A Method, Device, Equipment and Medium for Filtering Dust in Mining Areas Based on Point Cloud

Abstract

The present invention discloses a method, device, equipment and medium for filtering dust in a mining area based on point cloud. By projecting lidar point cloud data onto a depth image, a projected depth image is obtained, overcoming the difficulty that the point cloud density is affected by distance in the Cartesian coordinate system. After projection, combined with the obstacle clustering information obtained in the Cartesian coordinate system, the features of the clustering clusters on the point cloud projection depth image are extracted on the point cloud distribution projection map, and then the extracted features are input into a trained classification model to obtain the dust probability corresponding to each clustering cluster, so as to remove the obstacles corresponding to the dust with the cluster label from the obstacle sequence according to the obtained dust probability, achieving the effect of filtering dust.

Description

Technical Field

[0001] The present invention relates to the technical field of obtaining obstacle information in mining areas, and particularly to a method, device, equipment and medium for filtering dust in mining areas based on point clouds. Background Art

[0002] Dust flying in the mining area environment is a very common phenomenon, which endangers the safety of driverless vehicles in the mining area. Driverless vehicles using lidar are prone to taking the point clouds generated by dust flying as obstacle point clouds, and behaviors such as pre-obstacle parking or emergency obstacle avoidance caused thereby are prone to cause safety accidents under heavy load conditions. Therefore, dust filtering of lidar point clouds is a very important task. Since the road surface in public roads or other complex scenarios is mainly concrete, the improvement of lidar-based point cloud quality focuses on noise filtering, mainly using the spatial density of point clouds in the Cartesian coordinate system to filter out noise points or noise point clusters with relatively small density.

[0003] In the prior art, generally, point clouds are obtained by lidar, the point clouds are downsampled and their point cloud density is calculated as a reference threshold, the point cloud density within the neighborhood range of each point is obtained by traversal and compared with the reference threshold to obtain a noise point set, and after clustering the noise point set, clusters with a number less than the reference threshold in the noise point clustering clusters are filtered out as the denoising result. However, this method does not consider the influence of the reflection point distance on the point cloud density obtained by the lidar, and the calculation within the neighborhood space range has problems of high complexity and time consumption, and is suitable for scenarios with fewer noise points and low requirements for real-time performance. In an environment where dust flying is large in the mining area and the intensity of dust flying fluctuates due to weather influence, it will lead to insufficient performance and unstable performance. Summary of the Invention

[0004] The present invention provides a method, device, computer equipment and medium for filtering dust in mining areas based on point clouds, so as to achieve the effect of correcting misdetected obstacles and filtering dust for the obstacle perception function in a mine environment with more dust flying.

[0005] In a first aspect, a method for filtering dust in mining areas based on point clouds is provided, including:

[0006] According to the point cloud data of the lidar in the mining area received, a point cloud projection depth image is obtained;

[0007] Clustering the projection points of each point cloud data located on the point cloud projection depth image to obtain clustering clusters corresponding to each obstacle;

[0008] Extracting the features of each clustering cluster on the point cloud projection depth image;

[0009] Input the features extracted from each of the clustering clusters into the trained classification model to obtain the dust probability corresponding to each of the clustering clusters, and remove the obstacles corresponding to the clustering clusters with the dust probability higher than the preset threshold from the obstacle sequence.

[0010] Further, clustering the projection points of each point cloud data on the point cloud projection depth image to obtain clustering clusters corresponding to each obstacle, including:

[0011] Segment the point cloud projection depth image into a plurality of grids based on the planar size of the point cloud projection depth image;

[0012] Filter out the ground grids according to the height difference of the projection points of each point cloud data corresponding to each grid and the height difference between two adjacent grids to obtain non-ground grids;

[0013] Cluster the non-ground grids according to their planar relative positions to obtain clustering clusters corresponding to each of the obstacles, and generate cluster labels corresponding to each of the clustering clusters.

[0014] Further, inputting the features extracted from each of the clustering clusters into the trained classification model to obtain the dust probability corresponding to each of the clustering clusters, including:

[0015] Analyze the obstacle categories to which each of the clustering clusters belongs, and distinguish and calibrate the cluster labels corresponding to the clustering clusters whose obstacle category is dust and the cluster labels corresponding to the clustering clusters whose obstacle category is non-dust;

[0016] Input the calibration result of the cluster labels into a preset classification model for training to obtain the trained classification model.

[0017] Further, extracting the features of each of the clustering clusters on the point cloud projection depth image, including:

[0018] Determine the distribution range image of each of the clustering clusters on the point cloud projection depth image;

[0019] Extract features from the point cloud projection depth image corresponding to each of the distribution range images according to the preset feature extraction rules.

[0020] Further, the features include at least one of row sparsity feature, column density non-uniformity feature, and average density feature. Extracting features from the point cloud projection depth image corresponding to each of the distribution range images according to the preset feature extraction rules includes:

[0021] According to the distribution range image of each of the clustering clusters on the point cloud projected depth image, each of the distribution range images is divided by rows to obtain a row range image;

[0022] Based on the accommodable projection number and the actual projection number of the projection points of the point cloud data in the row range image, the row sparsity feature is obtained;

[0023] Based on the variance between the number of the row range images and the actual projection number of the projection points of the point cloud data in the row range images, the column density non-uniformity feature is obtained;

[0024] Based on the distribution range image of each of the clustering clusters on the point cloud projected depth image, and the accommodable projection number and the actual projection number of the projection points of the point cloud data in the distribution range image, the average density feature is obtained.

[0025] Further, the step of dividing each of the distribution range images by rows according to the distribution range image of each of the clustering clusters on the point cloud projected depth image to obtain a row range image includes:

[0026] Based on the preset weights of the row range images corresponding to the highest position and the lowest position located on the distribution range image, and the number of the row range images within the distribution range image, the weighted ratio corresponding to each row range image is determined;

[0027] According to the weighted ratio corresponding to each row range image, the row width of this row range image is weighted and calculated to obtain the weighted row width of this row range image;

[0028] Based on the weighted row width of each row range image, the distribution range image is divided by rows.

[0029] Further, the features include at least one of the penetration point number feature, the penetration ratio feature, and the penetration range ratio feature; the step of extracting features from the point cloud projected depth image corresponding to each distribution range image according to the preset feature extraction rules includes:

[0030] Judge whether the projection points of the point cloud data located within the distribution range image are penetration points, and count the number of penetration points to obtain the penetration point number feature corresponding to each distribution range image;

[0031] According to the actual projection number and the number of penetration points of the projection points of the point cloud data located within the distribution range image, the penetration ratio feature is obtained;

[0032] Obtain the penetration range proportion feature according to the area of the penetration point in the distribution range image and the number of penetration points.

[0033] In a second aspect, a point cloud-based dust filtering device for a mining area is provided, including:

[0034] A point cloud projection depth image acquisition module, configured to obtain a point cloud projection depth image according to the point cloud data of a lidar in a received mining area;

[0035] A clustering cluster acquisition module, configured to cluster the projection points of each point cloud data located on the point cloud projection depth image to obtain clustering clusters corresponding to each obstacle;

[0036] A feature extraction module, configured to extract the features of each clustering cluster on the point cloud projection depth image;

[0037] A dust removal module, configured to input the features extracted corresponding to each clustering cluster into a trained classification model to obtain the dust probability corresponding to each clustering cluster, and remove the obstacles corresponding to the clustering clusters with the dust probability higher than a preset threshold from the obstacle sequence.

[0038] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned point cloud-based dust filtering method for a mining area are implemented.

[0039] In a fourth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned point cloud-based dust filtering method for a mining area are implemented.

[0040] In the solution implemented by the above-mentioned method, device, computer equipment and storage medium for filtering dust in a mining area based on point cloud, a point cloud projection depth image obtained from the point cloud data of a lidar in the mining area can be used to cluster the projection points of each point cloud data on the point cloud projection depth image to obtain clustering clusters corresponding to each obstacle; then, the features of each clustering cluster on the point cloud projection depth image are extracted; the features extracted for each clustering cluster are input into a trained classification model to obtain the dust probability corresponding to each clustering cluster, and the obstacles corresponding to the clustering clusters with a dust probability higher than a preset threshold are removed from the obstacle sequence. By projecting the lidar point cloud data onto a depth image, the present invention overcomes the difficulty that the point cloud density is affected by distance in the Cartesian coordinate system. After projection, combined with the obstacle clustering information obtained in the Cartesian coordinate system, the features of the clustering clusters on the point cloud projection depth image are extracted on the point cloud distribution projection map, and then the features are input into a trained classification model to obtain the dust probability corresponding to each clustering cluster, so as to remove the obstacles corresponding to the cluster labels of dust from the obstacle sequence according to the obtained dust probability, achieving the effect of filtering dust. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for the description of the embodiments of the present invention. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0042] Figure 1 is a flowchart of a method for filtering dust in a mining area based on point cloud according to an embodiment of the present invention;

[0043] Figure 2 is Figure 1 a schematic flowchart of a specific implementation manner of step S102 in

[0044] Figure 3 is Figure 1 a schematic flowchart of a specific implementation manner of step S103 in

[0045] Figure 4 is Figure 1 a schematic flowchart of a specific implementation manner of step S104 in

[0046] Figure 5 is a schematic structural diagram of a device for filtering dust in a mining area based on point cloud according to an embodiment of the present invention;

[0047] Figure 6 is a schematic structural diagram of a computer device according to an embodiment of the present invention;

[0048] Figure 7 It is another schematic structural diagram of a computer device in an embodiment of the present invention. Detailed implementation manners

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0050] The driverless mining truck detects obstacles through various sensors including lidar, so as to avoid obstacles in time during the driverless process.

[0051] Due to the relatively harsh working environment in the mining area, lidar will generate a large number of false detections in the mining area scene, which are mainly caused by thick dust; thick dust will form dust point clouds on the lidar. Among them, thick dust refers to dust with a dust concentration greater than 10 mg / m³.

[0052] In the prior art, point cloud data is obtained through lidar, the point cloud data is downsampled and its point cloud density is calculated as a reference threshold. By traversing and solving the point cloud density within the neighborhood range of each point and comparing it with the reference threshold, a set of noise points is obtained. After clustering the set of noise points, clusters with fewer points than the reference threshold in the noise point clustering clusters are filtered out as the denoising result. However, this method does not consider the influence of the reflection point distance on the point cloud density obtained by the lidar, and the calculation within the neighborhood space range has problems of high complexity and time consumption, and is suitable for scenarios with fewer noise points and low requirements for real-time performance. In an environment where the dust in the mining area is large and the dust intensity fluctuates due to weather influence, it will lead to insufficient performance and unstable performance.

[0053] Based on this, a method for filtering dust in a mining area based on point cloud provided by an embodiment of the present invention is shown in Figure 1 as follows, and it includes the following steps:

[0054] S101. According to the point cloud data of the lidar in the mining area received, obtain a point cloud projection depth image.

[0055] Multiple laser pulses are emitted outward by the lidar on the driverless mining vehicle until the signal reaches an object, then reflects and returns to the receiver. Based on the known speed of the laser pulse and the reflection time, the distance between the detected object and the lidar receiver can be calculated. The laser emitter of the lidar rotates at high speed inside and can complete one rotation within 0.1 s. After one rotation, the position information of the objects around the vehicle relative to the lidar is obtained, and there is a large amount of point cloud data, which can well depict the environment around the vehicle. Then the lidar transmits the data to the computer through the network cable using the UDP protocol. Then the lidar driver code on the computer parses the transmitted data to obtain the point cloud data of the environment around the vehicle, including three-dimensional coordinates and reflection intensity (x, y, z, intensity).

[0056] The point cloud projection depth image can be obtained from the projection of the point cloud data on the range. And the point cloud data can be represented by (row, col). row is determined by the pitch angle of each point in the point cloud data, and col is determined by the horizontal angle of the point cloud data relative to the horizontal plane. Here, according to the mutual conversion relationship between the Cartesian coordinate system and the laser coordinate system (reflection distance r, pitch angle θ, azimuth angle ψ on the horizontal plane), the pitch angle θ and the horizontal azimuth angle ψ are calculated point by point, and finally the range image projection of the point cloud is obtained, which is the point cloud projection depth image.

[0057] S102. Cluster the projection points of each point cloud data located on the point cloud projection depth image to obtain a cluster corresponding to each obstacle.

[0058] By clustering the projection points of each point cloud data on the point cloud projection depth image, the projection points of the point cloud data can be classified according to each unknown obstacle, and then a cluster corresponding to each obstacle can be obtained, so as to extract the features on the point cloud projection depth image corresponding to each unknown obstacle through step S103.

[0059] See Figure 2 , the clustering of the projection points of each point cloud data located on the point cloud projection depth image to obtain a cluster corresponding to each obstacle may include the following steps:

[0060] S1021. Based on the planar size of the point cloud projection depth image, divide the point cloud projection depth image into multiple grids.

[0061] Here, the spatial coordinates of each point cloud data are represented as (x, y, z, intensity). The grid projection size (Sx, Sy) of the x-y plane is set. Starting from the origin of the x-y plane, the point cloud data is segmented into several grids in the x-y plane, and the size of each grid is (Sx, Sy). Here, each grid contains the point cloud data that falls within the spatial range of the grid.

[0062] S1022. Filter out the ground grids based on the height differences of the projection points of the point cloud data corresponding to each grid and the height differences between adjacent grids, and obtain non-ground grids.

[0063] It should be noted that the ground grids are extracted based on the height differences of the projection points of the point cloud data within each grid and the height differences between several adjacent grids, and then the ground grids are filtered out from the grids segmented from the point cloud projection depth image to obtain non-ground grids.

[0064] S1023. Cluster the non-ground grids according to their relative planar positions to obtain the clustering clusters corresponding to each obstacle, and generate the cluster labels corresponding to each clustering cluster.

[0065] Here, by clustering each non-ground grid according to its relative position relationship in the x-y plane, the clustering clusters corresponding to each obstacle are obtained. At the same time, the cluster labels of each clustering cluster are marked to obtain the clustering clusters with cluster labels in the point cloud projection depth image; for example, clustering the non-ground grids A, B, C, and D according to their planar relative positions, it is determined that the non-ground grids A, B, C, and D can form a clustering cluster corresponding to obstacle A. Since the attributes of obstacle A are unknown here, the cluster label corresponding to this clustering cluster is set as A.

[0066] S103. Extract the features of each clustering cluster on the point cloud projection depth image.

[0067] Here, the features of each clustering cluster on the point cloud projection depth image are extracted through a preset feature extraction rule.

[0068] See Figure 3 The extraction of the features of each clustering cluster on the point cloud projection depth image may include the following steps:

[0069] S1031. Determine the distribution range image of each clustering cluster on the point cloud projection depth image.

[0070] For each clustering cluster, determine its distribution range on the point cloud projection depth image. Assume that the distribution range image of the obstacle to which each clustering cluster belongs generally conforms to a rectangular distribution.

[0071] Extract the distribution range image of each cluster label on the point cloud projected depth image according to the cluster label corresponding to each point cloud data, that is, the range where the projected points of the point cloud data included in each cluster label start from the minimum column and minimum row and end at the maximum column and maximum row on the point cloud projected depth image.

[0072] S1032. Extract features from the point cloud projected depth image corresponding to each of the distribution range images according to a preset feature extraction rule.

[0073] In some embodiments, the features include at least one of a row sparsity feature, a column density non-uniformity feature, and an average density feature. The extracting features from the point cloud projected depth image corresponding to each of the distribution range images according to a preset feature extraction rule may include: dividing each of the distribution range images by rows according to the distribution range image of each clustering cluster on the point cloud projected depth image to obtain a row range image; obtaining the row sparsity feature based on the allowable projection number and the actual projection number of the projected points of the point cloud data in the row range image; obtaining the column density non-uniformity feature based on the variance between the number of the row range images and the actual projection number of the projected points of the point cloud data in the row range image; and obtaining the average density feature based on the distribution range image of each clustering cluster on the point cloud projected depth image and the allowable projection number and the actual projection number of the projected points of the point cloud data in the distribution range image.

[0074] Here, the projection range of each clustering cluster corresponding to each cluster label on the point cloud projected depth image is approximately represented as a rectangular frame. Based on the number of rows involved in the rectangular frame, determine the distribution range image to which each obstacle belongs. Divide each distribution range image by rows according to the distribution range image of the clustering cluster corresponding to each obstacle on the point cloud projected depth image to obtain a row range image; the row sparsity feature can be obtained by calculating the ratio between the allowable actual projection number and the projection number of the projected points of the point cloud data in the row range image; the average density feature can be obtained based on the ratio between the allowable projection number and the actual projection number of the projected points of the point cloud data in the distribution range image of each clustering cluster on the point cloud projected depth image.

[0075] Further, to weaken the influence of ground points within the obstacle range, here, dividing each of the distribution range images of each clustering cluster in the point cloud projection depth image into rows to obtain row range images includes: determining the weighting ratio corresponding to each row range image based on the preset weights of the row range images corresponding to the highest and lowest positions located on the distribution range image, and the number of row range images within the distribution range image; performing weighted calculation on the row width of this row range image according to the weighting ratio corresponding to each row range image to obtain the weighted row width of this row range image; and dividing the distribution range image based on the weighted row width of each row range image. Here, to weaken the influence of ground points within the obstacle range, the row width of the row range image is weighted. By assigning a lower weight to the row width of the row range image closer to the ground, the influence of ground points within the obstacle range can be weakened.

[0076] Further, the features include at least one of the number of penetration points feature, penetration ratio feature, and penetration range ratio feature; the extracting features from the point cloud projection depth image corresponding to each distribution range image according to the preset feature extraction rules includes: determining whether the projection points of each point cloud data located within the distribution range image are penetration points, and counting the number of penetration points to obtain the number of penetration points feature corresponding to each distribution range image; obtaining the penetration ratio feature according to the actual projection number of the projection points of the point cloud data located within the distribution range image and the number of penetration points; and obtaining the penetration range ratio feature according to the area of the penetration points in the distribution range image and the number of penetration points.

[0077] It should be noted that when determining whether the projection points of each point cloud data located within the distribution range image are penetration points, a point whose distance exceeds a preset distance outside the distance range of the projection point of the transport aircraft corresponding to the cluster label is defined as a penetration point.

[0078] S104. Input the features extracted for each clustering cluster into the trained classification model to obtain the dust probability corresponding to each clustering cluster, and remove the obstacles corresponding to the clustering clusters with dust probability higher than the preset threshold from the obstacle sequence.

[0079] Here, based on the features extracted for each clustering cluster, inputting these features into the trained classification model can obtain the dust probability of each clustering cluster. Here, the dust probability obtained by the classification model is used as the basis for filtering obstacles. Removing the obstacles corresponding to the clustering clusters with dust probability higher than 0.5 from the obstacle sequence can achieve the effect of dust filtering.

[0080] SeeFigure 4 Inputting the features extracted for each of the clustering clusters into a trained classification model to obtain the dust probability corresponding to each of the clustering clusters may include the following steps:

[0081] S1041. Analyze the obstacle categories to which each of the clustering clusters belongs, and distinguish and label the cluster labels corresponding to the clustering clusters whose obstacle category is dust and the cluster labels corresponding to the clustering clusters whose obstacle category is non-dust.

[0082] Here, by analyzing the obstacle categories to which each clustering cluster belongs, and for the cluster labels of dust and the other obstacle cluster labels, the labels are respectively distinguished and labeled as 1 and 0.

[0083] S1042. Input the calibration result of the cluster labels into a preset classification model for training to obtain the trained classification model.

[0084] To effectively distinguish and represent the probability of dust and other obstacles, a classifier is constructed using logistic regression for classification. By inputting the calibration result of the cluster labels into a preset classification model for training, a trained classification model is obtained, and then the trained classification model is used in subsequent tests and applications.

[0085] The present invention provides a method for filtering dust in a mining area based on point cloud. Compared with the prior art, the present invention obtains a point cloud projection depth image according to the point cloud data of a lidar in the received mining area, clusters the projection points of each point cloud data located on the point cloud projection depth image to obtain clustering clusters corresponding to each obstacle; then extracts the features of each clustering cluster on the point cloud projection depth image; inputs the features extracted for each clustering cluster into a trained classification model to obtain the dust probability corresponding to each clustering cluster, and removes the obstacles corresponding to the clustering clusters with a dust probability higher than a preset threshold from the obstacle sequence. The present invention overcomes the difficulty that the point cloud density is affected by distance in the Cartesian coordinate system by projecting the lidar point cloud data onto a depth image. After projection, combined with the obstacle clustering information obtained in the Cartesian coordinate system, the features of the clustering clusters on the point cloud projection depth image are extracted on the point cloud distribution projection map, and then the extracted features are input into a trained classification model to obtain the dust probability corresponding to each clustering cluster, so as to remove the obstacles corresponding to the cluster label of dust from the obstacle sequence according to the obtained dust probability, achieving the effect of filtering dust, and further realizing the effect of correcting and filtering misdetected obstacles for the obstacle perception function in a mine environment with more dust.

[0086] It should be understood that the sequence numbers of the steps in the above embodiments do not indicate the order of execution. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0087] In one embodiment, a dust filtering device for a mining area based on point cloud is provided, as Figure 5 shown. The device for obtaining mining area information includes a point cloud projection depth image obtaining module, a clustering cluster obtaining module, a feature extraction module, a dust removing module, and a first execution module. The detailed description of each functional module is as follows:

[0088] The point cloud projection depth image obtaining module 51 is used to obtain a point cloud projection depth image according to the point cloud data of the lidar in the mining area received;

[0089] The clustering cluster obtaining module 52 is used to cluster the projection points of each point cloud data located on the point cloud projection depth image to obtain clustering clusters corresponding to each obstacle;

[0090] The feature extraction module 53 is used to extract the features of each clustering cluster on the point cloud projection depth image;

[0091] The dust removing module 54 is used to input the features extracted corresponding to each clustering cluster into a trained classification model to obtain the dust probability corresponding to each clustering cluster respectively, and remove the obstacles corresponding to the clustering clusters with the dust probability higher than a preset threshold from the obstacle sequence.

[0092] In one embodiment, the clustering cluster obtaining module 52 includes:

[0093] The grid segmentation unit is used to segment the point cloud projection depth image into multiple grids based on the planar size of the point cloud projection depth image;

[0094] The non-ground grid determination unit is used to filter out the ground grids according to the height difference of the projection points of each point cloud data corresponding to each grid and the height difference between two adjacent grids to obtain non-ground grids;

[0095] The clustering cluster obtaining unit is used to cluster the non-ground grids according to their planar relative positions to obtain the clustering clusters corresponding to each obstacle, and generate cluster labels corresponding to each clustering cluster respectively.

[0096] In one embodiment, the dust removing module 54 includes:

[0097] A discrimination and calibration unit, configured to analyze the obstacle categories to which each of the clustering clusters belongs, and discriminate and calibrate the cluster labels corresponding to the clustering clusters whose belonging obstacle category is dust, and the cluster labels corresponding to the clustering clusters whose belonging obstacle category is non-dust;

[0098] A classification model training unit, configured to input the calibration result of the cluster label into a preset classification model for training to obtain the trained classification model.

[0099] In one embodiment, the feature extraction module 53 includes:

[0100] A distribution range image determination unit, configured to determine the distribution range image of each of the clustering clusters on the point cloud projection depth image;

[0101] A feature extraction unit, configured to extract features from the point cloud projection depth image corresponding to each of the distribution range images according to a preset feature extraction rule.

[0102] In one embodiment, the features include at least one of a row sparsity feature, a column density non-uniformity feature, and an average density feature, and the feature extraction unit includes:

[0103] A row range image acquisition subunit, configured to divide each of the distribution range images into rows according to the distribution range image of each of the clustering clusters on the point cloud projection depth image to obtain a row range image;

[0104] A row sparsity feature acquisition subunit, configured to obtain the row sparsity feature based on the allowable projection number and the actual projection number of the projection points of the point cloud data in the row range image;

[0105] A column density non-uniformity feature acquisition subunit, configured to obtain the column density non-uniformity feature based on the variance between the number of the row range images and the actual projection number of the projection points of the point cloud data in the row range images;

[0106] An average density feature acquisition subunit, configured to obtain the average density feature based on the distribution range image of each of the clustering clusters on the point cloud projection depth image, and the allowable projection number and the actual projection number of the projection points of the point cloud data in the distribution range image.

[0107] In one embodiment, the row range image acquisition subunit includes:

[0108] A weighted ratio determination subunit, configured to determine a weighted ratio corresponding to each of the row range images based on preset weights of the row range images corresponding to the highest position and the lowest position located on the distribution range image, and the number of row range images within the distribution range image;

[0109] A weighted row width determination subunit, configured to perform weighted calculation on the row width of each of the row range images according to the weighted ratio corresponding to each of the row range images, to obtain the weighted row width of each of the row range images;

[0110] A row division subunit, configured to perform row division on the distribution range image based on the weighted row width of each of the row range images.

[0111] In an embodiment, the feature includes at least one of a penetration point number feature, a penetration ratio feature, and a penetration range ratio feature; the feature extraction unit includes:

[0112] A penetration point number feature acquisition subunit, configured to determine whether the projection points of the respective point cloud data located within the distribution range image are penetration points, and count the number of penetration points, to obtain the penetration point number feature corresponding to each of the distribution range images;

[0113] A penetration point number feature acquisition subunit, configured to obtain the penetration ratio feature according to the actual projection number of the projection points of the point cloud data located within the distribution range image and the number of penetration points;

[0114] A penetration range ratio feature acquisition subunit, configured to obtain the penetration range ratio feature according to the area of the penetration points in the distribution range image and the number of penetration points.

[0115] The present invention provides a dust filtering device for a mining area based on point cloud. Compared with the prior art, the present invention obtains a point cloud projection depth image from the point cloud data of a lidar in the mining area, clusters the projection points of each point cloud data on the point cloud projection depth image to obtain a clustering cluster corresponding to each obstacle; then extracts the features of each clustering cluster on the point cloud projection depth image; inputs the features extracted corresponding to each clustering cluster into a trained classification model to obtain the dust probability corresponding to each clustering cluster respectively, and removes the obstacles corresponding to the clustering clusters with a dust probability higher than a preset threshold from the obstacle sequence. The present invention overcomes the difficulty that the point cloud density is affected by distance in the Cartesian coordinate system by projecting the lidar point cloud data onto a depth image. After projection, combined with the obstacle clustering information obtained in the Cartesian coordinate system, the features of the clustering clusters on the point cloud projection depth image are extracted on the point cloud distribution projection map, and then the features are input into a trained classification model to obtain the dust probability corresponding to each clustering cluster respectively, so as to remove the obstacles corresponding to the cluster labels as dust from the obstacle sequence according to the obtained dust probability, achieving the effect of filtering dust.

[0116] For the specific limitations of the dust filtering device for the mining area based on point cloud, reference can be made to the limitations of the dust filtering method for the mining area based on point cloud in the above text, which will not be elaborated here. Each module in the above dust filtering device for the mining area based on point cloud can be implemented in whole or in part by software, hardware and their combination. The above modules can be embedded in the processor in the computer device in hardware form or be independent of it, or be stored in the memory in the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0117] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 6 shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the server side of a dust filtering method for a mining area based on point cloud.

[0118] In one embodiment, a computer device is provided. The computer device can be a client, and its internal structure diagram can be as Figure 7As shown in the figure. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server through a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the client side of a method for filtering dust in a mining area based on point cloud

[0119] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the following steps are realized:

[0120] According to the received point cloud data of the lidar in the mining area, obtain a point cloud projection depth image;

[0121] Cluster the projection points of each point cloud data located on the point cloud projection depth image to obtain a cluster corresponding to each obstacle;

[0122] Extract the features of each cluster on the point cloud projection depth image;

[0123] Input the features extracted corresponding to each cluster into a trained classification model to obtain the dust probability corresponding to each cluster respectively, and remove the obstacles corresponding to the clusters with the dust probability higher than a preset threshold from the obstacle sequence.

[0124] In one embodiment, another computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the following steps are realized:

[0125] According to the received point cloud data of the lidar in the mining area, obtain a point cloud projection depth image;

[0126] Cluster the projection points of each point cloud data located on the point cloud projection depth image to obtain a cluster corresponding to each obstacle;

[0127] Extract the features of each cluster on the point cloud projection depth image;

[0128] Input the features extracted corresponding to each cluster into a trained classification model to obtain the dust probability corresponding to each cluster respectively, and remove the obstacles corresponding to the clusters with the dust probability higher than a preset threshold from the obstacle sequence.

[0129] It should be noted that for the functions or steps that can be achieved by the above computer-readable storage medium or computer device, reference can be made to the relevant descriptions on the server side and the client side in the foregoing method embodiments. To avoid repetition, they will not be described in detail here.

[0130] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above various methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0131] Those skilled in the art can clearly understand that for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0132] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for filtering dust in a mining area based on point cloud, characterized in that, Including: Obtaining a point cloud projection depth image based on the received point cloud data of a lidar in a mining area; Clustering the projection points of each point cloud data located on the point cloud projection depth image to obtain clustering clusters corresponding to each obstacle; Extracting the features of each clustering cluster on the point cloud projection depth image, including: determining the distribution range image of each clustering cluster on the point cloud projection depth image, dividing each distribution range image by rows according to the distribution range image of each clustering cluster on the point cloud projection depth image to obtain a row range image, obtaining a row sparsity feature based on the allowable projection number and the actual projection number of the projection points of the point cloud data in the row range image, obtaining a column density non-uniformity feature based on the variance between the number of the row range images and the actual projection number of the projection points of the point cloud data in the row range image, obtaining an average density feature based on the distribution range image of each clustering cluster on the point cloud projection depth image and the allowable projection number and the actual projection number of the projection points of the point cloud data in the distribution range image, and the features include at least one of the row sparsity feature, the column density non-uniformity feature, and the average density feature; Inputting the features extracted corresponding to each clustering cluster into a trained classification model to obtain the dust probability corresponding to each clustering cluster respectively, and removing the obstacles corresponding to the clustering clusters with the dust probability higher than a preset threshold from the obstacle sequence.

2. The method according to claim 1, characterized in that, The clustering the projection points of each point cloud data located on the point cloud projection depth image to obtain clustering clusters corresponding to each obstacle includes: Based on the planar size of the point cloud projection depth image, dividing the point cloud projection depth image into multiple grids; Filtering out the ground grids according to the height difference of the projection points of each point cloud data corresponding to each grid and the height difference between two adjacent grids to obtain non-ground grids; Clustering the non-ground grids according to their relative positions in the plane to obtain clustering clusters corresponding to each obstacle, and generating cluster labels corresponding to each clustering cluster respectively.

3. The method according to claim 2, characterized in that, The inputting the features extracted corresponding to each clustering cluster into a trained classification model to obtain the dust probability corresponding to each clustering cluster respectively includes: Analyzing the obstacle categories to which each clustering cluster belongs, and distinguishing and calibrating the cluster labels corresponding to the clustering clusters whose belonging obstacle categories are dust and the cluster labels corresponding to the clustering clusters whose belonging obstacle categories are non-dust; Inputting the calibration result of the cluster labels into a preset classification model for training to obtain the trained classification model.

4. The method according to claim 1, characterized in that, The dividing each distribution range image by rows according to the distribution range image of each clustering cluster on the point cloud projection depth image to obtain a row range image includes: Determine the weighted ratio corresponding to each row range image based on the preset weights of the row range images corresponding to the highest position and the lowest position respectively located on the distribution range image, and the number of row range images within the distribution range image; Perform weighted calculation on the row width of this row range image according to the weighted ratio corresponding to each row range image to obtain the weighted row width of this row range image; Based on the weighted row width of each row range image, perform row division on the distribution range image to obtain each row range image.

5. The method according to claim 1, characterized in that, The features include at least one of the number of penetration points feature, penetration ratio feature, and penetration range ratio feature; extracting features from the point cloud projection depth image corresponding to each distribution range image according to the preset feature extraction rules includes: Judge whether the projection points of each point cloud data located within the distribution range image are penetration points, and count the number of penetration points to obtain the number of penetration points feature corresponding to each distribution range image; Obtain the penetration ratio feature according to the actual projection number and the number of penetration points of the projection points of the point cloud data located within the distribution range image; Obtain the penetration range ratio feature according to the area of the penetration points in the distribution range image and the number of penetration points.

6. A device for filtering dust in a mining area based on point cloud, characterized in that, Include: A point cloud projection depth image acquisition module, configured to obtain a point cloud projection depth image according to the received point cloud data of the lidar in the mining area; A clustering cluster acquisition module, configured to cluster the projection points of each point cloud data located on the point cloud projection depth image to obtain clustering clusters corresponding to each obstacle; A feature extraction module, configured to extract the features of each clustering cluster on the point cloud projection depth image, including: determining the distribution range image of each clustering cluster on the point cloud projection depth image, and performing row division on each distribution range image according to the distribution range image of each clustering cluster on the point cloud projection depth image to obtain row range images, obtaining the row sparsity feature based on the allowable projection number and the actual projection number of the projection points of the point cloud data in the row range image, obtaining the column density non-uniformity feature based on the variance between the number of row range images and the actual projection number of the projection points of the point cloud data in the row range image, and obtaining the average density feature based on the distribution range image of each clustering cluster on the point cloud projection depth image and the allowable projection number and the actual projection number of the projection points of the point cloud data in the distribution range image, and the features include at least one of the row sparsity feature, the column density non-uniformity feature, and the average density feature; A dust removal module, configured to input the features extracted corresponding to each clustering cluster into the trained classification model to obtain the dust probability corresponding to each clustering cluster respectively, and remove the obstacles corresponding to the clustering clusters with the dust probability higher than the preset threshold from the obstacle sequence.

7. A computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, the steps of the point cloud-based dust filtering method for mining areas described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, the steps of the point cloud-based dust filtering method for mining areas described in any one of claims 1 to 5 are implemented.

Images