A warehouse environment data collection method and system based on the Internet of Things

Abstract

The present application relates to the technical field of data processing, and more particularly to a warehouse environment data collection method and system based on the Internet of Things. The method comprises: collecting warehouse environment data using a laser radar, the warehouse environment data being point cloud data, and denoising the warehouse environment data using a radius filtering algorithm to achieve collection of the warehouse environment data; in the process of denoising using the radius filtering algorithm, comprising: for a target grid in a plurality of grids, determining an outlier feature of data points in the target grid; determining a data density of the target grid; determining an intra-cluster aggregation degree of a cluster in the target grid; determining an inter-cluster separation degree of the cluster in the target grid; and determining a minimum neighbor number of the target grid. The radius filtering of the dynamic minimum neighbor number can dynamically adjust the filtering parameters according to the local features of the point cloud data in each grid, more accurately remove noise and outlier data points, and collect more accurate warehouse environment data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a warehouse environment data collection method and system based on the Internet of Things. BACKGROUND

[0002] In the automobile supply chain warehouse, a large number of automobile production parts are stored, and these parts are of various types and shapes, and have strict requirements on the storage environment. With the development of artificial intelligence and Internet of Things technology, its application in warehouse management has gradually been valued and promoted. Warehouse robots collect point cloud data of the warehouse environment through devices such as laser radars or three-dimensional scanners, and transmit the data through Internet of Things technology to realize intelligent management of the warehouse environment. However, point cloud data often contains a large amount of noise and redundant information, which will affect the accuracy and reliability of the data, and thus affect the effectiveness of warehouse management. Therefore, effective filtering of point cloud data to remove noise and redundant information is the key to improving data collection quality and efficiency. In the process of collecting point cloud data in the automobile supply chain warehouse, radius filtering, as an effective filtering method, can be applied in the preprocessing stage of point cloud data. Radius filtering searches the neighborhood of each point in the point cloud data by setting a reasonable radius range, and then determines whether the point is a noise point or a redundant point according to the number or attribute characteristics of the points in the neighborhood, and performs corresponding processing.

[0003] The patent application file with publication number CN117422622A discloses a sonar image denoising method based on adaptive combination filtering, S1: importing the point cloud data measured by the sonar into operation; S2: performing radius filtering according to the point cloud spacing, setting the search radius and domain value to remove point clouds with high dispersion; S3: setting the slice thickness and the distance between adjacent slices to perform slice processing on the point cloud data, simplifying the data set and improving the denoising effect; S4: calculating the mean and standard deviation to set the corresponding threshold and using a statistical filtering method to process the sonar image; S5: using combination filtering to perform point cloud filtering on the sonar data; S6: performing precision inspection and data visualization comparison.

[0004] However, the point cloud data in the warehouse environment is often non-uniform. Due to the placement, shape, size and occlusion of objects, the density and distribution of point cloud data will have significant differences in different areas. The above patent application file does not solve the problem that when using traditional radius filtering for denoising, due to the use of the minimum number of neighbors in the global neighborhood, it cannot adapt to the differences in density and distribution of point cloud data in different areas, resulting in over-filtering of some areas and under-filtering of some areas, and thus reducing the accuracy of the collected warehouse environment data. SUMMARY

[0005] In order to solve the problem that when the traditional radius filtering is used for denoising, due to the minimum number of neighbors in the global neighborhood, the density and distribution of point cloud data in different regions will appear differences, leading to over-filtering of data in some regions and under-filtering of data in some regions, and further reducing the accuracy of the collected warehouse environment data, the application provides a warehouse environment data collection method and system based on Internet of Things.

[0006] In the first aspect, the application provides a warehouse environment data collection method based on Internet of Things, which adopts the following technical scheme:

[0007] A warehouse environment data collection method based on Internet of Things, comprising:

[0008] For a target grid in the plurality of grids, the outlying feature of a target data point in the target grid is determined according to the distance between the target data point and the adjacent data points, and the number of adjacent data points; the point corresponding to the mean value of the coordinates of each data point is taken as a feature point, the distance from each data point to the feature point is weighted and summed according to the outlying feature of each data point, to obtain the point cloud aggregation degree of the target grid, and further determine the data density of the target grid; clustering is performed on each data point to obtain a plurality of class clusters, and for a target class cluster in the plurality of class clusters, the intra-cluster aggregation degree of the target class cluster is determined according to the mean value of the distance from each data point in the target class cluster to the cluster center; the distance between the edge points of the target class cluster and the nearest class cluster to the target class cluster is taken as the inter-cluster separation degree of the target class cluster; the target class cluster is split and merged according to the size of the intra-cluster aggregation degree and the inter-cluster separation degree, and the minimum number of neighbors of the target grid is determined according to the number of all class clusters after splitting and merging, and the data density.

[0009] The beneficial effects are that by adjusting the denoising parameters according to the local density of point cloud data, the problems of "over-filtering" and "under-filtering" in traditional radius filtering are avoided, and the data processing accuracy in different regions is improved; by combining grid division, outlier detection and clustering analysis, the denoising process is optimized, the real data can be more accurately preserved, and the influence of noise and abnormal points is reduced; by weighted calculation of the relationship between data points and neighbors, the accuracy of data is enhanced, and the point cloud data in different regions is properly processed; higher quality point cloud data is provided, so that the warehouse automation equipment can navigate and position more accurately, and the stability and accuracy of the system are improved.

[0010] Further, the point cloud data includes three-dimensional space coordinates, intensity information and color information.

[0011] Further, the outlying feature satisfies the following relationship:

[0012] ; in the formula, is the Within the first grid Outlier characteristics of each data point For the first Within the first grid The number of neighboring data points of a data point For the first Within the first grid From the data point to the The distance between neighboring data points.

[0013] The beneficial effects are as follows: by calculating the average distance between each data point and its neighboring points, outliers can be effectively identified, which helps to remove noise from point cloud data and improve data accuracy; based on the calculation of outlier features, outliers can be accurately removed in areas with uneven data density or a lot of noise, thus optimizing data quality.

[0014] Furthermore, the data density satisfies the following relationship:

[0015] In the formula, For the first Data density of each grid For the first The number of data points within each grid. For the first The degree of point cloud aggregation of each grid, For the first Within the first grid Outlier characteristics of each data point For the first Within the first grid The distance from each data point to the feature point This is the normalization function.

[0016] The beneficial effects are as follows: by combining outlier features and the distance from data points to feature points, the degree of data aggregation can be described more accurately, especially when the point cloud is unevenly distributed, it can provide more accurate local density estimation; it can adaptively adjust the density estimation according to the outlier features of data points, avoid the errors generated by traditional methods in areas with large density changes, and help identify sparse and dense areas.

[0017] Furthermore, the degree of clustering within the cluster satisfies the following relationship:

[0018] In the formula, For the first Within the first grid The degree of intra-cluster aggregation of each class of clusters, For the first Within the first grid The number of data points within each cluster For the first Within the first grid Within the cluster, the th Data points to cluster center distance, It is a natural exponential function.

[0019] The beneficial effects are as follows: by calculating the distance of each data point to the cluster center and applying the natural exponential function, the degree of clustering of data points within a cluster can be accurately measured; by weighted averaging the distances of data points within a cluster, the degree of clustering can be adapted to different point densities within the cluster, and even in cases of uneven density, the overall degree of clustering can be judged more accurately.

[0020] Furthermore, the edge point is the data point on the straight line connecting the cluster centers of two clusters that is farthest from the cluster center of its own cluster; if there is no data point on the straight line, then the data point within the cluster that is farthest from the cluster center and closest to the straight line is selected as the edge point.

[0021] Further, the splitting and merging of the target cluster includes: splitting the target cluster in response to the intra-cluster aggregation degree being less than a preset cluster splitting threshold; merging the target cluster and the cluster closest to the target cluster in response to the inter-cluster separation degree being less than a preset cluster merging threshold; and stopping the splitting and merging of the target cluster in response to the intra-cluster aggregation degree being less than the preset cluster splitting threshold and the inter-cluster separation degree being less than the preset cluster merging threshold.

[0022] Furthermore, the minimum number of neighbors satisfies the following relationship:

[0023] In the formula, For the first Minimum number of neighbors for each grid. For the first Data density of each grid For the first The number of clusters within a grid, For hyperparameters, It is a natural exponential function. This is the floor symbol.

[0024] The beneficial effect is that by combining the minimum number of neighbors with data density and the number of clusters, an adaptive adjustment of the number of neighbors for each grid is achieved. In high-density areas, the number of neighbors automatically increases, thereby capturing local structure more accurately, while in low-density areas, the number of neighbors decreases, avoiding overcomputation.

[0025] Further, the collection of the warehouse environment data comprises: taking twice the minimum number of neighbors as a filtering radius of a target grid, in response to the number of neighbors within the filtering radius of the target grid being less than the minimum number of neighbors of the target grid, the warehouse environment data corresponding to the target grid is excluded, otherwise, the warehouse environment data corresponding to the target grid is denoised by radius filtering to obtain denoised warehouse environment data corresponding to the target grid, and the denoised warehouse environment data is transmitted to a database to complete the collection of the warehouse environment data based on the Internet of Things.

[0026] In a second aspect, the present application provides a warehouse environment data collection system based on the Internet of Things, which adopts the following technical solution:

[0027] The warehouse environment data collection system based on the Internet of Things comprises a processor and a memory, and the memory stores computer program instructions, which, when executed by the processor, implement the warehouse environment data collection method based on the Internet of Things.

[0028] By adopting the above technical solution, the warehouse environment data collection method based on the Internet of Things is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is made according to the memory and the processor, and the use is facilitated.

[0029] The present application has the following technical effects:

[0030] This invention utilizes radius filtering based on the dynamic minimum neighbor count to dynamically adjust filtering parameters according to the local characteristics of point cloud data within each grid. This allows for more accurate removal of noisy and outlier data points while retaining valid data points. It solves the problem of traditional radius filtering, which uses the minimum neighbor count within the global neighborhood and cannot adapt to differences in density and distribution across different regions of the point cloud data, resulting in some areas being filtered excessively while others are under-filtered, thus reducing the accuracy of warehouse environment data. This targeted processing better ensures the accuracy of the collected warehouse environment data. By dividing the warehouse environment point cloud data into grids with a preset grid width, the relevant characteristics of each grid can be obtained. For example, outlier characteristics of each data point can be determined first, and then... The calculation of the point cloud clustering degree of the grid determines the data density. It can also analyze the intra-cluster clustering degree and inter-cluster separation degree of the data points within the grid after clustering, so as to accurately grasp the data situation within the grid from multiple aspects. At the same time, grid partitioning divides the large-scale point cloud data into multiple smaller processing units for parallel processing, making the data processing within each grid more efficient and improving the efficiency of data processing. Based on factors such as the intra-cluster clustering degree, inter-cluster separation degree, and the number of clusters after splitting and merging, as well as the data density, the minimum number of neighbors of the grid is determined. This allows the minimum number of neighbors to better fit the actual situation of each grid during the radius filtering algorithm denoising process, optimize the denoising effect, and improve the overall quality of warehouse environment data collection. Attached Figure Description

[0031] Figure 1 This is a flowchart of a warehouse environment data collection method based on the Internet of Things according to an embodiment of the present invention.

[0032] Figure 2 This is a flowchart of step S2 in an embodiment of the present invention, which is a warehouse environment data acquisition method based on the Internet of Things.

[0033] Figure 3 This is a schematic diagram of edge points in a warehouse environment data acquisition method based on the Internet of Things according to an embodiment of the present invention; wherein, 1 represents a cluster. The cluster center, 2 is the cluster The cluster center of the nearest neighbor cluster, 3 is the cluster. Edge points, 4 is a cluster The edge point of the nearest neighbor cluster. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] This invention discloses a warehouse environment data collection method based on the Internet of Things (IoT), referring to... Figure 1 This includes steps S1-S3:

[0036] S1: Use lidar to collect warehouse environmental data, which is point cloud data.

[0037] Warehouse robots use LiDAR to scan the warehouse environment, thereby obtaining warehouse environmental data.

[0038] Specifically, the point cloud data includes three-dimensional spatial coordinates, intensity information (objects with high reflectivity, such as mirrors or metal surfaces, will have higher intensity of reflected light, while objects with low reflectivity, such as rough black fabric, will have lower intensity of reflected light) and color information, etc.

[0039] S2: Use the radius filtering algorithm to denoise the warehouse environment data.

[0040] It should be noted that the point cloud data of the warehouse environment usually contains a large number of discrete three-dimensional points, which reflect the shape, position and attribute of the objects inside the warehouse. When processing these data, it is necessary to accurately identify the boundaries, features of the objects and the spatial relationship between them. At the same time, due to the different types, sizes and positions of the objects in the warehouse, the distribution and noise level of the point cloud data may also differ, that is, the point cloud data in the warehouse environment is often non-uniform, and the radius filtering is performed on the data according to the filtering radius and the minimum number of neighbors. The minimum number of neighbors in the global neighborhood used in the traditional radius filtering cannot adapt to this non-uniform distribution, which may cause over-filtering of some area data and under-filtering of some area data. Therefore, the point cloud data is divided into grids, and different minimum numbers of neighbors are set for the grids to perform parallel (at the same time) filtering processing, which can not only improve the filtering efficiency, but also improve the phenomenon of insufficient or over-filtering of some data points. When obtaining the minimum number of neighbors corresponding to each grid, the density characteristics of the point cloud data distribution in the grid and the types (number of clusters after clustering) of the point cloud data can be obtained. This is because: on the one hand, in the warehouse environment, due to the placement, shape and size of the objects, the distribution density of the point cloud data may differ significantly, for example, in areas with dense shelves, the point cloud data may be very dense, and if the number of neighbors is too small, the local features of the objects may not be accurately reflected, and in open or less object areas, the point cloud data may be relatively sparse, and if the number of neighbors is too large, unnecessary calculation and noise may be introduced; on the other hand, in the warehouse environment, the point cloud data may contain objects of different categories, such as various parts of cars, shelves, etc., and objects of different categories have different shapes, sizes and features, so their point cloud data may also differ. If the number of categories of point cloud data in the grid is greater, it indicates that there may be objects of different shapes, sizes and features in the grid, and in order to accurately describe and distinguish these objects, more local feature information is needed, that is, more neighbors are needed to capture such details.

[0041] In summary, the minimum number of neighbors corresponding to the grid is obtained according to the density of the point cloud data distribution in the grid and the types (number of clusters after clustering) of the point cloud data. The higher the density of the point cloud data distribution in the grid or the more types, the more neighbors are needed, and the larger the corresponding minimum number of neighbors.

[0042] Reference Figure 2 In the process of denoising using the radius filtering algorithm, the warehouse environment data is divided into multiple grids according to a predetermined grid width (20 in the present application, which can be set by the implementer according to the specific implementation), and the minimum number of neighbors for determining the filtering radius of each grid is obtained, including steps S201-S205, as follows:

[0043] S201: For a target grid in the plurality of grids, determine an outlier feature of a data point in the target grid.

[0044] It should be noted that the point cloud data density in the grid specifically refers to the relationship between the amount of data in the grid and the degree of data aggregation. The more data in the grid and the more aggregated the data, the higher the data density in the grid. When obtaining the degree of data aggregation in the grid, the distance between the point cloud data point in the grid and a certain feature point is generally used to describe it. However, since there may be noise in the point cloud data point, some data may deviate, and the distance between such data point and feature point is far, which will have a great influence on the degree of aggregation when participating in operation, and needs to be eliminated to improve the accuracy of the degree of aggregation. Therefore, the outlier feature of each data point in the grid needs to be calculated first.

[0045] According to the distance between the target data point in the target grid and the adjacent data points, and the number of adjacent data points, the outlier feature of the target data point is determined, and the distance is the Euclidean distance.

[0046] Specifically, the outlier feature satisfies the following relationship:

[0047] ;

[0048] In the formula, is the outlier feature of the i-th data point in the j-th grid, is the number of adjacent data points of the i-th data point in the j-th grid, is the distance from the i-th data point in the j-th grid to the k-th adjacent data point. wherein, represents the mean value of the distance from the i-th data point in the j-th grid to the adjacent data point, the greater the value, the farther the distance between the i-th data point and its adjacent data point, and the more obvious the outlier feature of the i-th data point. S202: Determine the data density of the target grid. The point corresponding to the mean value of the coordinates of each data point is taken as a feature point, the distance from each data point to the feature point is weighted and summed according to the outlier feature of each data point, the point cloud aggregation degree of the target grid is obtained, and the data density of the target grid is determined.

[0049]

[0050] S202: Determine the data density of the target grid.

[0051] The point corresponding to the mean value of the coordinates of each data point is taken as a feature point, the distance from each data point to the feature point is weighted and summed according to the outlier feature of each data point, the point cloud aggregation degree of the target grid is obtained, and the data density of the target grid is determined.

[0052] ​​​​​​​​​Specifically, the data density satisfies the following relationship:

[0053] ;

[0054] In the formula, For the first Data density of each grid For the first The number of data points within each grid. For the first The degree of point cloud aggregation of each grid, For the first Within the first grid Outlier characteristics of each data point For the first Within the first grid The distance from each data point to the feature point This is the normalization function.

[0055] in, Indicates the first Within the first grid The weight of the i-th data point, within the grid. The more obvious the outlier characteristics of a data point, the lower the reliability of its distance from the feature points in the grid, and the smaller the corresponding weight. Indicates the first The degree of point cloud aggregation in a grid is determined by the smaller the weighted distance between data points and feature points within the grid. The more data points in the grid and the more clustered they are, the higher the grid density.

[0056] S203: Determine the degree of intra-cluster aggregation of clusters within the target grid.

[0057] It should be noted that the simplest and most effective way to obtain the types of point cloud data within a grid is to cluster the data points within the grid. The clustering process groups similar point cloud data points into one class, and the final number of clusters is the number of point cloud categories within the grid. However, noise is unavoidable during the point cloud data acquisition process, which leads to poor clustering results. Therefore, the clustering results can be analyzed, and some clusters can be split or merged to improve the accuracy of the final number of clusters.

[0058] Clustering is performed on each data point to obtain multiple clusters. For the target cluster among the multiple clusters, the degree of clustering within the target cluster is determined based on the average distance from each data point within the target cluster to the cluster center.

[0059] Implementers can choose clustering methods based on the specific implementation situation, such as K-means clustering.

[0060] Specifically, the degree of clustering within the cluster satisfies the following relationship:

[0061] ;

[0062] In the formula, For the first Within the first grid The degree of intra-cluster aggregation of each cluster, For the first Within the first grid The number of data points within each cluster For the first Within the first grid Within the cluster, the th Data points to cluster center distance, It is a natural exponential function.

[0063] in, Indicates the first Within the first grid The mean distance between data points within each cluster and the cluster center; the smaller this value, the better. The higher the degree of intra-cluster aggregation of a class of clusters.

[0064] S204: Determine the inter-cluster separation degree of clusters within the target grid.

[0065] The distance between the edge points of the target cluster and the cluster closest to the target cluster is used as the inter-cluster separation degree of the target cluster.

[0066] Specifically, the edge point is the data point on the straight line connecting the centers of two clusters that is farthest from the center of its own cluster; if there is no data point on the straight line, then the data point within the cluster that is farthest from the cluster center and closest to the straight line is selected as the edge point, such as... Figure 3 As shown, assume point 1 is a cluster. The cluster center, point 2 is the cluster The cluster center of the nearest neighbor cluster, the cluster The edge points are the clusters on the line connecting points 1 and 2. The point 3 furthest from the cluster center, the cluster The edge point of the nearest neighbor cluster is the point 4, which is the farthest from the cluster center in the nearest neighbor cluster on the line connecting 1 and 2. If there is no point in the cluster on the line, the point that is farthest from the cluster center and closest to the line is selected as the edge point.

[0067] Among them, the The greater the distance between an edge point within a cluster and the edge point of its nearest cluster, the greater the distance between the edge point of the i-th cluster. The higher the inter-cluster separation of each cluster, the better.

[0068] S205: Determine the minimum neighbor number of the target grid.

[0069] It should be noted that some clusters with poor clustering effect are split because the clusters may contain subsets; some clusters with fuzzy boundaries are merged because the clusters may belong to the same cluster.

[0070] According to the size of the cluster aggregation degree and the cluster separation degree, the target cluster is split and merged.

[0071] Specifically, the splitting and merging of the target cluster comprises:

[0072] In response to the cluster aggregation degree being less than a preset cluster splitting threshold, the target cluster is split;

[0073] In response to the cluster separation degree being less than a preset cluster merging threshold, the target cluster and the nearest cluster to the target cluster are merged;

[0074] In response to the cluster aggregation degree being less than a preset cluster splitting threshold and the cluster separation degree being less than a preset cluster merging threshold, the splitting and merging of the target cluster is stopped.

[0075] The cluster splitting threshold is the average of the cluster aggregation degrees of all clusters in the grid. When the cluster aggregation degree of a cluster in the grid is less than the preset threshold, it is considered that the clustering effect of the cluster is poor and there may be a sub-class structure, and the cluster is split. The cluster merging threshold is the average of the cluster separation degrees of all clusters in the grid. When the cluster separation degree of a cluster in the grid is less than the preset threshold, it is considered that the distance between the cluster and the nearest neighbor cluster is fuzzy, and they may belong to the same cluster, so the cluster is merged. When the cluster aggregation degree and the cluster separation degree of a cluster in the grid are both less than the preset threshold, it is considered that the cluster may have poor clustering effect and sub-class structure, and the distance between the cluster and the nearest neighbor cluster is fuzzy, and they may belong to the same cluster, which is a complex situation and cannot be handled by simple splitting or merging operation, so the cluster is not split or merged.

[0076] According to the number of all clusters after splitting and merging and the data density, the minimum neighbor number of the target grid is determined.

[0077] Specifically, the minimum neighbor number satisfies the following relationship:

[0078] ;

[0079] In the formula, is the minimum neighbor number of the i-th grid, is the minimum neighbor number of the i-th grid, and is the minimum neighbor number of the i-th grid. data density of a grid, the number of clusters in the first grid, a hyper parameter, is a natural exponential function, is a ceiling symbol.

[0080] The hyper parameter can be set by the implementer according to the specific implementation, for example, 10.

[0081] When the data density of the grid is high or the number of categories is large, more neighbors are needed to capture such details, and the minimum number of neighbors is also larger.

[0082] S3: to realize the collection of warehouse environment data.

[0083] Specifically, the collection of warehouse environment data includes:

[0084] The double of the minimum number of neighbors is used as the filtering radius of the target grid, and the warehouse environment data corresponding to the target grid is excluded in response to the number of neighbors in the filtering radius of the target grid being less than the minimum number of neighbors of the target grid. Otherwise, the warehouse environment data corresponding to the target grid is denoised by radius filtering to obtain denoised warehouse environment data corresponding to the target grid, and the denoised warehouse environment data is transmitted to the database, so as to facilitate more accurate analysis of the warehouse environment in the future, and complete the warehouse environment data collection based on the Internet of Things.

[0085] The embodiment of the application also discloses a warehouse environment data collection system based on the Internet of Things, comprising a processor and a memory, and the memory stores computer program instructions, which realize the warehouse environment data collection method based on the Internet of Things when executed by the processor.

[0086] The above system also includes a communication bus and a communication interface and other components familiar to those skilled in the art, the setting and function of which are known in the art, so hereinafter will not be repeated.

[0087] The above are the preferred embodiments of the application, not limited to the protection scope of the application, therefore: any equivalent changes made according to the structure, shape, principle of the application, should be covered within the protection scope of the application.

Claims

1. An Internet of Things based warehouse environment data collection method, characterized by, The warehouse environment data is collected by using a laser radar, and the warehouse environment data is point cloud data. In the process of denoising by using the radius filtering algorithm, the warehouse environment data is divided into a plurality of grids, and the minimum neighbor number of each grid is obtained. For a target grid in the plurality of grids, the outlying feature of a target data point is determined according to the distance between the target data point and adjacent data points in the target grid and the number of adjacent data points. The coordinates of the data points are weighted and summed to obtain the point cloud aggregation degree of the target grid, and the data density of the target grid is determined. 2.The warehouse environment data collection method based on the Internet of Things according to claim 1, characterized in that, The point cloud data includes three-dimensional space coordinates, intensity information and color information. 3.The warehouse environment data collection method based on the Internet of Things according to claim 1, characterized in that, The outlying feature satisfies the following relationship: ； wherein is the number of data points in the first grid, is the number of data points in the first grid, is the outlier characteristic of the data point in the first grid, is the number of neighboring data points of the data point in the first grid, is the number of neighboring data points of the data point in the first grid, is the distance of the data point in the first grid to the neighboring data point, is the distance of the data point in the first grid to the neighboring data point, is the distance of the data point in the first grid to the neighboring data point, is the distance of the data point in the first grid to the neighboring data point, is the distance of the data point in the first grid to the neighboring data point. 4.The warehouse environment data collection method based on the Internet of Things according to claim 1, characterized in that, The data density satisfies the following relationship: ； wherein, is the data density of the th grid, is the number of data points in the th grid, is the point cloud clustering degree of the th grid, is the outlier feature of the th data point in the th grid, is the distance from the th data point in the th grid to the feature point, is a normalization function. 5.The warehouse environment data collection method based on the Internet of Things according to claim 1, characterized in that, The intra-cluster aggregation degree satisfies the following relationship: ； In the formula, For the first Within the first grid The degree of intra-cluster aggregation of each class of clusters, For the first Within the first grid The number of data points within each cluster For the first Within the first grid Within the cluster, the th Data points to cluster center distance, It is a natural exponential function. 6.The warehouse environment data collection method based on the Internet of Things according to claim 1, characterized in that, The edge point is a data point farthest from the cluster center of the cluster on the straight line connecting the cluster centers of the two clusters. 7.The warehouse environment data collection method based on the Internet of Things according to claim 1, characterized in that, If there is no data point on the straight line, the data point farthest from the cluster center in the cluster and closest to the straight line is selected as the edge point. The target cluster is split and merged, including: In response to the intra-cluster aggregation degree being less than a preset cluster splitting threshold, the target cluster is split; In response to the inter-cluster separation degree being less than a preset cluster merging threshold, the target cluster and the cluster closest to the target cluster are merged; 8.The warehouse environment data collection method based on the Internet of Things according to claim 1, characterized in that, In response to the intra-cluster aggregation degree being less than a preset cluster splitting threshold and the inter-cluster separation degree being less than a preset cluster merging threshold, the splitting and merging of the target cluster are stopped. ； wherein is the minimum number of neighbors for the th grid, is the data density for the th grid, is the number of clusters within the th grid, is a hyperparameter, is the natural exponential function, is the ceiling function. 9.The warehouse environment data collection method based on the Internet of Things according to claim 1, characterized in that, The minimum neighbor number satisfies the following relationship: The warehouse environment data is collected by using a laser radar, and the warehouse environment data is point cloud data.

10. An Internet of Things based warehouse environment data collection system, characterized by, The minimum neighbor number is doubled as the filtering radius of the target grid, and in response to the neighbor number in the filtering radius of the target grid being less than the minimum neighbor number of the target grid, the warehouse environment data corresponding to the target grid is excluded. Otherwise, the warehouse environment data corresponding to the target grid is denoised by using radius filtering to obtain denoised warehouse environment data corresponding to the target grid, and the denoised warehouse environment data is transmitted to a database to complete the warehouse environment data collection based on the Internet of Things. The warehouse environment data is collected by using a laser radar, and the warehouse environment data is point cloud data. A processor and a memory, the memory storing computer program instructions which, when executed by the processor, implement an Internet of Things based warehouse environment data collection method according to any one of claims 1-9.

Images