A 3D point cloud filtering method and system based on depth information

Abstract

The present application relates to a 3D point cloud filtering method and system based on depth information, and relates to the field of image processing technology. The method includes acquiring image information; generating a depth map and generating three-dimensional coordinate information; generating clustering information; calculating cluster center information; calculating point cloud radius information; calculating depth average information; calculating point cloud radius threshold information; when the point cloud radius information is less than the point cloud radius threshold information, the current clustering information is regarded as noise information and removed from the image information, thereby improving the problem that the data difference is too large during the filtering process, which causes interference to the results. The present application has the effect of filtering out data points formed by non-obstacles, avoiding misjudgment or missed judgment, and improving the accuracy of noise removal.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a 3D point cloud filtering method and system based on depth information. Background Art

[0002] With the advancement of modern technology, emerging applications based on 3D vision are booming, and 3D point clouds are attracting increasing attention. Point clouds have a wide range of applications, including robotics, 3D graphics, autonomous driving, and virtual reality (AR / VR / MR).

[0003] Regarding the aforementioned technologies, the inventors believe that researching and developing algorithms for efficient storage and processing of point clouds is becoming increasingly important to keep pace with growing application needs. Traditional point cloud analysis algorithms primarily focus on encoding the local geometric features of points. This can lead to significant data discrepancies during the filtering process, which can interfere with the results, leaving room for improvement. Summary of the Invention

[0004] In order to improve the problem of excessive data differences that may interfere with the results during the filtering process, the present application provides a 3D point cloud filtering method, system, storage medium, and intelligent terminal based on depth information.

[0005] In a first aspect, the present application provides a 3D point cloud filtering method based on depth information, which adopts the following technical solutions:

[0006] A 3D point cloud filtering method based on depth information, comprising:

[0007] Obtain image information;

[0008] Generate a depth map based on the image information and generate three-dimensional coordinate information based on the coordinates of the depth map combined with the depth value, wherein the three-dimensional coordinate information includes horizontal coordinate information, vertical coordinate information and depth information;

[0009] Generate clustering information based on three-dimensional coordinate information;

[0010] Arbitrarily select cluster information and calculate cluster center information, and define the selected cluster as the current cluster information;

[0011] Calculate the point cloud radius information based on the cluster center information;

[0012] Calculate the average depth information based on the depth information in the three-dimensional coordinate information corresponding to the current cluster information;

[0013] Calculate the point cloud radius threshold information based on the depth average information;

[0014] When the point cloud radius information is less than the point cloud radius threshold information, the current cluster information is regarded as noise information and removed from the image information.

[0015] By adopting the above technical solution, under normal circumstances, real obstacles will appear as a series of relatively coherent data points in the depth map, while noise points in the depth map are usually manifested as discrete jump points. Therefore, clustering can be used to detect whether they are obstacles or noise. The maximum distribution range of the points in the cluster is calculated and converted into a point cloud radius threshold. Then, based on whether the average depth value is less than the point cloud radius threshold, it is determined whether the point cloud corresponds to a real obstacle or noise in the scene, thereby filtering out data points formed by non-obstacles. Regardless of whether the noise points are far away from the obstacles or at a long distance, this method can accurately process them through adaptive adjustment, avoiding misjudgment or missed judgment, and improving the accuracy of noise removal.

[0016] Optionally, a method for calculating point cloud radius information based on cluster center information includes:

[0017] Calculate the three-dimensional distance values of all data points and cluster center information in the current cluster information and filter out the value with the largest three-dimensional distance value as the point cloud radius information.

[0018] By adopting the above technical solution, the point cloud obtained by the clustering algorithm has relatively coherent data points and no isolated jump points. Therefore, taking the maximum distance from the center as the point cloud radius can reflect the maximum distribution range of the point cloud. Based on this, it is judged whether the point cloud corresponds to a real obstacle or noise in the scene, thereby improving the accuracy of eliminating noise points.

[0019] Optionally, another method for calculating point cloud radius information based on cluster center information includes:

[0020] Calculate the one-dimensional distance value between all data points and cluster center information in the current cluster information, and define the one-dimensional distance value as the one-dimensional distance information;

[0021] Filter the maximum and minimum values in any single-dimensional distance information respectively and calculate the difference between the two, and define the difference as the single-dimensional span information;

[0022] The maximum value in the single-dimensional span information is selected and half of the maximum value is calculated to be defined as the point cloud radius information.

[0023] By adopting the above technical solution, for different objects, their spans in the three directions of X, Y, and Z axes may vary greatly, and the distribution density of point clouds in different directions may not be consistent. By calculating half of the span as the point cloud radius, the obtained point cloud radius can accurately describe the distribution of the point cloud, thereby improving the accuracy of point cloud distribution determination.

[0024] Optionally, the method of generating clustering information based on three-dimensional coordinate information includes:

[0025] Selecting one of the three-dimensional coordinate information to form initial cluster information including the three-dimensional coordinate information, defining the three-dimensional coordinate information as selected three-dimensional coordinate information, and defining the three-dimensional coordinate information that has not been selected as unselected three-dimensional coordinate information;

[0026] Calculate an average value based on the depth information of the data points in the initial clustering information, and define the average value as dynamic depth average information;

[0027] Calculate dynamic clustering threshold information based on dynamic depth average information;

[0028] arbitrarily selecting unselected three-dimensional coordinate information and calculating a distance value between the unselected three-dimensional coordinate information and the selected three-dimensional coordinate information, and defining the distance value as distance information;

[0029] Filter out the smallest distance value in the distance information and define it as the minimum distance information;

[0030] When the minimum distance information is less than the dynamic clustering threshold information, the unselected three-dimensional coordinate information is updated to the corresponding selected three-dimensional coordinate information in the initial clustering information;

[0031] When the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the dynamic clustering threshold information, the dynamic clustering threshold information is updated and it is re-determined whether the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the updated dynamic clustering threshold information;

[0032] When the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the updated dynamic clustering threshold information, the initial clustering information is updated to the clustering information;

[0033] The unselected three-dimensional coordinate information is reselected and the initial clustering information is re-formed until the unselected three-dimensional coordinate information no longer exists.

[0034] By adopting the above technical solution, the threshold used in clustering data is selected based on the average depth of the pixels currently being processed. After processing each cluster, a new threshold needs to be calculated to process the remaining pixels. The threshold used each time will adaptively change with the remaining data points. By utilizing the distribution characteristics of obstacles on the depth map, the densely distributed data points formed by obstacles on the depth map are formed into a cluster, and the data points formed by non-obstacles are filtered out. This can effectively avoid the interference of excessive data differences on the results, more accurately process all pixels in the image, and improve the accuracy of the clustering results.

[0035] Optionally, a method for selecting one of the three-dimensional coordinate information to form initial clustering information including the three-dimensional coordinate information includes:

[0036] According to the depth information, the three-dimensional coordinate information corresponding to the maximum value is used as the first three-dimensional coordinate information selected, the three-dimensional coordinate information is defined as the initial three-dimensional coordinate information, and the initial clustering information is defined as the first initial clustering information;

[0037] Calculating initial threshold information based on depth information of initial three-dimensional coordinate information;

[0038] selecting the three-dimensional coordinate information having the largest depth information other than the initial three-dimensional coordinate information as the second three-dimensional coordinate information, and defining the three-dimensional coordinate information as the secondary selected three-dimensional coordinate information;

[0039] Calculating the difference between the secondary three-dimensional coordinate information and the initial three-dimensional coordinate information, and defining the difference as distance difference information;

[0040] Determine whether the distance difference information is greater than the initial threshold information;

[0041] If it is greater than, the second selected three-dimensional coordinate information is used as the initial three-dimensional coordinate information of the next cluster, and the next cluster is defined as the second initial cluster information;

[0042] If it is less than, the second selected three-dimensional coordinate information is added to the first initial clustering information as the selected three-dimensional coordinate information;

[0043] selecting the three-dimensional coordinate information having the largest depth information other than the initial three-dimensional coordinate information and the second-selected three-dimensional coordinate information as the third three-dimensional coordinate information, and defining the three-dimensional coordinate information as the second-selected three-dimensional coordinate information;

[0044] determining in sequence whether the reselected three-dimensional coordinate information is to be added to the first initial clustering information or the second initial clustering information;

[0045] When the reselected three-dimensional coordinate information is not added to the first initial clustering information and the second initial clustering information, the third initial clustering information is established until the three-dimensional coordinate information has become the selected three-dimensional coordinate information.

[0046] By adopting the above technical solution, by selecting the maximum depth as the range of the initial clustering, more data points can be included in the initial clustering, thereby quickly parsing images with large amounts of data and reducing the amount of repeated calculations in the algorithm, thereby improving the efficiency of the algorithm.

[0047] Optionally, after all three-dimensional coordinate information forms initial cluster information, the method for forming cluster information includes:

[0048] Updating the initial threshold information into dynamic clustering threshold information according to the three-dimensional coordinate information in the first initial clustering information;

[0049] respectively calculating the distance values between the three-dimensional coordinate information of other initial clustering information and the three-dimensional coordinate information in the first initial clustering information and selecting the minimum distance value, which is defined as the verification distance information;

[0050] Determine whether the verification distance information is less than the dynamic clustering threshold information;

[0051] If it is less than, delete the corresponding initial clustering information;

[0052] If it is greater than, the corresponding initial clustering information is maintained;

[0053] After the three-dimensional coordinate information in all other initial clustering information is verified, the first initial clustering information is updated and the other initial clustering information is updated.

[0054] By adopting the above technical solution, by re-judging whether other initial clustering information falls into the cluster from the largest to the smallest after the first initial clustering information is completed, the first initial clustering information continues to contain other data and is not prone to data omissions. The data that did not enter the first initial clustering information during the first clustering process is re-checked, thereby improving the accuracy of clustering.

[0055] Optionally, another method for selecting one of the three-dimensional coordinate information to form initial cluster information including the three-dimensional coordinate information includes:

[0056] According to the depth information, the three-dimensional coordinate information corresponding to the minimum value is selected as the first three-dimensional coordinate information, the three-dimensional coordinate information is defined as the reverse initial three-dimensional coordinate information, and the initial clustering information is defined as the fourth initial clustering information;

[0057] Calculating reverse initial threshold information based on depth information of reverse initial three-dimensional coordinate information;

[0058] Selecting the three-dimensional coordinate information with the smallest depth information except the reverse initial three-dimensional coordinate information as the second three-dimensional coordinate information, and defining the three-dimensional coordinate information as the reverse secondary three-dimensional coordinate information;

[0059] Calculating the difference between the reverse secondary 3D coordinate information and the reverse initial 3D coordinate information, and defining the difference as reverse distance difference information;

[0060] Determine whether the reverse distance difference information is greater than the reverse initial threshold information;

[0061] If it is greater than, the reverse second-selected three-dimensional coordinate information is used as the initial three-dimensional coordinate information of the next cluster, and the next cluster is defined as the fifth initial cluster information;

[0062] If it is less than, the reverse second-selected three-dimensional coordinate information is added as the selected three-dimensional coordinate information to the fourth initial clustering information;

[0063] selecting the three-dimensional coordinate information with the smallest depth information except the reverse initial three-dimensional coordinate information and the reverse second-selected three-dimensional coordinate information as the third three-dimensional coordinate information, and defining the three-dimensional coordinate information as the reverse second-selected three-dimensional coordinate information;

[0064] determining in sequence whether the reversely selected three-dimensional coordinate information is to be added to the fourth initial clustering information or the fifth initial clustering information;

[0065] When the reselected three-dimensional coordinate information is not added to the fourth initial clustering information and the fifth initial clustering information, the sixth initial clustering information is established until the three-dimensional coordinate information has become the selected three-dimensional coordinate information.

[0066] By adopting the above technical solution, clustering is performed starting from the smallest depth, the number of clusters is smaller, and when noise points are determined, fewer data points in the cluster are covered, the noise point analysis is more accurate, and the accuracy of filtering out noise points is improved.

[0067] Optionally, when the number of three-dimensional coordinate information corresponding to the depth information with the largest value is two or more, the method for selecting the initial three-dimensional coordinate information includes:

[0068] The three-dimensional coordinate information corresponding to the depth information with the largest value is defined as the deepest three-dimensional coordinate information;

[0069] Calculate the median value of the horizontal and vertical coordinates with the largest values, and define the straight line corresponding to the horizontal and vertical coordinates corresponding to the average value as the central axis information;

[0070] Calculate the distance between the deepest three-dimensional coordinate information and the central axis respectively, and define the distance value as the horizontal distance information;

[0071] After sorting the deepest three-dimensional coordinate information in ascending order according to the value of the horizontal distance to the middle distance information, the deepest three-dimensional coordinate information with the smallest value of the horizontal distance to the middle distance information is defined as the initial three-dimensional coordinate information, and the next deepest three-dimensional coordinate information is defined as the secondary three-dimensional coordinate information.

[0072] By adopting the above technical solution, by selecting the point closest to the middle point as the initial three-dimensional coordinate information, although the depth is the same, the coverage area closest to the middle point is the largest, which can reduce the calculation after the cluster is formed exponentially, greatly reducing the calculation workload and improving the filtering efficiency.

[0073] Optionally, a method for calculating the point cloud radius threshold information and the dynamic clustering threshold information includes:

[0074] Obtain target information;

[0075] Calculate depth parameter information and radius parameter information according to the required target information;

[0076] Calculate the product between the depth average information and the radius parameter information to obtain the point cloud radius threshold information;

[0077] The dynamic clustering threshold information is obtained by calculating the product between the dynamic depth average information and the depth parameter information.

[0078] By adopting the above technical solution, the coefficients of the two thresholds are established according to the demand goals, which are essentially determined according to the design goals of the product. Therefore, the parameters are more in line with actual needs, the judgment process is more accurate, and the accuracy of filtering and clustering is improved.

[0079] In a second aspect, the present application provides a 3D point cloud filtering system based on depth information, which adopts the following technical solutions:

[0080] A 3D point cloud filtering system based on depth information, comprising:

[0081] An image acquisition module, used for acquiring image information;

[0082] A processing module, connected to the image acquisition and judgment module, is used for storing and processing information;

[0083] A generation module, connected to the processing module, is used to generate a depth map according to the image information and generate three-dimensional coordinate information according to the coordinates of the depth map combined with the depth value, wherein the three-dimensional coordinate information includes horizontal coordinate information, vertical coordinate information and depth information;

[0084] The generation module generates clustering information based on the three-dimensional coordinate information;

[0085] A calculation module, connected to the processing module, is used to arbitrarily select cluster information and calculate cluster center information, and define the selected cluster as current cluster information;

[0086] The calculation module calculates the point cloud radius information based on the cluster center information;

[0087] The calculation module calculates the depth average information based on the depth information in the three-dimensional coordinate information corresponding to the current cluster information;

[0088] The calculation module calculates the point cloud radius threshold information based on the depth average information;

[0089] A judgment module is used to judge the size of the point cloud radius information and the point cloud radius threshold information;

[0090] When the judgment module determines that the point cloud radius information is less than the point cloud radius threshold information, the processing module regards the current clustering information as noise information and removes it from the image information.

[0091] By adopting the above technical solution, under normal circumstances, real obstacles will appear as a series of relatively coherent data points in the depth map, while noise points in the depth map are usually manifested as discrete jump points. Therefore, clustering can be used to detect whether they are obstacles or noise. The maximum distribution range of the points in the cluster is calculated and converted into a point cloud radius threshold. Then, based on whether the average depth value is less than the point cloud radius threshold, it is determined whether the point cloud corresponds to a real obstacle or noise in the scene, thereby filtering out data points formed by non-obstacles. Regardless of whether the noise points are far away from the obstacles or at a long distance, this method can accurately process them through adaptive adjustment, avoiding misjudgment or missed judgment, and improving the accuracy of noise removal.

[0092] In summary, this application includes at least one of the following beneficial technical effects:

[0093] 1. By calculating the maximum distribution range of points within the cluster, data points not formed by obstacles are filtered out to avoid misjudgment or missed judgment, thereby improving the accuracy of noise removal;

[0094] 2. Taking the maximum distance from the center as the point cloud radius can reflect the maximum distribution range of the point cloud and improve the accuracy of noise elimination;

[0095] 3. By selecting the maximum depth as the range of the starting clustering, images with large amounts of data can be quickly parsed, reducing the amount of repeated calculations in the algorithm and improving the efficiency of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 This is a flowchart of a 3D point cloud filtering method based on depth information in an embodiment of the present application.

[0097] Figure 2 This is a flowchart of a method for calculating point cloud radius information based on cluster center information in an embodiment of the present application.

[0098] Figure 3 This is a flowchart of another method for calculating point cloud radius information based on cluster center information in an embodiment of the present application.

[0099] Figure 4 This is a flowchart of a method for generating clustering information based on three-dimensional coordinate information in an embodiment of the present application.

[0100] Figure 5 This is a flowchart of a method for selecting initial clustering information containing one of the three-dimensional coordinate information in an embodiment of the present application.

[0101] Figure 6 It is a flowchart of a method for forming cluster information after all three-dimensional coordinate information forms initial cluster information in an embodiment of the present application.

[0102] Figure 7 This is a flowchart of another method for selecting initial clustering information containing one of the three-dimensional coordinate information in an embodiment of the present application.

[0103] Figure 8 It is a flow chart of the method for opening and closing each continuously arranged conveyor belt in the embodiment of the present application.

[0104] Figure 9 This is a flowchart of a method for calculating point cloud radius threshold information and dynamic clustering threshold information in an embodiment of the present application.

[0105] Figure 10 This is a module diagram of a 3D point cloud filtering method based on depth information in an embodiment of the present application. DETAILED DESCRIPTION

[0106] In order to make the purpose, technical solutions and advantages of this application more clear, the following Figure 1-10 It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0107] The embodiments of the present invention are described in further detail below with reference to the accompanying drawings.

[0108] See also Figure 1 The embodiment of the present invention provides a 3D point cloud filtering method based on depth information. The main process of the 3D point cloud filtering method based on depth information is described as follows:

[0109] Step 100: Acquire image information.

[0110] The image information is the information of the image to be analyzed, and can be obtained by any device capable of taking a picture, such as a camera.

[0111] Step 101: Generate a depth map according to image information and generate three-dimensional coordinate information according to the coordinates of the depth map combined with the depth value, wherein the three-dimensional coordinate information includes horizontal coordinate information, vertical coordinate information and depth information.

[0112] A depth map is a graph where each pixel in an image is represented by its depth value. The map consists of an area with I rows and J columns, each representing a pixel with a depth value of Z. The generation method is similar to 3D imaging and will not be described in detail here. The three-dimensional coordinate information is defined by defining the pixel's position in the image as its X and Y coordinates, with its depth value as the Z coordinate. The horizontal coordinate is the X coordinate, the vertical coordinate is the Y coordinate, and the depth information is the Z coordinate.

[0113] Step 102: Generate clustering information based on the three-dimensional coordinate information.

[0114] Clustering is the process of dividing a collection of physical or abstract objects into clusters of similar objects. Cluster information is one of these clusters. Three-dimensional coordinate information is classified according to a set of rules so that all data points are evenly assigned to a particular cluster.

[0115] Step 103: Randomly select cluster information and calculate cluster center information, and define the selected cluster as current cluster information.

[0116] The current cluster information is the information of the selected cluster. The cluster center information is the information of the center of all data points in the current cluster information. The calculation method is to add the X coordinates, Y coordinates, and Z coordinates of all data points separately to obtain the sum of X coordinates, Y coordinates, and Z coordinates, and then divide them by the number of data points to obtain the average X coordinates, Y coordinates, and Z coordinates, respectively, as the X coordinates, Y coordinates, and Z coordinates of the cluster center information.

[0117] Step 104: Calculate point cloud radius information based on cluster center information.

[0118] The point cloud radius information is the assumed radius of the point cloud distribution range. The calculation method is to classify according to a certain rule.

[0119] Step 105: Calculate average depth information based on the depth information in the three-dimensional coordinate information corresponding to the current clustering information.

[0120] The depth average value information is the average value of the depth information, which is calculated by adding up the Z coordinates of all the three-dimensional coordinate information in the current cluster information and dividing the sum by the number of data points.

[0121] Step 106: Calculate point cloud radius threshold information based on the depth average value information.

[0122] The point cloud radius threshold information is the maximum limit of the point cloud. The calculation method is to calculate the depth average information according to a certain rule, which is related to the depth average.

[0123] Step 107: When the point cloud radius information is less than the point cloud radius threshold information, the current cluster information is regarded as noise information and removed from the image information.

[0124] Clusters with a point cloud radius r less than the point cloud radius threshold are considered noise points. This is because, in general, real obstacles appear as a series of relatively coherent data points in the depth map, while noise points in the depth map typically appear as discrete transition points. Therefore, clustering can be used to detect obstacles and noise. Compared to traditional algorithms, this noise removal scheme has the advantage of accurately processing noise points far from obstacles or at a greater distance through adaptive adjustments, avoiding misjudgments or missed detections.

[0125] Reference Figure 2 ,The method of calculating the point cloud radius information based on the cluster center information includes:

[0126] Step 200: Calculate the three-dimensional distance values between all data points in the current cluster information and the cluster center information and select the value with the largest three-dimensional distance value as the point cloud radius information.

[0127] The point cloud radius is determined by the maximum distance from the center of the cluster. This is because the point cloud obtained by the clustering algorithm is relatively coherent, without isolated jump points. Therefore, the maximum distance from the center is used as the point cloud radius to reflect the maximum distribution range of the point cloud. This is used as a basis for determining whether the point cloud corresponds to a real obstacle or noise in the scene. Compared to traditional algorithms, this clustering method can utilize the distribution characteristics of obstacles in the depth map, forming a cluster of densely distributed data points formed by obstacles in the depth map, and filtering out data points formed by non-obstacles. This clustering method can more accurately eliminate noise.

[0128] Reference Figure 3 Another method for calculating the point cloud radius information based on the cluster center information includes:

[0129] Step 300: Calculate the one-dimensional distance value between all data points in the current cluster information and the cluster center information, and define the one-dimensional distance value as the one-dimensional distance information.

[0130] The single-dimensional distance information is the difference between any coordinate of the data point and the value of the same coordinate of the cluster center information. The calculation method is to subtract the two.

[0131] Step 301: Filter the maximum value and the minimum value in any one-dimensional distance information respectively and calculate the difference between the two, and define the difference as the one-dimensional span information.

[0132] The single-dimensional span information is the difference between the maximum and minimum values in the unit distance information, calculated by subtracting the two. It should be noted that the values subtracted here are at the same coordinate.

[0133] Step 302: Select the maximum value in the single-dimensional span information and calculate half of the maximum value to be defined as the point cloud radius information.

[0134] The method chosen here is that different objects may have large differences in their spans in the X, Y, and Z axes, and the distribution density of the point cloud in different directions may not be consistent. If the traditional algorithm is used, it may not be possible to accurately describe the distribution of the point cloud. By selecting half of the maximum span as the point cloud radius, the interference of this situation can be avoided.

[0135] Reference Figure 4 The method for generating clustering information based on three-dimensional coordinate information includes:

[0136] Step 400: Select one of the three-dimensional coordinate information to form initial cluster information including the three-dimensional coordinate information, define the three-dimensional coordinate information as selected three-dimensional coordinate information, and define the three-dimensional coordinate information that has not been selected as unselected three-dimensional coordinate information.

[0137] Initial cluster information refers to the cluster information formed based on the first selected 3D coordinate information when clustering is first formed. This cluster information contains the first selected 3D coordinate information and may also contain other 3D coordinate information. Selected 3D coordinate information refers to the 3D coordinate information that has already been included in the cluster information. Unselected 3D coordinate information refers to the 3D coordinate information that has not yet been included in any cluster information.

[0138] Step 401: Calculate an average value based on the depth information of the data points in the initial clustering information, and define the average value as dynamic depth average value information.

[0139] The dynamic depth average value information is the average value of the depth information of the data points in the initial cluster information, and is calculated by dividing the sum of the depth information of the data points in the initial cluster information by the number of data points in the initial cluster information.

[0140] Step 402: Calculate dynamic clustering threshold information based on the dynamic depth average value information.

[0141] The dynamic clustering threshold is a critical value calculated based on the dynamic depth average. This value is calculated by multiplying the dynamic depth average by a reasonable parameter. This is intended to make the threshold more accurate and better distinguish between clusters.

[0142] Step 403: arbitrarily select unselected three-dimensional coordinate information and calculate the distance value between the unselected three-dimensional coordinate information and the selected three-dimensional coordinate information, and define the distance value as distance information.

[0143] The distance information is the distance between the unselected 3D coordinate information and the selected 3D coordinate information. The distance value is a 3D distance value. The purpose of the calculation is to determine whether it can be classified into a cluster.

[0144] Step 404: Filter out the minimum distance value in the distance information and define it as the minimum distance information.

[0145] The minimum distance information is the smallest distance value in the distance information. The screening method is numerical comparison.

[0146] Step 405: When the minimum distance information is less than the dynamic clustering threshold information, the unselected three-dimensional coordinate information is updated to the corresponding selected three-dimensional coordinate information in the initial clustering information.

[0147] When the minimum distance is less than the dynamic clustering threshold, the selected unselected 3D coordinates are within the threshold range, indicating that the unselected 3D coordinates meet the clustering rules and can be included in the cluster. The update method is to update the unselected 3D coordinates to selected 3D coordinates and attach the label of the initial cluster information.

[0148] Step 406: When the minimum distance information corresponding to the unselected 3D coordinate information is greater than the dynamic clustering threshold information, the dynamic clustering threshold information is updated and it is re-determined whether the minimum distance information corresponding to the unselected 3D coordinate information is greater than the updated dynamic clustering threshold information.

[0149] When the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the dynamic clustering threshold information, it means that the data points corresponding to the unselected three-dimensional coordinate information cannot enter the initial clustering information under the previous screening conditions. The dynamic clustering threshold information in the initial clustering information at this time is updated to produce adaptive changes, thereby effectively avoiding interference with the results caused by excessive data differences.

[0150] Step 407: When the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the updated dynamic clustering threshold information, the initial clustering information is updated to the clustering information.

[0151] If the initial clustering information cannot be entered even after the dynamic clustering threshold information and the minimum distance information are updated, it means that the clustering is completed at this time, and the initial clustering information is updated to the clustering information, indicating that the clustering is completed at this time.

[0152] Step 408: Reselect unselected three-dimensional coordinate information and re-form the initial clustering information until no unselected three-dimensional coordinate information exists.

[0153] After all the data are processed in the manner of steps 400-408, all the data points can be clustered to obtain clustering information.

[0154] Reference Figure 5 , a method for selecting initial clustering information containing one of the three-dimensional coordinate information includes:

[0155] Step 500: The three-dimensional coordinate information corresponding to the maximum value is selected as the first three-dimensional coordinate information according to the depth information, and the three-dimensional coordinate information is defined as initial three-dimensional coordinate information and the initial clustering information is defined as first initial clustering information.

[0156] The initial three-dimensional coordinate information is the first three-dimensional coordinate information selected to form the first cluster. The selection method is to compare the depth information of all three-dimensional coordinate information and select the three-dimensional coordinate information corresponding to the largest depth information. The first initial clustering information is the information of the cluster containing the initial three-dimensional coordinate information. At this time, since there is only one data point in the cluster, the dynamic clustering threshold information is formed according to the depth of the data point. At this time, the larger the depth value, the larger the dynamic clustering threshold information, and the larger the range, the wider the spatial range that can be included. Therefore, more data points can be included in the initial clustering, thereby quickly parsing images with large amounts of data and reducing the amount of repeated calculations of the algorithm, thereby improving the efficiency of the algorithm.

[0157] Step 501: Calculate initial threshold information based on depth information of initial three-dimensional coordinate information.

[0158] The initial threshold information is an initial value of the dynamic clustering threshold information obtained by calculating according to a rule based on the depth information of the initial three-dimensional coordinate information.

[0159] Step 502: Select the three-dimensional coordinate information with the maximum depth information except the initial three-dimensional coordinate information as the second three-dimensional coordinate information, and define the three-dimensional coordinate information as the secondary three-dimensional coordinate information.

[0160] The second selected three-dimensional coordinate information is the largest three-dimensional coordinate information except the initial three-dimensional coordinate information. The selection method is to delete the initial three-dimensional coordinate information from the depth map and then compare all depth values.

[0161] Step 503: Calculate the difference between the secondary three-dimensional coordinate information and the initial three-dimensional coordinate information, and define the difference as distance difference information.

[0162] The distance difference information is the difference between the secondary 3D coordinate information and the initial 3D coordinate information. Since the first initial cluster information only includes the initial 3D coordinate information at this time, it is only necessary to calculate the difference between the two to determine whether it falls into the first initial cluster information.

[0163] Step 504: Determine whether the distance difference information is greater than the initial threshold information.

[0164] Step 5041: If it is greater than, the second selected three-dimensional coordinate information is used as the initial three-dimensional coordinate information of the next cluster, and the next cluster is defined as the second initial cluster information.

[0165] The second initial clustering information is the initial clustering information containing the secondary three-dimensional coordinate information. If it is greater than, it means that it does not fall within the first initial clustering information, so it can be used as the initial three-dimensional coordinate information of the next cluster to form the second initial clustering information. It should be noted that the cluster formed here is only for reference. When the distance value between it and the data point in the updated first initial clustering information is less than the threshold, the second initial clustering information is deleted. However, when the secondary three-dimensional coordinate information forming the second initial clustering information is still within the second initial clustering information, it is not deleted.

[0166] Step 5042: If it is less than, the second-selected three-dimensional coordinate information is added to the first initial clustering information as the selected three-dimensional coordinate information.

[0167] If it is less than, it means that it meets the requirements and can be added to the first initial clustering information, and it can be updated to the selected three-dimensional coordinate information to reduce the number of repeated selections.

[0168] Step 505: Select the three-dimensional coordinate information with the maximum depth information except the initial three-dimensional coordinate information and the second-selected three-dimensional coordinate information as the third three-dimensional coordinate information, and define the three-dimensional coordinate information as the second-selected three-dimensional coordinate information.

[0169] The reselected three-dimensional coordinate information is the maximum three-dimensional coordinate information of the depth information except the initial three-dimensional coordinate information and the second-selected three-dimensional coordinate information.

[0170] Step 506: determining in sequence whether the reselected three-dimensional coordinate information is to be added to the first initial clustering information or the second initial clustering information.

[0171] The purpose of the judgment here is the same as that of step 504.

[0172] Step 507: When the reselected three-dimensional coordinate information is not added to the first initial clustering information and the second initial clustering information, the third initial clustering information is established until the three-dimensional coordinate information has become the selected three-dimensional coordinate information.

[0173] The purpose here is the same as step 5041, and will not be described in detail here. The screening performed here is ultimately a first pass screening of all three-dimensional coordinate information.

[0174] Reference Figure 6 , after all three-dimensional coordinate information forms initial cluster information, the cluster information forming method includes:

[0175] Step 600: updating the initial threshold information to dynamic clustering threshold information according to the three-dimensional coordinate information in the first initial clustering information.

[0176] Here is the three-dimensional coordinate information in the first initial clustering information formed after a round of screening. The purpose of updating is to check whether the data points in other clustering information are consistent with the data points in the updated first initial clustering information.

[0177] Step 601: Calculate the distance values between the three-dimensional coordinate information of other initial clustering information and the three-dimensional coordinate information in the first initial clustering information respectively, and select the minimum distance value, which is defined as the verification distance information.

[0178] The verification distance information is the minimum distance between the 3D coordinate information of the other initial clustering information and the 3D coordinate information in the first initial clustering information. It is important to note that the initial clustering information other than the first initial clustering information is only a relatively fragile classification information. The data within it is not the point that is ultimately classified within the cluster information and can be filtered out of the cluster and classified into other clusters at any time.

[0179] Step 602: Determine whether the verification distance information is less than the dynamic clustering threshold information.

[0180] The purpose of the judgment is to determine whether it belongs to the first initial clustering information.

[0181] Step 6021: If it is less than, delete the corresponding initial clustering information.

[0182] When a certain three-dimensional coordinate information was originally in the Nth initial clustering information, but when it is reclassified it is found to belong to the first initial clustering information, it will be classified into the first initial clustering information, and if the three-dimensional coordinate information is the initial point for establishing the Nth initial clustering information, the Nth initial clustering information will be disbanded, and the remaining data points in the Nth initial clustering information will be re-clustered according to steps 500-507.

[0183] Step 6022: If it is greater than, maintain the corresponding initial clustering information.

[0184] If it is greater than, the initial clustering information will continue to remain unchanged.

[0185] Step 603: After the three-dimensional coordinate information in all other initial clustering information is verified, the first initial clustering information is updated and the other initial clustering information is updated.

[0186] Reference Figure 7 Another method for selecting one of the three-dimensional coordinate information to form initial cluster information including the three-dimensional coordinate information includes:

[0187] Step 700: Based on the depth information, the three-dimensional coordinate information corresponding to the minimum value is selected as the first three-dimensional coordinate information, the three-dimensional coordinate information is defined as the reverse initial three-dimensional coordinate information, and the initial clustering information is defined as the fourth initial clustering information.

[0188] The reverse initial three-dimensional coordinate information is the three-dimensional coordinate information of the first three-dimensional coordinate information selected to form the first cluster. The selection method is to compare the depth information of all three-dimensional coordinate information and select the three-dimensional coordinate information corresponding to the smallest depth information. The fourth initial clustering information is the information of the cluster containing the reverse initial three-dimensional coordinate information. At this time, since there is only one data point in the cluster, the dynamic clustering threshold information is formed according to the depth of the data point. At this time, the larger the depth value, the smaller the dynamic clustering threshold information, and the smaller the range, the smaller the spatial range that can be included, and the accuracy of the clustering is relatively more accurate. The determination of noise points is relatively accurate and it is not easy to determine nearby areas that are not noise points as noise points because of the presence of noise points in them and the large coverage area, thereby improving the accuracy of filtering out noise points.

[0189] Step 701: Calculate reverse initial threshold information based on depth information of reverse initial three-dimensional coordinate information.

[0190] Step 702: Select the three-dimensional coordinate information with the smallest depth information except the reverse initial three-dimensional coordinate information as the second three-dimensional coordinate information, and define the three-dimensional coordinate information as the reverse secondary three-dimensional coordinate information.

[0191] The reverse secondary 3D coordinate information is the smallest 3D coordinate information except the reverse initial 3D coordinate information, and is obtained by deleting the reverse initial 3D coordinate information from the depth map and then comparing all depth values.

[0192] Step 703: Calculate the difference between the reverse secondary 3D coordinate information and the reverse initial 3D coordinate information, and define the difference as reverse distance difference information.

[0193] The reverse distance difference information is the difference between the reverse secondary 3D coordinate information and the reverse initial 3D coordinate information. Since the fourth initial cluster information only includes the reverse initial 3D coordinate information at this time, it is only necessary to calculate the difference between the two to determine whether it falls within the fourth initial cluster information.

[0194] Step 704: Determine whether the reverse distance difference information is greater than the reverse initial threshold information.

[0195] Step 7041: If it is greater than, the reverse second-selected three-dimensional coordinate information is used as the initial three-dimensional coordinate information of the next cluster, and the next cluster is defined as the fifth initial cluster information.

[0196] The fifth initial clustering information is the initial clustering information that includes the reverse secondary three-dimensional coordinate information. If it is greater than, it means that it does not fall within the fourth initial clustering information, so it can be used as the initial three-dimensional coordinate information for the next cluster to form the fifth initial clustering information. It should be noted that the cluster formed here is only for reference. When the distance value between it and the data point in the updated fourth initial clustering information is less than the threshold, the second initial clustering information is deleted. However, if the reverse secondary three-dimensional coordinate information forming the fifth initial clustering information is still within the fifth initial clustering information, it is not deleted.

[0197] Step 7042: If it is less than, the reverse second-selected three-dimensional coordinate information is added as the selected three-dimensional coordinate information to the fourth initial clustering information.

[0198] If it is less than, it means that it meets the requirements and can be added to the fourth initial clustering information, and it is updated to the selected three-dimensional coordinate information to reduce the number of repeated selections.

[0199] Step 705: Select the three-dimensional coordinate information with the smallest depth information except the reverse initial three-dimensional coordinate information and the reverse secondary three-dimensional coordinate information as the third three-dimensional coordinate information, and define the three-dimensional coordinate information as the reverse secondary three-dimensional coordinate information.

[0200] The reverse reselected three-dimensional coordinate information is the smallest three-dimensional coordinate information except the reverse initial three-dimensional coordinate information and the reverse second-selected three-dimensional coordinate information.

[0201] Step 706: determine in sequence whether the reversely selected three-dimensional coordinate information is added to the fourth initial clustering information or the fifth initial clustering information.

[0202] Step 707: When the reselected three-dimensional coordinate information is not added to the fourth initial clustering information and the fifth initial clustering information, the sixth initial clustering information is created until all the three-dimensional coordinate information has become the selected three-dimensional coordinate information.

[0203] Reference Figure 8 When the number of three-dimensional coordinate information corresponding to the depth information with the largest value is two or more, the method for selecting the initial three-dimensional coordinate includes:

[0204] Step 800: define the three-dimensional coordinate information corresponding to the depth information with the largest value as the deepest three-dimensional coordinate information.

[0205] The deepest three-dimensional coordinate information is the three-dimensional coordinate information corresponding to the depth information with the largest depth value. The purpose of defining it is to distinguish it from other three-dimensional coordinate information.

[0206] Step 801: Calculate the median value of the maximum horizontal coordinate information and the maximum vertical coordinate information, and define the straight line corresponding to the horizontal and vertical coordinates corresponding to the average value as the central axis information.

[0207] The central axis information is the axis perpendicular to the image information formed when the X and Y coordinates in the depth map are at their intermediate values and the Z coordinate is not restricted. Since the depth image necessarily covers all X and Y coordinates, the central axis information in the depth map only needs to be calculated from the intermediate X and Y coordinates.

[0208] Step 802: Calculate the distance between the deepest three-dimensional coordinate information and the central axis respectively, and define the distance value as the horizontal distance information.

[0209] The horizontal distance is the distance between the deepest 3D coordinate and the central axis. The vertical distance calculated here is the distance from a point to a line. Since the deepest 3D coordinate is a point and the central axis is a line parallel to the Z axis, the distance between the deepest 3D coordinate and the central axis does not have a vector in the Z axis direction. Therefore, we only need to calculate the square root of the difference between the X and Y coordinates.

[0210] Step 803: Sort the deepest three-dimensional coordinate information in ascending order according to the value of the horizontal distance to the middle distance information, and define the deepest three-dimensional coordinate information with the smallest value of the horizontal distance to the middle distance information as the initial three-dimensional coordinate information, and define the next deepest three-dimensional coordinate information as the secondary three-dimensional coordinate information.

[0211] By selecting the point closest to the middle point as the initial three-dimensional coordinate information, the area where the sphere formed intersects with the entire depth map is the largest. As a result, although the depth is the same, the coverage area closest to the middle point is the largest, which can reduce the calculation after the cluster is formed exponentially, greatly reducing the calculation workload and improving the efficiency of filtering.

[0212] Reference Figure 9 ,The calculation method of point cloud radius threshold information and dynamic clustering threshold information includes:

[0213] Step 900: Obtain demand target information.

[0214] The target information is the information of the object that needs to be identified and is manually input.

[0215] Step 901: Calculate depth parameter information and radius parameter information according to required target information.

[0216] Depth parameter information is used to calculate the dynamic clustering threshold. It's calculated as α = wide_min / dis, where wide_min is the product width and dis is the distance from the camera. For example, when detecting an object 3 meters away and wider than 10 cm, α = 10 cm / 3 meters = 1 / 30. Radius parameter information is used to calculate the point cloud radius threshold. It's calculated as β = radius_wide / dis, where β is determined based on the product's design objectives and radius_wide is the average distance between data points within a cluster. Both parameters are determined based on the product's design objectives, making it easier to identify products and improving filtering accuracy.

[0217] Step 902: Calculate the product of the depth average value information and the radius parameter information to obtain point cloud radius threshold information.

[0218] The calculation method is to multiply the two.

[0219] Step 903: Calculate the product of the dynamic depth average value information and the depth parameter information to obtain dynamic clustering threshold information.

[0220] The calculation method is to multiply the two.

[0221] Based on the same inventive concept, an embodiment of the present invention provides a 3D point cloud filtering system based on depth information, comprising:

[0222] Reference Figure 10 , a 3D point cloud filtering system based on depth information, comprising:

[0223] Image acquisition module 1003, used to acquire image information;

[0224] The processing module 1001 is connected to the image acquisition and judgment module 1002 and is used for storing and processing information;

[0225] The generating module 1004 is connected to the processing module 1001 and is used to generate a depth map according to the image information and generate three-dimensional coordinate information according to the coordinates of the depth map combined with the depth value, wherein the three-dimensional coordinate information includes abscissa information, ordinate information and depth information;

[0226] The selection module 1005 is connected to the processing module 1001 and is used to select one of the three-dimensional coordinate information that forms the initial cluster information including the three-dimensional coordinate information;

[0227] The generating module 1004 generates clustering information according to the three-dimensional coordinate information;

[0228] The calculation module 1006 is connected to the processing module 1001 and is used to arbitrarily select cluster information and calculate cluster center information, and define the selected cluster as the current cluster information;

[0229] The calculation module 1006 calculates the point cloud radius information based on the cluster center information;

[0230] The calculation module 1006 calculates the average depth information according to the depth information in the three-dimensional coordinate information corresponding to the current cluster information;

[0231] The calculation module 1006 calculates the point cloud radius threshold information based on the depth average value information;

[0232] A determination module 1002 is configured to determine the size of the point cloud radius information and the point cloud radius threshold information;

[0233] When the determination module 1002 determines that the point cloud radius information is less than the point cloud radius threshold information, the processing module 1001 regards the current clustering information as noise information and removes it from the image information.

[0234] Those skilled in the art will clearly understand that for the sake of convenience and brevity, the division of the above-mentioned functional modules is only used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working processes of the above-mentioned systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0235] The above are all preferred embodiments of the present application and are not intended to limit the scope of protection of this application. Unless otherwise stated, any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features. In other words, unless otherwise stated, each feature is merely an example of a series of equivalent or similar features.

Claims

1. A 3D point cloud filtering method based on depth information, characterized in that: include: Obtain image information; Generate a depth map based on the image information and generate three-dimensional coordinate information based on the coordinates of the depth map combined with the depth value, wherein the three-dimensional coordinate information includes horizontal coordinate information, vertical coordinate information and depth information; Generate clustering information based on three-dimensional coordinate information; The method for generating clustering information according to three-dimensional coordinate information includes: Selecting one of the three-dimensional coordinate information to form initial cluster information including the three-dimensional coordinate information, defining the three-dimensional coordinate information as selected three-dimensional coordinate information, and defining the three-dimensional coordinate information that has not been selected as unselected three-dimensional coordinate information; Calculate an average value based on the depth information of the data points in the initial clustering information, and define the average value as dynamic depth average information; Calculate dynamic clustering threshold information based on dynamic depth average information; arbitrarily selecting unselected three-dimensional coordinate information and calculating a distance value between the unselected three-dimensional coordinate information and the selected three-dimensional coordinate information, and defining the distance value as distance information; Filter out the smallest distance value in the distance information and define it as the minimum distance information; When the minimum distance information is less than the dynamic clustering threshold information, the unselected three-dimensional coordinate information is updated to the corresponding selected three-dimensional coordinate information in the initial clustering information; When the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the dynamic clustering threshold information, the dynamic clustering threshold information is updated and it is re-determined whether the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the updated dynamic clustering threshold information; When the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the updated dynamic clustering threshold information, the initial clustering information is updated to the clustering information; Reselecting unselected three-dimensional coordinate information and re-forming initial clustering information until no unselected three-dimensional coordinate information exists; Arbitrarily select cluster information and calculate cluster center information, and define the selected cluster as the current cluster information; Calculate the point cloud radius information based on the cluster center information; Calculate the average depth information based on the depth information in the three-dimensional coordinate information corresponding to the current cluster information; Calculate the point cloud radius threshold information based on the depth average information; When the point cloud radius information is less than the point cloud radius threshold information, the current cluster information is regarded as noise information and removed from the image information.

2. The 3D point cloud filtering method based on depth information according to claim 1, characterized in that: Methods for calculating point cloud radius information based on cluster center information include: Calculate the three-dimensional distance values of all data points and cluster center information in the current cluster information and filter out the value with the largest three-dimensional distance value as the point cloud radius information.

3. The 3D point cloud filtering method based on depth information according to claim 1, characterized in that: Another method for calculating point cloud radius information based on cluster center information includes: Calculate the one-dimensional distance value between all data points and cluster center information in the current cluster information, and define the one-dimensional distance value as the one-dimensional distance information; Filter the maximum and minimum values in any single-dimensional distance information respectively and calculate the difference between the two, and define the difference as the single-dimensional span information; The maximum value in the single-dimensional span information is selected and half of the maximum value is calculated to be defined as the point cloud radius information.

4. The 3D point cloud filtering method based on depth information according to claim 1, characterized in that: A method for selecting one of the three-dimensional coordinate information to form initial cluster information including the three-dimensional coordinate information includes: The three-dimensional coordinate information corresponding to the maximum value according to the depth information is used as the first three-dimensional coordinate information selected, the three-dimensional coordinate information is defined as the initial three-dimensional coordinate information, and the initial clustering information is defined as the first initial clustering information; Calculating initial threshold information based on depth information of initial three-dimensional coordinate information; selecting the three-dimensional coordinate information having the largest depth information other than the initial three-dimensional coordinate information as the second three-dimensional coordinate information, and defining the three-dimensional coordinate information as the secondary selected three-dimensional coordinate information; Calculating the difference between the secondary three-dimensional coordinate information and the initial three-dimensional coordinate information, and defining the difference as distance difference information; Determine whether the distance difference information is greater than the initial threshold information; If it is greater than, the second selected three-dimensional coordinate information is used as the initial three-dimensional coordinate information of the next cluster, and the next cluster is defined as the second initial cluster information; If it is less than, the second selected three-dimensional coordinate information is added to the first initial clustering information as the selected three-dimensional coordinate information; selecting the three-dimensional coordinate information having the largest depth information other than the initial three-dimensional coordinate information and the second-selected three-dimensional coordinate information as the third three-dimensional coordinate information, and defining the three-dimensional coordinate information as the second-selected three-dimensional coordinate information; determining in sequence whether the reselected three-dimensional coordinate information is to be added to the first initial clustering information or the second initial clustering information; When the reselected three-dimensional coordinate information is not added to the first initial clustering information and the second initial clustering information, the third initial clustering information is established until the three-dimensional coordinate information has become the selected three-dimensional coordinate information.

5. The 3D point cloud filtering method based on depth information according to claim 4, characterized in that: When all three-dimensional coordinate information forms initial cluster information, the cluster information forming method includes: Updating the initial threshold information into dynamic clustering threshold information according to the three-dimensional coordinate information in the first initial clustering information; respectively calculating the distance values between the three-dimensional coordinate information of other initial clustering information and the three-dimensional coordinate information in the first initial clustering information and selecting the minimum distance value, which is defined as the verification distance information; Determine whether the verification distance information is less than the dynamic clustering threshold information; If it is less than, delete the corresponding initial clustering information; If it is greater than, the corresponding initial clustering information is maintained; After the three-dimensional coordinate information in all other initial clustering information is verified, the first initial clustering information is updated and the other initial clustering information is updated.

6. The 3D point cloud filtering method based on depth information according to claim 1, characterized in that: Another method for selecting one of the three-dimensional coordinate information to form initial cluster information including the three-dimensional coordinate information includes: According to the depth information, the three-dimensional coordinate information corresponding to the minimum value is selected as the first three-dimensional coordinate information, the three-dimensional coordinate information is defined as the reverse initial three-dimensional coordinate information, and the initial clustering information is defined as the fourth initial clustering information; Calculating reverse initial threshold information based on depth information of reverse initial three-dimensional coordinate information; Selecting the three-dimensional coordinate information with the smallest depth information except the reverse initial three-dimensional coordinate information as the second three-dimensional coordinate information, and defining the three-dimensional coordinate information as the reverse secondary three-dimensional coordinate information; Calculating the difference between the reverse secondary 3D coordinate information and the reverse initial 3D coordinate information, and defining the difference as reverse distance difference information; Determine whether the reverse distance difference information is greater than the reverse initial threshold information; If it is greater than, the reverse second-selected three-dimensional coordinate information is used as the initial three-dimensional coordinate information of the next cluster, and the next cluster is defined as the fifth initial cluster information; If it is less than, the reverse second-selected three-dimensional coordinate information is added as the selected three-dimensional coordinate information to the fourth initial clustering information; selecting the three-dimensional coordinate information with the smallest depth information except the reverse initial three-dimensional coordinate information and the reverse second-selected three-dimensional coordinate information as the third three-dimensional coordinate information, and defining the three-dimensional coordinate information as the reverse second-selected three-dimensional coordinate information; determining in sequence whether the reversely selected three-dimensional coordinate information is to be added to the fourth initial clustering information or the fifth initial clustering information; When the reselected three-dimensional coordinate information is not added to the fourth initial clustering information and the fifth initial clustering information, the sixth initial clustering information is established until the three-dimensional coordinate information has become the selected three-dimensional coordinate information.

7. The 3D point cloud filtering method based on depth information according to claim 4, characterized in that: When the number of three-dimensional coordinate information corresponding to the depth information with the largest value is two or more, the method for selecting the initial three-dimensional coordinate information includes: The three-dimensional coordinate information corresponding to the depth information with the largest value is defined as the deepest three-dimensional coordinate information; Calculate the median value of the horizontal and vertical coordinates with the largest values, and define the straight line corresponding to the horizontal and vertical coordinates corresponding to the average value as the central axis information; Calculate the distance between the deepest three-dimensional coordinate information and the central axis respectively, and define the distance value as the horizontal distance information; After sorting the deepest three-dimensional coordinate information in ascending order according to the value of the horizontal distance to the middle distance information, the deepest three-dimensional coordinate information with the smallest value of the horizontal distance to the middle distance information is defined as the initial three-dimensional coordinate information, and the next deepest three-dimensional coordinate information is defined as the secondary three-dimensional coordinate information.

8. The 3D point cloud filtering method based on depth information according to claim 1, characterized in that: The calculation method of point cloud radius threshold information and dynamic clustering threshold information includes: Obtain target information; Calculate depth parameter information and radius parameter information according to the required target information; Calculate the product between the depth average information and the radius parameter information to obtain the point cloud radius threshold information; The dynamic clustering threshold information is obtained by calculating the product between the dynamic depth average information and the depth parameter information.

9. A 3D point cloud filtering system based on depth information, characterized in that: include: An image acquisition module, used for acquiring image information; A processing module, connected to the image acquisition and judgment module, is used for storing and processing information; A generation module, connected to the processing module, is used to generate a depth map according to the image information and generate three-dimensional coordinate information according to the coordinates of the depth map combined with the depth value, wherein the three-dimensional coordinate information includes horizontal coordinate information, vertical coordinate information and depth information; The generation module generates clustering information based on the three-dimensional coordinate information; A calculation module, connected to the processing module, is used to arbitrarily select cluster information and calculate cluster center information, and define the selected cluster as current cluster information; The calculation module calculates the point cloud radius information based on the cluster center information; The calculation module calculates the depth average information based on the depth information in the three-dimensional coordinate information corresponding to the current cluster information; The calculation module calculates the point cloud radius threshold information based on the depth average information; A judgment module is used to judge the size of the point cloud radius information and the point cloud radius threshold information; When the judgment module determines that the point cloud radius information is less than the point cloud radius threshold information, the processing module regards the current cluster information as noise information and removes it from the image information; The generating module may also be used to select one of the three-dimensional coordinate information to form initial clustering information including the three-dimensional coordinate information, define the three-dimensional coordinate information as selected three-dimensional coordinate information, and define the three-dimensional coordinate information that has not been selected as unselected three-dimensional coordinate information; Calculate an average value based on the depth information of the data points in the initial clustering information, and define the average value as dynamic depth average information; Calculate dynamic clustering threshold information based on dynamic depth average information; arbitrarily selecting unselected three-dimensional coordinate information and calculating a distance value between the unselected three-dimensional coordinate information and the selected three-dimensional coordinate information, and defining the distance value as distance information; Filter out the smallest distance value in the distance information and define it as the minimum distance information; When the minimum distance information is less than the dynamic clustering threshold information, the unselected three-dimensional coordinate information is updated to the corresponding selected three-dimensional coordinate information in the initial clustering information; When the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the dynamic clustering threshold information, the dynamic clustering threshold information is updated and it is re-determined whether the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the updated dynamic clustering threshold information; When the minimum distance information corresponding to the unselected three-dimensional coordinate information is greater than the updated dynamic clustering threshold information, the initial clustering information is updated to the clustering information; The unselected three-dimensional coordinate information is reselected and the initial clustering information is re-formed until the unselected three-dimensional coordinate information no longer exists.

Images