloud is a noise point, including: for each cluster point cloud, the first judgment indicator, the second judgment indicator, and the third judgment indicator of the cluster point cloud are combined to determine whether the cluster point cloud is a noise point.

[0014] Among them, determining the linear correlation index of the cluster point cloud includes: fitting the coordinates of each spatial point in the cluster point cloud using the least squares method to obtain the fitting straight line of the cluster point cloud; calculating the mean of the distance between each spatial point in the cluster point cloud and the fitting straight line to obtain the linear correlation index.

[0015] Wherein, determining the spatial autocorrelation index of the clustered point cloud includes: obtaining the third distance between each spatial point in the clustered point cloud and a preset origin, and obtaining the mean of the third distances corresponding to each spatial point in the clustered point cloud as the distance mean of the clustered point cloud; and obtaining the difference between the third distance corresponding to each spatial point in the clustered point cloud and the distance mean; taking each spatial point in the clustered point cloud as a candidate point, respectively, multiplying the difference corresponding to the candidate point with the difference corresponding to any other point and dividing it by the distance between the candidate point and the other point to obtain a first parameter of the candidate point; taking the square of the difference corresponding to the candidate point as the second parameter of the candidate point; and determining the linear correlation index of the clustered point cloud based on the first parameter of each candidate point, the second parameter, the number of candidate points and the autocorrelation weight; wherein the autocorrelation weight is related to the distance between the candidate point and the other point.

[0016] Among them, combining the first judgment index, the second judgment index and the third judgment index of the cluster point cloud to determine whether the cluster point cloud belongs to a noise point, including: performing weighted summation on the first judgment index, the second judgment index and the third judgment index to obtain a total judgment index; based on the total judgment index, determining whether the cluster point cloud belongs to a noise point.

[0017] Among them, clustering the target point cloud to obtain at least one cluster point cloud includes: performing a third-perspective projection on the target point cloud to obtain a third projection image, the third projection image containing third projection points of at least some of the spatial points in the target point cloud; obtaining a fourth distance between a spatial point corresponding to each third projection point in the third projection image and a preset origin; and determining whether each third projection point belongs to the same cluster point cloud based on a difference in the fourth distance corresponding to each third projection point.

[0018] Among them, the first perspective is a bird's-eye view, and the second perspective is a front view.

[0019] Among them, the noise determination method also includes: obtaining an initial point cloud; preprocessing the initial point cloud to obtain a target point cloud; wherein the preprocessing includes downsampling and / or filtering out point clouds in the initial point cloud that belong to a preset area, and the preset area includes the ground.

[0020] To solve the above technical problems, the second aspect of the present application provides an electronic device, which includes a memory and a processor, wherein the memory stores program instructions; the processor is used to execute the program instructions stored in the memory to implement the method provided by the first aspect above.

[0021] In order to solve the above technical problem, the third aspect of the present application provides a computer-readable storage medium, which is used to store program instructions, and the program instructions can be executed by a processor to implement the method provided by the first aspect above.

[0022] The beneficial effects of the present application are as follows: Different from the prior art, the present application projects the target point cloud from the first perspective and the second perspective respectively, obtains the first projection image and the second projection image accordingly, and clusters the target point cloud to obtain at least one cluster point cloud; based on the first projection image, determines the first judgment index of each cluster point cloud; and based on the second projection image, determines the second judgment index of each cluster point cloud; at least combines the first judgment index and the second judgment index of each cluster point cloud to determine whether each cluster point cloud is a noise point. By combining multiple perspectives to analyze and judge the cluster point cloud, and combining the judgment indexes of multiple perspectives, it is determined whether the cluster point cloud is a noise point, so as to improve the accuracy of noise point determination. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flowchart of an implementation method of a noise point determination method provided by the present application;

[0024] Figure 2 is a flow chart of another implementation of the noise point determination method provided by the present application;

[0025] Figure 3 It is a schematic diagram of the framework structure of an embodiment of an electronic device provided by the present application;

[0026] Figure 4 It is a schematic diagram of the framework structure of an implementation scheme of a computer-readable storage medium provided in this application. DETAILED DESCRIPTION

[0027] The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] It should be noted that in the embodiments of the present application, there are descriptions involving "first", "second", etc., which are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features.

[0029] In the description of the application, the meaning of "multiple" is at least two, for example, two, three, etc., unless otherwise clearly and specifically limited. In the embodiment of the present application, all directional indications (such as up, down, left, right, front, back ...) are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their deformations are intended to cover non-exclusive inclusions. For example, the process, method, system, product or equipment comprising a series of steps or units is not limited to the steps or units listed, but optionally also includes the steps or units not listed, or optionally also includes other steps or units inherent to these processes, methods, products or equipment.

[0030] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0031] The present application is used to determine whether a clustered point cloud in a target point cloud belongs to a noise point cloud. The target point cloud may be acquired by a laser radar. The laser radar is mainly composed of a laser emitting unit and a laser receiving unit. The laser emitting unit emits a laser with a wavelength of 850nm to 1550nm. When the laser is transmitted through the air and irradiated onto an object, it returns to the laser receiving unit after diffuse reflection.

[0032] See also Figure 1 , Figure 1 : is a flow chart of an embodiment of a noise determination method provided by the present application, the method comprising:

[0033] S11: Perform first-view projection and second-view projection on the target point cloud respectively, and obtain a first projection image and a second projection image accordingly.

[0034] In one embodiment, the target point cloud may be an initial point cloud collected by a laser radar on a vehicle. In another embodiment, the target point cloud may be obtained by preprocessing the initial point cloud, and the preprocessing may include downsampling, coordinate conversion, and filtering out point clouds in the initial point cloud that belong to a preset area. Among them, the initial point cloud may be downsampled by a voxel sampling method to reduce the amount of data of the target point cloud, so as to reduce the amount of data for subsequent processing. Coordinate conversion is to convert spatial points from a world coordinate system to a vehicle coordinate system to perform relevant calculations in the vehicle coordinate system. The preset area includes at least one of the ground and an irrelevant area. The irrelevant area may be an area outside the preset area. Since the area collected by the laser radar is far away, and the data in the too far area is less accurate and not referenceable, a preset area may be set to filter out point clouds in the initial point cloud that are outside the preset area. In one embodiment, the noise points judged by the present application belong to rain and snow noise points. At this time, it is only necessary to judge whether the point cloud in the air belongs to rain and snow noise points, without judging the point cloud on the ground. Therefore, a ground filtering algorithm can be used to filter out the point cloud located on the ground in the initial point cloud. The ground filtering algorithm includes but is not limited to RANSAC, CSF, and PCL.

[0035] After acquiring the target point cloud, the target point cloud can be projected from multiple perspectives to analyze the point cloud data from multiple perspectives. In one embodiment, the target point cloud can be projected from a first perspective and a second perspective to obtain a first projection image and a second projection image. For any three-dimensional point in the target point cloud, there is a corresponding projection point in each projection image. The first perspective can be a top-down perspective, and the second perspective can be a front-view perspective. It can be understood that in other embodiments, the first perspective and the second perspective can also be other perspectives.

[0036] S12: Clustering the target point cloud to obtain at least one clustered point cloud.

[0037] In one implementation, an existing point cloud clustering algorithm may be used to cluster the target point cloud, and the point cloud clustering algorithm includes but is not limited to a k-means algorithm, a Gaussian mixture clustering, a hierarchical clustering, and DBSCAN.

[0038] In another embodiment, the target point cloud may be projected onto a two-dimensional plane, and the positional relationship of each projection point in the two-dimensional plane and the distance of the target point cloud in three-dimensional space may be used to determine whether each target point cloud belongs to the same cluster point cloud.

[0039] In a specific embodiment, the target point cloud is projected from a third perspective to obtain a third projection image, which includes the third projection points of at least some of the spatial points in the target point cloud; the fourth distance between the spatial point corresponding to each third projection point in the third projection image and the preset origin is obtained; based on the difference in the fourth distance corresponding to each third projection point, it is determined whether each third projection point belongs to the same cluster point cloud. Among them, the third perspective can be a front-view perspective, and any third projection point in the third projection image corresponds to a spatial point in the target point cloud. For any two third projection points, the spatial points corresponding to them are obtained, and the fourth distance between each spatial point and the preset origin is calculated. According to the relationship between the difference between the two fourth distances and the preset difference, it is determined whether the two third projection points belong to the same cluster target, that is, whether the two spatial points belong to the same cluster target. If the difference between the two fourth distances is less than the preset difference, it can be determined whether the two third projection points belong to the same cluster target. By traversing all third projection points, clustering of the target point cloud can be completed.

[0040] S13: Determine a first judgment index for each cluster point cloud based on the first projection image; and determine a second judgment index for each cluster point cloud based on the second projection image.

[0041] In one embodiment, different judgment indicators can be determined from multiple perspectives. The first judgment indicator can include the intensity and spatial discreteness of the spatial point, and the characteristics of each point cloud can be analyzed from the spatial point level. The second judgment indicator can include the average penetration rate. When the laser is irradiated on the water vapor group, part of the light beam is reflected back to the receiver by the water vapor to form noise points, and the other part penetrates the water vapor to irradiate the real target, so the penetration rate can be used as a judgment indicator.

[0042] In a specific embodiment, the first projection image includes the first projection points of at least some of the spatial points in the target point cloud. In order to further reduce the amount of calculation, the projection image can be rasterized. Specifically, the first projection image is rasterized to obtain a first raster image; the initial raster index of each grid in the first raster image is obtained; wherein the initial raster index includes the number of first projection points in the grid, and at least one of the intensity mean and the distance mean of each first projection point in the grid, and the distance mean is the mean of each first distance between the spatial point corresponding to each first projection point and the preset origin; each grid in the first raster image is respectively used as the first target grid, and the initial raster index of the first target grid is used as the first judgment index of the spatial point corresponding to each first projection point in the first target grid; or, based on the initial raster index of the first target grid and the initial raster index of the neighborhood grid of the first target grid, the final raster index of the first target grid is determined; the final raster index of the first target grid is used as the first judgment index of the spatial point corresponding to each first projection point in the first target grid.

[0043] In the rasterization process, the rasterization range can be set. For example, the area 70 meters in front of the vehicle and 20 meters to the left and right of the vehicle can be rasterized, and the area in the first projection image that belongs to the rasterization range can be rasterized. When rasterizing, the rasterized area resolution can be set to a 10 cm grid, so that a 700*400 raster image can be obtained. After the first projection image is rasterized to obtain the first raster image, it can be determined in which grid each first projection point in the first projection image is located, and the corresponding relationship between the spatial point and the grid index can be obtained.

[0044] Further, for each grid in the first grid image, the number of first projection points in the grid is calculated, and the intensity mean and distance mean of each first projection point in the grid are calculated. The intensity mean of each first projection point is the mean of the intensity of the space point corresponding to each first projection point. The intensity of the space point is the reflection intensity value returned by the space point when the laser radar sensor scans the environment. The distance mean of each first projection point is the mean of the first distances between the space point corresponding to each first projection point and the preset origin. The preset origin is set by the user, for example, it can be the center point of the rear axle of the vehicle. In one embodiment, the coordinates of the preset origin are (0, 0, 0), and the coordinates obtained by converting the coordinates of the space point are (x, y, z), then the first distance is In this way, the initial grid index of each grid in the first grid image can be obtained. The coordinate transformation of the space point is to transform the space point into the vehicle coordinate system.

[0045] In a specific implementation, the initial grid index of each grid may be used as the first judgment index of the space point corresponding to each first projection point in each grid.

[0046] In another embodiment, the final grid index of each grid can be determined by combining the initial grid index of each grid and the initial grid index of the neighborhood grid of each grid; the final grid index of each grid is used as the first judgment index of the spatial point corresponding to each first projection point in each grid. Among them, the neighborhood grid is the grid within a preset range around the first target grid as the intermediate grid. The neighborhood size can be set as needed and is not limited here.

[0047] Among them, the final grid index may include target intensity and spatial discreteness, and the spatial discreteness includes at least one of spatial distance uniformity, spatial quantity uniformity, and density. Each grid in the first grid image is respectively used as the first target grid, and the initial grid index of the first target grid and the initial grid index of the neighborhood grid are averaged to obtain the target intensity. The specific calculation is shown in the following formula 1. The variance of the distance means included in the initial grid index of the first target grid and the initial grid index of the neighborhood grid is calculated to obtain the spatial distance uniformity. The specific calculation is shown in the following formula 2. The variance of the number of first projection points in the first target grid and the neighborhood grid is calculated to obtain the spatial quantity uniformity. The specific calculation is shown in the following formula 3. The total number of first projection points in the first target grid and the neighborhood grid is divided by the number of target grids and neighborhood grids to obtain density. The specific calculation is shown in the following formula 4.

[0048]

[0049] Among them, intensity is the target intensity, K is the number of the first target grid and the neighboring grid, int i is the mean intensity of the first projection point of the first target grid and the i-th grid in the neighborhood grid; uniform s is the spatial distance uniformity, dis i is the average of the first distances between the spatial points corresponding to the first projection points in the i-th grid and the preset origin, dis ave is K dis i The mean of uniform num is the spatial quantity uniformity, num i is the number of first projection points in the i-th grid, dis ave is K num i The mean of ; density is the density.

[0050] In one embodiment, after obtaining the first judgment index of each spatial point, it is also possible to directly determine whether each spatial point is a noise point. For example, it is possible to determine whether the first judgment index reaches a threshold, and count the number of spatial points belonging to a cluster point cloud whose first judgment index reaches the threshold. If the number exceeds a preset number, it is considered that the cluster point cloud belongs to a noise point.

[0051] In another specific embodiment, the second judgment index includes an average penetration rate, and the second projection image includes second projection points of at least some of the spatial points in the target point cloud; the second projection image is rasterized to obtain a second raster image; the rasterization process here is similar to the process of rasterizing the first projection image, and will not be repeated here. Each grid covered by the cluster point cloud is used as the second target grid, and the distance difference between the maximum distance and the minimum distance corresponding to each second target grid is obtained; wherein the maximum distance and the minimum distance are respectively the maximum and minimum values ​​of the second distance between the spatial point corresponding to each second projection point in the second target grid and the preset origin; the average of each distance difference is calculated to obtain the average penetration rate. The specific calculation process of the average penetration rate is shown in the following formula 5.

[0052]

[0053] Where penetration is the average penetration rate, K is the number of second target grids covered by a cluster point cloud, is the maximum distance corresponding to the i-th second target grid, is the minimum distance corresponding to the i-th second target grid.

[0054] The first target grid and the second target grid are both valid grids in the projection image. The valid grid is the grid with projection points, that is, the grid with the first projection point in the first projection image is the first target grid, and the grid with the second projection point in the second projection image is the second target grid.

[0055] S14: Determine whether each cluster point cloud is a noise point by at least combining the first judgment index and the second judgment index of each cluster point cloud.

[0056] In one embodiment, the total judgment index of the cluster point cloud can be determined based on the first judgment index and the second judgment index, and then whether the cluster point cloud belongs to noise points can be determined based on the total judgment index. For example, the first judgment index and the second judgment index can be weighted and summed to obtain the total judgment index. When the first judgment index and the second judgment index contain multiple indicators, a corresponding weight can be set for each indicator, and all indicators are weighted and summed to obtain the total judgment index. Whether the cluster point cloud belongs to noise points is determined based on the size relationship between the total judgment index and the preset indicator threshold. For example, when the total judgment index of the cluster point cloud is greater than the preset indicator threshold, it is determined that the cluster point cloud belongs to noise points.

[0057] In another embodiment, the total judgment index of the cluster point cloud is determined based on the first judgment index and the second judgment index. This is not limited to weighted summation, and the first judgment index and the second judgment index may be directly summed to determine the total judgment index.

[0058] In other implementations, a first indicator threshold and a second indicator threshold may also be set. When the first judgment indicator of the clustering target meets the first indicator threshold and the second judgment indicator meets the second indicator threshold, it is determined that the clustering target belongs to a noise point.

[0059] In the above method, the target point cloud is projected from the first perspective and the second perspective respectively, and the first projection image and the second projection image are obtained correspondingly, and the target point cloud is clustered to obtain at least one cluster point cloud; based on the first projection image, the first judgment index of each cluster point cloud is determined; and based on the second projection image, the second judgment index of each cluster point cloud is determined; at least the first judgment index and the second judgment index of each cluster point cloud are combined to determine whether each cluster point cloud belongs to a noise point. By analyzing and judging the cluster point cloud in combination with multiple perspectives, the judgment indicators of multiple perspectives are combined to determine whether the cluster point cloud belongs to a noise point, so as to improve the accuracy of noise point determination. In addition, the present application judges whether the entire cluster point cloud belongs to a noise point from the perspective of the cluster point cloud, rather than judging whether each spatial point in the point cloud belongs to a noise point, so that it can be controlled as a whole to avoid wrong judgment and deletion of spatial points that do not belong to noise points.

[0060] See also Figure 2 , Figure 2 : is a flow chart of another embodiment of the noise determination method provided by the present application, the method comprising:

[0061] S21: Perform first-view projection and second-view projection on the target point cloud respectively, and obtain a first projection image and a second projection image accordingly.

[0062] S22: Clustering the target point cloud to obtain at least one clustered point cloud.

[0063] S23: Determine a first judgment index for each cluster point cloud based on the first projection image; and determine a second judgment index for each cluster point cloud based on the second projection image.

[0064] For the specific implementation of steps S21-S23, please refer to the aforementioned steps S11-S13, which will not be repeated here.

[0065] S24: Determine a third judgment index of the cluster point cloud, where the third judgment index includes a linear correlation index and / or a spatial autocorrelation index.

[0066] In one embodiment, in order to further improve the accuracy of noise point determination, a third judgment index can be further introduced. The third judgment index can describe the correlation between each spatial point in the cluster point cloud. For example, the third judgment index includes a linear correlation index and a spatial autocorrelation index, which can analyze the correlation between each spatial point contained in the cluster target from the cluster target level. The spatial autocorrelation index can describe the correlation between each spatial point in a group of cluster point clouds and other nearby spatial points. Low spatial autocorrelation means that the group of spatial points is scattered, and high spatial autocorrelation means that the group of spatial points is clustered. If the group of cluster point clouds is generated by laser irradiation of actual objects, then the spatial autocorrelation of the group of point clouds is relatively large, because common road obstacles such as cars, non-motor vehicles, pedestrians, etc. usually have large and continuous reflective surfaces. If a group of cluster point clouds is generated by rain and snow noise points, the spatial autocorrelation is relatively small.

[0067] Among them, the linear correlation index of the clustered point cloud can be determined in the following way: the coordinates of each spatial point in the clustered point cloud are fitted using the least squares method to obtain the fitting straight line of the clustered point cloud; the mean of the distance between each spatial point in the clustered point cloud and the fitting straight line is calculated to obtain the linear correlation index.

[0068] Specifically, for each spatial point, the first-dimensional coordinate x of each spatial point is used i and the second dimension coordinate is y i , the slope k and intercept b of the fitting line are obtained by the least squares method.

[0069]

[0070] Among them, M represents the number of spatial points belonging to the same cluster point cloud, x i is the first dimension coordinate of the i-th spatial point, y i is the second-dimensional coordinate of the i-th spatial point, and linearity is the linear correlation index.

[0071] The spatial autocorrelation index of the clustered point cloud can be determined in the following way: obtain the third distance between each spatial point in the clustered point cloud and the preset origin, and obtain the mean of the third distance corresponding to each spatial point in the clustered point cloud as the distance mean of the clustered point cloud; and obtain the difference between the third distance corresponding to each spatial point in the clustered point cloud and the distance mean; the calculation of the third distance is the same as the calculation method of the first distance mentioned above, which will not be repeated here. Take each spatial point in the clustered point cloud as a candidate point, take the product of the difference corresponding to the candidate point and the difference corresponding to any other point and divide it by the distance between the candidate point and the other point to obtain the first parameter of the candidate point; take the square of the difference corresponding to the candidate point as the second parameter of the candidate point; determine the linear correlation index of the clustered point cloud based on the first parameter of each candidate point, the second parameter, the number of candidate points and the autocorrelation weight; wherein the autocorrelation weight is related to the distance between the candidate point and the other point.

[0072] In a specific implementation, the spatial autocorrelation index calculation process is shown in the following formula.

[0073]

[0074] Among them, autocorrelation is the spatial autocorrelation index, M is the number of spatial points included in the cluster point cloud, that is, the number of candidate points, and w ij is the weight parameter, W is the autocorrelation weight, r i is the third distance from the i-th spatial point to the preset origin, r j is the third distance from the jth spatial point to the preset origin, r is the mean of the third distances from the M spatial points to the preset origin, dis i,j is the distance from the i-th spatial point to the j-th spatial point.

[0075] It can be understood that there is no particular order in which steps S24 and S25 are executed. They can be executed simultaneously, or one can be executed first and then the other.

[0076] S25: For each cluster point cloud, determine whether the cluster point cloud is a noise point by combining the first judgment index, the second judgment index, and the third judgment index of the cluster point cloud.

[0077] In one embodiment, the first judgment index, the second judgment index, and the third judgment index may be weighted and summed to obtain a total judgment index; based on the total judgment index, it is determined whether the clustered point cloud is a noise point. Step S25 is similar to step S14 and will not be described in detail here.

[0078] This application is mainly used to determine whether clustered point clouds belong to rain and snow noise points. In other scenarios, it can also be used to determine whether clustered point clouds belong to other types of noise points. For example, it can also determine whether clustered point clouds belong to obstacles. Comprehensive index evaluation of the target point cloud from two perspectives and multi-perspective processing combine the advantages of the two perspectives to better avoid misjudgment caused by occlusion. By making full use of the characteristics of three-dimensional point clouds, it can not only distinguish outlier noise points, but also identify noise points generated by large water vapor groups. The solution of this application is simple, and only a simple algorithm is used to determine the noise points, which can meet the real-time requirements of vehicle-mounted radars.

[0079] See also Figure 3 , Figure 3 It is a schematic diagram of the framework structure of an embodiment of an electronic device provided in the present application.

[0080] The electronic device 30 includes a memory 31 and a processor 32 coupled to each other, the memory 31 stores program instructions, and the processor 32 is used to execute the program instructions to implement the steps in any of the above method embodiments. Specifically, the electronic device 30 may include but is not limited to: a desktop computer, a laptop computer, a server, a mobile phone, a tablet computer, etc., which are not limited here.

[0081] Specifically, the processor 32 is used to control itself and the memory 31 to implement the steps in any of the above method embodiments. The processor 32 can also be called a CPU (Central Processing Unit). The processor 32 may be an integrated circuit chip with signal processing capabilities. The processor 32 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field-programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. In addition, the processor 32 can be implemented by an integrated circuit chip.

[0082] See also Figure 4 , Figure 4 It is a schematic diagram of the framework structure of an implementation scheme of a computer-readable storage medium provided in this application.

[0083] The computer-readable storage medium 40 stores program instructions 41 , and when the program instructions 41 are executed by a processor, they are used to implement the steps in any of the above method embodiments.

[0084] The computer-readable storage medium 40 may specifically be a medium that can store computer programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or may be a server that stores the computer program. The server may send the stored computer program to other devices for execution, or may also execute the stored computer program itself.

[0085] If the technical solution of this application involves personal information, the product using the technical solution of this application has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using the technical solution of this application has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, clear and prominent signs are set to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that he or she agrees to the collection of his or her personal information; or on the device that processes personal information, the personal information processing rules are notified by obvious signs / information, and the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

[0086] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A noise point determination method, characterized in that: include: Perform first-view projection and second-view projection on the target point cloud respectively, and obtain a first projection image and a second projection image correspondingly; Clustering the target point cloud to obtain at least one clustered point cloud; Based on the first projection image, determining a first judgment index for each of the clustered point clouds; and based on the second projection image, determining a second judgment index for each of the clustered point clouds; At least the first judgment index and the second judgment index of each cluster point cloud are combined to determine whether each cluster point cloud is a noise point.

2. The method according to claim 1, characterized in that: The first projection image includes first projection points of at least some of the spatial points in the target point cloud; The determining, based on the first projection image, a first judgment index of each of the cluster point clouds comprises: Performing rasterization processing on the first projection image to obtain a first raster image; Acquire initial grid indicators of each grid in the first grid image; wherein the initial grid indicators include the number of first projection points in the grid, and at least one of an intensity mean and a distance mean of each first projection point in the grid, and the distance mean is the mean of each first distance between the space point corresponding to each first projection point and a preset origin; Using each grid in the first grid image as a first target grid and using the initial grid index of the first target grid as a first judgment index of a spatial point corresponding to each first projection point in the first target grid; or, Based on the initial grid index of the first target grid and the initial grid index of the neighborhood grid of the first target grid, a final grid index of the first target grid is determined; and the final grid index of the first target grid is used as a first judgment index of the spatial points corresponding to each of the first projection points in the first target grid.

3. The method according to claim 2, characterized in that The final grid index includes target intensity and spatial dispersion, and the spatial dispersion includes at least one of spatial distance uniformity, spatial quantity uniformity, and density; The determining the final grid index of the first target grid based on the initial grid index of the first target grid and the initial grid index of the neighborhood grid of the first target grid includes: Averaging the intensity means included in the initial grid index of the first target grid and the initial grid index of the neighborhood grid to obtain the target intensity; Calculating the variance of each of the distance means included in the initial grid index of the first target grid and the initial grid index of the neighborhood grid to obtain the spatial distance uniformity; Calculating the variance of the number of the first projection points in the first target grid and the neighborhood grid to obtain the spatial quantity uniformity; The density is obtained by dividing the total number of the first projection points in the first target grid and the neighborhood grid by the number of the target grid and the neighborhood grid.

4. The method according to claim 1, characterized in that The second judgment index includes an average penetration rate, and the second projection image includes second projection points of at least some of the spatial points in the target point cloud; The determining, based on the second projection image, a second judgment index of each cluster point cloud comprises: performing rasterization processing on the second projection image to obtain a second raster image; Taking each grid covered by the clustered point cloud as a second target grid, obtaining the distance difference between the maximum distance and the minimum distance corresponding to each second target grid; wherein the maximum distance and the minimum distance are respectively the maximum value and the minimum value of the second distance between the spatial point corresponding to each second projection point in the second target grid and the preset origin; The average of the distance differences is calculated to obtain the average penetration rate.

5. The method according to claim 2 or 4, characterized in that: The target grid is a grid having projection points in the grid image.

6. The method according to claim 1, characterized in that After clustering the target point cloud to obtain at least one clustered point cloud, the method further includes: Determining a third judgment index of the cluster point cloud, wherein the third judgment index includes a linear correlation index and / or a spatial autocorrelation index; The determining whether each of the clustered point clouds is a noise point by at least combining the first judgment index and the second judgment index of each of the clustered point clouds comprises: For each of the clustered point clouds, the first judgment index, the second judgment index, and the third judgment index of the clustered point cloud are combined to determine whether the clustered point cloud is a noise point.

7. The method according to claim 6, characterized in that Determining a linear correlation index of the clustered point cloud includes: Fitting the coordinates of each spatial point in the cluster point cloud using the least squares method to obtain a fitting straight line of the cluster point cloud; Calculating the mean of the distances between each spatial point in the cluster point cloud and the fitting straight line to obtain the linear correlation index; And / or, determining the spatial autocorrelation index of the clustered point cloud, comprising: Obtaining a third distance between each spatial point in the clustered point cloud and a preset origin, and obtaining a mean of the third distances corresponding to each spatial point in the clustered point cloud as a distance mean of the clustered point cloud; and obtaining a difference between the third distance corresponding to each spatial point in the clustered point cloud and the distance mean; Taking each spatial point in the cluster point cloud as a candidate point, respectively, calculating the product of the difference value corresponding to the candidate point and the difference value corresponding to any other point and dividing the product by the distance between the candidate point and the other point, to obtain a first parameter of the candidate point; taking the square of the difference value corresponding to the candidate point as a second parameter of the candidate point; Based on the first parameter, the second parameter, the number of candidate points and the autocorrelation weight of each candidate point, a linear correlation index of the clustered point cloud is determined; wherein the autocorrelation weight is related to the distance between the candidate point and the other points.

8. The method according to claim 6, characterized in that The determining whether the clustered point cloud is a noise point by combining the first judgment index, the second judgment index, and the third judgment index of the clustered point cloud includes: Performing weighted summation on the first judgment index, the second judgment index, and the third judgment index to obtain a total judgment index; Based on the overall judgment index, it is determined whether the clustered point cloud belongs to a noise point.

9. The method according to claim 1, characterized in that: The step of clustering the target point cloud to obtain at least one clustered point cloud comprises: Performing a third viewing angle projection on the target point cloud to obtain a third projection image, wherein the third projection image includes third projection points of at least some of the spatial points in the target point cloud; Acquire a fourth distance between a spatial point corresponding to each third projection point in the third projection image and a preset origin; Based on the difference of the fourth distances corresponding to the third projection points, it is determined whether the third projection points belong to the same cluster point cloud.

10. The method according to claim 1, characterized in that The first perspective is a top-down perspective, and the second perspective is a front-view perspective; And / or, the method further comprises: Get the initial point cloud; The initial point cloud is preprocessed to obtain the target point cloud; wherein the preprocessing includes downsampling and / or filtering out point clouds in the initial point cloud that belong to a preset area, and the preset area includes the ground.

11. An electronic device, characterized in that: include: A memory and a processor, wherein the memory stores program instructions; The processor is used to execute the program instructions stored in the memory to implement the method according to any one of claims 1 to 10.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program instructions, and the program instructions can be executed by a processor to implement the method according to any one of claims 1 to 10.

Images