A complex internal structure train carriage frozen coal unloading automatic obstacle avoidance path planning system

Abstract

The present application relates to the technical field of path planning of transportation devices, and more particularly to an automatic obstacle avoidance path planning system for a train compartment with complex internal structure and frozen coal unloading. The system comprises: a data acquisition module for acquiring frozen coal point cloud data inside the compartment; a target neighborhood radius module for calculating the target neighborhood radius of the frozen coal point cloud data according to the local density difference between each neighborhood radius and its adjacent neighborhood radius; a target minimum point number module for obtaining the initial minimum point number of the frozen coal point cloud data; adjusting the initial minimum point number of the frozen coal point cloud data according to the relationship between the overlap degree and the set threshold to obtain the target minimum point number; and a path planning module for clustering the target cleaning area based on the target neighborhood radius and the target minimum point number of the frozen coal point cloud data, and controlling the cleaning robot to perform cleaning work along the cleaning path, thereby effectively improving the accuracy of the obstacle avoidance path planning.

Description

Technical Field

[0001] This invention relates to the field of path planning technology for transportation devices. More specifically, this invention relates to an automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure. Background Technology

[0002] In cold regions, C80 open coal wagons used for transporting coal often have a large amount of frozen coal remaining inside the wagon after unloading due to the coal freezing with moisture. To achieve efficient cleaning, a cleaning robot mounted on a tipper is typically used for high-pressure water jet cleaning. To ensure that the cleaning robot can complete the task comprehensively, efficiently, and safely, a working path needs to be planned for it that can accurately locate all frozen coal and effectively avoid obstacles such as support beams inside the wagon.

[0003] Currently, the mainstream technical approach to achieving this goal mainly involves acquiring point cloud data of the interior of the train carriage using equipment such as LiDAR, then employing clustering algorithms to identify regions representing frozen coal from the complex point cloud, and finally performing path planning based on the identification results. Among numerous clustering algorithms, the density-based DBSCAN algorithm is widely used because it can discover clusters of arbitrary shapes and does not require pre-specifying the number of clusters.

[0004] However, due to the uneven thickness and varying attachment patterns of the residual frozen coal inside the carriage, the corresponding point cloud data exhibits a distribution characteristic with extremely large density differences. If a fixed set of neighborhood radii and minimum number of points is used, when processing high-density areas such as the center of frozen coal, a complete piece of frozen coal may be incorrectly divided into multiple small clusters, resulting in the planning of redundant cleaning paths; while when processing low-density areas such as the edge of frozen coal, the number of points may not meet the density requirements and thus fail to identify them as valid clusters, causing some frozen coal to be missed.

[0005] Therefore, how to overcome the limitations of the traditional DBSCAN algorithm in processing multi-density frozen coal point clouds and achieve accurate identification of all frozen coal areas is a key problem that current technology urgently needs to solve. Summary of the Invention

[0006] To address the limitations of the traditional DBSCAN algorithm in processing multi-density frozen coal point clouds and to achieve accurate identification of all frozen coal areas, this invention proposes an automatic obstacle avoidance path planning system for unloading frozen coal from train carriages with complex internal structures. The system includes the following modules:

[0007] Data Acquisition Module: Used to acquire frozen coal point cloud data inside the car; Target Neighborhood Radius Module: Used to increase the neighborhood radius along a set radius range, taking the number of frozen coal point cloud data contained under each neighborhood radius as the local density; Determining the neighborhood radius at the point where the local density change shows a decreasing inflection point based on the local density difference between each neighborhood radius and its adjacent neighborhood radii, as the target neighborhood radius for that frozen coal point cloud data; Target Minimum Point Count Module: Used to acquire the initial minimum number of points in the frozen coal point cloud data, the initial minimum number of points being positively correlated with the global average point density and the neighborhood volume under the target neighborhood radius; [The module then further details the module's functions and their corresponding parameters.] All frozen coal point cloud data are clustered using the minimum and maximum initial minimum point counts to obtain a first cluster region and a second cluster region. The overlap between the first and second cluster regions of the frozen coal point cloud data is calculated, and the initial minimum point count of the frozen coal point cloud data is adjusted according to the relationship between the overlap and a set threshold to obtain a target minimum point count. The path planning module is used to cluster the frozen coal point cloud data based on the target neighborhood radius and the target minimum point count to obtain a target cleaning area. Based on the target cleaning area, the cleaning path of the cleaning robot is planned, and the cleaning robot is controlled to perform cleaning operations along the cleaning path.

[0008] This invention improves the accuracy of frozen coal cleaning path planning by planning the obstacle avoidance and cleaning path of the cleaning robot during the unloading of frozen coal from train carriages, thereby increasing the efficiency of frozen coal unloading. During path planning, this invention adaptively determines the neighborhood radius and minimum number of points for each frozen coal point cloud data. This allows the clustering algorithm to adapt to both high-density thick frozen coal areas and low-density thin frozen coal edge areas, reducing the possibility of over-segmentation or under-segmentation when processing frozen coal point cloud data with uneven density. This effectively improves the accuracy and completeness of identifying all forms of frozen coal in the carriage, avoiding redundancy or omissions in the cleaning path due to identification errors. In obtaining the minimum number of points, this invention dynamically adjusts the minimum number of points based on the overlap of two different clustering results. This precise adjustment of the minimum number of points ensures accurate matching between the minimum number of points and the neighborhood radius, thereby effectively improving the accuracy of the final clustering identification result.

[0009] According to the present invention, an automatic obstacle avoidance path planning system for unloading frozen coal in a train carriage with a complex internal structure is provided. The step of acquiring frozen coal point cloud data inside the carriage includes: collecting raw point cloud data inside the carriage, and preprocessing to remove background point cloud data from the raw point cloud data to obtain frozen coal point cloud data of the frozen coal area in the carriage.

[0010] This invention takes into account that background point cloud data in the original point cloud data can interfere with the accuracy of clustering and identification of subsequent frozen coal point cloud data. Therefore, this invention preprocesses and filters out background point cloud data to prepare a foundation for subsequent frozen coal area identification.

[0011] According to the present invention, an automatic obstacle avoidance path planning system for unloading frozen coal in a train carriage with a complex internal structure is provided. The step of increasing the neighborhood radius along a set radius range, and using the number of frozen coal point cloud data contained in each neighborhood radius as the local density, includes: a preset radius range and a neighborhood radius growth step size; obtaining a sequence of neighborhood radii increasing from left to right based on the growth step size within the radius range; obtaining the maximum Z-axis value in all original point cloud data, using the maximum Z-axis value as the height of the cylindrical neighborhood corresponding to each neighborhood radius of the frozen coal point cloud data, and using each neighborhood radius of the frozen coal point cloud data as the radius of the cylindrical neighborhood; and using the number of frozen coal point cloud data contained in the cylindrical neighborhood under each neighborhood radius as the local density of the frozen coal point cloud data under each neighborhood radius.

[0012] According to the present invention, an automatic obstacle avoidance path planning system for unloading frozen coal in a train carriage with a complex internal structure is provided. The step of determining the neighborhood radius at the point where the local density change shows a downward inflection point based on the local density difference between each neighborhood radius and its adjacent neighborhood radii, and using this as the target neighborhood radius of the frozen coal point cloud data, includes: obtaining adjacent neighborhood radii on both sides of the neighborhood radius of the frozen coal point cloud data based on a three-fold growth step size; obtaining the density change rate of the neighborhood radius based on the normalized value of the local density difference between the left and right adjacent neighborhood radii; and using the first downward inflection point that satisfies the condition, where the density change rate of the current neighborhood radius is less than 0.5 times the density change rate of the previous neighborhood radius and is obtained by traversing from the smallest neighborhood radius, as the target neighborhood radius of the frozen coal point cloud data.

[0013] This invention takes into account that the density within a frozen coal region changes with the increase of its neighborhood radius. When the neighborhood radius exceeds the thickness, the density exhibits a significant inflection point. Therefore, this invention obtains the inflection point of density decrease during the process of increasing the neighborhood radius of each frozen coal point cloud data, and uses this as the optimal neighborhood radius for that frozen coal point cloud data. This allows for accurate capture of density changes, reducing the possibility of undersegmentation or oversegmentation.

[0014] According to the present invention, an automatic obstacle avoidance path planning system for unloading frozen coal in a train carriage with a complex internal structure is provided. The method for obtaining the initial minimum number of points in the frozen coal point cloud data includes: obtaining the global average point density based on the local density of the target neighborhood radius of all frozen coal point cloud data; and rounding up the product of the neighborhood volume under the target neighborhood radius of the frozen coal point cloud data and the global average point density to obtain the initial minimum number of points in the frozen coal point cloud data.

[0015] According to the present invention, an automatic obstacle avoidance path planning system for unloading frozen coal in a train carriage with a complex internal structure is provided. The calculation of the overlap between the first cluster region and the second cluster region of the frozen coal point cloud data includes: recording the number of frozen coal point cloud data that simultaneously belong to both the first and second cluster regions as a first index of the frozen coal point cloud data; recording the total number of frozen coal point cloud data in the first and second cluster regions as a second index of the frozen coal point cloud data; and recording the ratio of the first index to the second index as the overlap of the frozen coal point cloud data.

[0016] This invention uses the minimum initial minimum number of points and the maximum initial minimum number of points to perform clustering to obtain a first cluster region and a second cluster region. Based on the overlap, the structural differences between the frozen coal point cloud data in the first cluster region and the second cluster region can be accurately determined.

[0017] According to the present invention, an automatic obstacle avoidance path planning system for unloading frozen coal in a train carriage with a complex internal structure is provided. The step of adjusting the initial minimum number of points in the frozen coal point cloud data to obtain a target minimum number of points based on the relationship between overlap and a set threshold includes: if the overlap of the frozen coal point cloud data is greater than the set threshold, multiplying the initial minimum number of points in the frozen coal point cloud data by a scaling factor to obtain the target minimum number of points, where the scaling factor is greater than 1; if the overlap of the frozen coal point cloud data is less than the set threshold, multiplying the initial minimum number of points in the frozen coal point cloud data by a scaling factor to obtain the target minimum number of points, where the scaling factor is less than 1; if the overlap of the frozen coal point cloud data is equal to the set threshold, then using the initial minimum number of points in the frozen coal point cloud data as the target minimum number of points.

[0018] This invention takes into account that when the overlap is high, it indicates that the clustering result is relatively stable and may be a dense core region. In this case, the initial minimum number of points can be increased to improve the clustering standard and prevent a complete region from being mistakenly divided into multiple small blocks. When the overlap is low, it indicates that the clustering result is unstable and may be a sparse edge region. In this case, the initial minimum number of points can be reduced to lower the clustering threshold and ensure that such regions are not missed, thereby improving the accuracy of the clustering result.

[0019] According to the present invention, an automatic obstacle avoidance path planning system for unloading frozen coal in a train carriage with a complex internal structure is provided. The method for clustering frozen coal point cloud data based on the target neighborhood radius and the target minimum number of points to obtain a target cleaning region includes: for a frozen coal point cloud data, if the number of frozen coal point cloud data in the target neighborhood radius is greater than or equal to the target minimum number of points, then the frozen coal point cloud data is marked as a core point; if the number of frozen coal point cloud data in the target neighborhood radius is less than the target minimum number of points, and the frozen coal point cloud data is located within the target neighborhood radius of other frozen coal point cloud data, then the frozen coal point cloud data is marked as a boundary point; connecting adjacent core points to form clusters, and boundary points are assigned to the clusters of their respective core points to obtain the target cleaning region.

[0020] According to the present invention, an automatic obstacle avoidance path planning system for unloading frozen coal in a train carriage with a complex internal structure is provided. The method for planning the cleaning path of the cleaning robot based on the target cleaning area includes: obtaining the projected area of ​​all frozen coal point cloud data in the target cleaning area on the horizontal plane, and using each projected area as the size of the coal block corresponding to the target cleaning area; setting a cleaning priority for each target cleaning area according to the size of the coal block, and sequentially accessing each target cleaning area based on the cleaning priority to obtain the cleaning path of the cleaning robot.

[0021] According to the present invention, an automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure is provided. The system controls a cleaning robot to perform cleaning operations along a cleaning path, including: converting the cleaning path into a sequence of execution instructions for the cleaning robot, and controlling the cleaning robotic arm to carry a high-pressure nozzle to sequentially complete the cleaning operations of each target cleaning area.

[0022] The present invention has the following technical effects:

[0023] Based on the above technical solution, this invention provides an automatic obstacle avoidance path planning system for unloading frozen coal from train carriages with complex internal structures. By planning the obstacle avoidance and cleaning path for the cleaning robot during the unloading of frozen coal from the train carriages, the accuracy of the frozen coal cleaning path planning can be effectively improved, thereby increasing the efficiency of unloading frozen coal from the train carriages. During the path planning process, this invention adaptively determines the neighborhood radius and minimum number of points for each frozen coal point cloud data, enabling the clustering algorithm to simultaneously adapt to high-density thick frozen coal areas and low-density thin frozen coal edge areas. This reduces the possibility of over-segmentation or under-segmentation when processing frozen coal point cloud data with uneven density, effectively improving the accuracy and completeness of identifying all forms of frozen coal within the carriage, and avoiding redundancy or omissions in the cleaning path due to identification errors. In obtaining the minimum number of points, this invention dynamically adjusts the minimum number of points based on the overlap of two different clustering results, allowing for precise adjustment of the minimum number of points so that the minimum number of points and the neighborhood radius can be accurately matched, thereby effectively improving the accuracy of the final clustering identification result. Attached Figure Description

[0024] Figure 1 This is a system block diagram of an automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure, according to an embodiment of the present invention. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0026] This invention provides an automatic obstacle avoidance path planning system for unloading frozen coal in train carriages with complex internal structures. Through an adaptive neighborhood radius determination method and a dynamic minimum point adjustment mechanism, it can accurately identify frozen coal piles of different thicknesses and densities, effectively improving the efficiency of cleaning operations.

[0027] like Figure 1 As shown, Figure 1 This is a system block diagram of an automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure, according to an embodiment of the present invention. Specifically, it includes modules 101-104, which are described in detail below.

[0028] The data acquisition module 101 is used to acquire point cloud data of frozen coal inside the car.

[0029] For example, in an embodiment of the present invention, obtaining frozen coal point cloud data inside the carriage includes: collecting raw point cloud data inside the carriage, and preprocessing to remove background point cloud data from the raw point cloud data to obtain frozen coal point cloud data of the frozen coal area inside the carriage.

[0030] For example, a 50 Hz LiDAR scanner can be used to acquire raw point cloud data of the interior of the carriage. The raw point cloud data includes effective frozen coal point cloud data of different thicknesses of frozen coal, background point cloud data of the carriage's struts, walls, and bottom, and a large amount of sensor noise.

[0031] Directly using raw point cloud data for analysis may lead to noise points being misidentified as frozen coal edges, resulting in oversegmentation and interfering with subsequent identification and segmentation of frozen coal deposits. Furthermore, residual background structures can increase the risk of collisions during path planning. Therefore, preprocessing of the raw point cloud data is necessary to extract valid frozen coal point cloud data, thereby reducing the impact of noise and invalid background data on the final segmentation results.

[0032] For example, each raw point cloud data point corresponds to a three-dimensional coordinate system including the X, Y, and Z axes and the reflection intensity. A statistical filtering algorithm is used to remove outlier noise points from the acquired raw point cloud data. The RANSAC algorithm is used to fit a planar model representing the carriage floor from the raw point cloud data and separate it into the background. The RANSAC algorithm is also used to fit a plane perpendicular to the ground plane normal vector, corresponding to the carriage sidewalls, and separate them into the background. A region-growing-based segmentation method is used to identify and remove structures such as struts from the raw point cloud data.

[0033] Specifically, when using a statistical filtering algorithm to remove outlier noise points from the collected raw point cloud data, the mean Euclidean distance and standard deviation of reflection intensity between the raw point cloud data and its 10 neighboring raw point cloud data can be obtained. If the mean Euclidean distance of the raw point cloud data is greater than 3 times the standard deviation of reflection intensity, then the raw point cloud data is removed as an outlier noise point.

[0034] The scanning frequency, filtering method, and number of adjacent original point cloud data can be set according to actual needs, and the embodiments of the present invention do not impose too many restrictions here.

[0035] By preprocessing the raw point cloud data according to the above steps, the frozen coal point cloud data of the frozen coal area inside the car can be accurately extracted. In order to set appropriate neighborhood radii and minimum number of points for frozen coal areas of different thicknesses, this embodiment of the invention analyzes the local density changes during the neighborhood radius growth process, obtains the neighborhood radius corresponding to the significant change in local density as the target neighborhood radius of the frozen coal point cloud data, and adjusts the minimum number of points based on this, which can accurately achieve clustering and identification of frozen coal of different thicknesses, i.e., the following steps are performed.

[0036] The target neighborhood radius module 102 is used to increase the neighborhood radius along a set radius range, and to take the number of frozen coal point cloud data contained in each neighborhood radius as the local density; and to determine the neighborhood radius at the point where the local density change decreases based on the local density difference between each neighborhood radius and its adjacent neighborhood radii, and to take it as the target neighborhood radius of the frozen coal point cloud data.

[0037] It should be noted that, since frozen coal point cloud data comprises frozen coal regions of varying thicknesses, the local density distribution characteristics of the frozen coal point cloud data within each region differ. For thinner frozen coal, the overall thickness range is relatively small, and a small increase in the neighborhood radius will lead to a significant increase in the frozen coal point cloud data within the inner neighborhood, thereby increasing the local density change rate. When the neighborhood radius is continuously increased until it exceeds the thickness of the frozen coal, the density will rapidly reach saturation, and the local density change rate will drop sharply to a very low level, forming a clear inflection point. This inflection point usually corresponds to a relatively small neighborhood radius.

[0038] For thicker frozen coal, the overall thickness range is large. During a long period of increasing neighborhood radius, the frozen coal point cloud data in its neighborhood will continue to increase significantly, and the local density change rate can remain at a high level within a large radius range. Only when the neighborhood radius is large enough to cover the entire frozen coal block will the density tend to saturate, and the local density change rate will then show a decreasing inflection point, which usually corresponds to a large neighborhood radius.

[0039] Based on this, embodiments of the present invention can obtain an inflection point with a large rate of density change based on the density difference between adjacent neighborhood radii during the increase of the neighborhood radius of each frozen coal point cloud data, and take this inflection point as the optimal target neighborhood radius of the frozen coal point cloud data.

[0040] For example, in an embodiment of the present invention, increasing the neighborhood radius along a set radius range, and using the number of frozen coal point cloud data contained in each neighborhood radius as the local density, includes: a preset radius range and a neighborhood radius growth step size; obtaining a neighborhood radius sequence that increases from left to right based on the growth step size within the radius range; obtaining the maximum Z-axis value in all original point cloud data, using the maximum Z-axis value as the height of the cylindrical neighborhood corresponding to each neighborhood radius of the frozen coal point cloud data, and using each neighborhood radius of the frozen coal point cloud data as the radius of the cylindrical neighborhood; and using the number of frozen coal point cloud data contained in the cylindrical neighborhood under each neighborhood radius as the local density of the frozen coal point cloud data under each neighborhood radius.

[0041] Among them, the radius range is set. At that time, the lower limit of the radius range can be set. Set to 1 meter, upper limit of radius range The radius is set to 10 meters, and the growth step size of the neighborhood radius is set to 0.2 meters, meaning that the difference between adjacent neighborhood radii in the obtained neighborhood radius sequence is 0.2 meters; the radius range and growth step size can be set according to actual needs.

[0042] Understandably, for thin-walled frozen coal, when its neighborhood radius is small, the number of point clouds increases rapidly with the increase of the radius, and the density change rate is high. When the radius exceeds the thickness of the frozen coal, the number of point clouds tends to saturate, and the density change rate drops sharply, forming a relatively steep inflection point. For thick-walled frozen coal, when the neighborhood radius increases over a large range, the number of point clouds continues to increase, and the density change rate remains high. Only when the radius is sufficient to enclose the entire frozen coal block does the density tend to saturate, and the change rate also decreases, forming a relatively gentle inflection point.

[0043] Therefore, identifying inflection points solely based on the local density change rate between adjacent neighborhood radii may fail to accurately identify them due to the slow density decrease before the inflection point in thick frozen coal. To improve the accuracy of inflection point identification, embodiments of the present invention can analyze local density changes by increasing the step size by three times, thereby accurately capturing the turning point of density change.

[0044] For example, in an embodiment of the present invention, determining the neighborhood radius at the point where the local density change shows a decreasing inflection point based on the local density difference between each neighborhood radius and its adjacent neighborhood radii, and using this as the target neighborhood radius of the frozen coal point cloud data, includes: obtaining adjacent neighborhood radii on both sides of the neighborhood radius of the frozen coal point cloud data based on a three-fold growth step size; obtaining the density change rate of the neighborhood radius based on the normalized value of the local density difference between the left and right adjacent neighborhood radii; and using the current neighborhood radius whose density change rate is less than 0.5 times the density change rate of the previous neighborhood radius, and which is the first decreasing inflection point obtained by traversing from the smallest neighborhood radius, as the target neighborhood radius of the frozen coal point cloud data.

[0045] It is understandable that when collecting frozen coal point cloud data, some frozen coal areas may be obstructed by support rods or other obstructions in the carriage. Therefore, the first descent inflection point obtained based on the above steps, satisfying this condition, may be a point inside the frozen coal and cannot reflect changes in thickness. Therefore, this embodiment of the invention can further increase the neighborhood radius after obtaining the first descent inflection point that satisfies this condition to determine whether a second, more significant inflection point appears. That is, if the density change rate of the current neighborhood radius is less than 0.6 or 0.7 times the density change rate of the previous neighborhood radius, then this second inflection point is the target neighborhood radius of the frozen coal point cloud data.

[0046] The specific factor for the change rate of neighborhood radius density can be set according to actual needs.

[0047] Specifically, when obtaining the target neighborhood radius of the frozen coal point cloud data based on the descending inflection point, the inflection point is identified where the density change rate of the current neighborhood radius and the density change rate of the previous neighborhood radius show a clear shift. Therefore, when obtaining the target neighborhood radius of the frozen coal point cloud data, the average of the current neighborhood radius and the previous neighborhood radius corresponding to the inflection point can be rounded up as the target neighborhood radius of the frozen coal point cloud data. If no neighborhood radius matching the descending inflection point is obtained after traversing the neighborhood radius sequence, then the radius range can be... Assign the target neighborhood radius to the current frozen coal point cloud data.

[0048] The target neighborhood radius of all frozen coal point cloud data can be obtained by following the steps above.

[0049] The target minimum point count module 103 is used to obtain the initial minimum point count of frozen coal point cloud data. The initial minimum point count is positively correlated with the global average point density and the neighborhood volume under the target neighborhood radius. All frozen coal point cloud data are clustered using the minimum and maximum initial minimum point counts respectively to obtain the first cluster region and the second cluster region. The overlap between the first and second cluster regions of the frozen coal point cloud data is calculated, and the initial minimum point count of the frozen coal point cloud data is adjusted according to the relationship between the overlap and a set threshold to obtain the target minimum point count.

[0050] It should be noted that in the DBSCAN algorithm, the minimum number of points and the neighborhood radius together define the core attributes of a frozen coal point cloud data set. The minimum number of points should be proportional to the expected number of points within the space defined by the neighborhood radius. For frozen coal blocks of different sizes, the number of frozen coal point cloud data points per unit volume is relatively stable. Therefore, the larger the volume within the cylindrical neighborhood space defined by the neighborhood radius of a frozen coal point cloud data set, the more frozen coal point cloud data points need to be included to be identified as core points, and then clusters are formed by connecting these core points.

[0051] Based on this, embodiments of the present invention can calculate the initial minimum number of points by using the global average point density and the cylindrical neighborhood volume corresponding to the target neighborhood radius, thereby achieving dynamic parameter adaptation. For thin frozen coal, the corresponding cylindrical neighborhood volume is small, and the calculated initial minimum number of points is less, which can lower the threshold for determining core points and ensure that the point cloud of thin frozen coal can form effective clustering. For thick frozen coal, the neighborhood volume is large, and the calculated initial minimum number of points is more, which raises the standard for determining core points and avoids over-segmentation caused by excessively low point number requirements.

[0052] For example, in an embodiment of the present invention, obtaining the initial minimum number of points for each frozen coal point cloud data includes: obtaining the global average point density based on the local density of the target neighborhood radius of all frozen coal point cloud data; and rounding up the product of the neighborhood volume under the target neighborhood radius of the frozen coal point cloud data and the global average point density to obtain the initial minimum number of points for the frozen coal point cloud data.

[0053] Specifically, when obtaining the global average point density based on the local density of the target neighborhood radius of all frozen coal point cloud data, the maximum and minimum values ​​of the cylindrical neighborhood volume corresponding to the target neighborhood radius of all frozen coal point cloud data can be removed, and the average local density in the cylindrical neighborhood corresponding to the target neighborhood radius of the remaining frozen coal point cloud data can be used as the global average point density.

[0054] Wherein, if the target neighborhood radius of the frozen coal point cloud data is Then the area of ​​the base circle of the corresponding cylindrical neighborhood is The height of the cylindrical neighborhood is the maximum value of the Z-axis in all the original point cloud data. The final obtained neighborhood volume of the frozen coal point cloud data under the target neighborhood radius is .

[0055] Understandably, if the initial minimum number of points in the frozen coal point cloud data is too small, the algorithm may become overly sensitive. Therefore, a minimum lower limit can be set for the initial minimum number of points in the frozen coal point cloud data. If the initial minimum number of points in the frozen coal point cloud data is less than this minimum lower limit, then this minimum lower limit will be used as the initial minimum number of points in the frozen coal point cloud data.

[0056] The minimum lower limit can be set to 3; the minimum lower limit can be set according to actual needs.

[0057] The above steps assume that the local point density of all frozen coal blocks is similar to the global average density, thus obtaining the initial minimum number of points for all frozen coal point cloud data. However, differences in adhesion, ice crystal structure, and moisture content among different frozen coal blocks result in non-uniform point cloud densities, necessitating adjustments to the initial minimum number of points.

[0058] It should be noted that for frozen coal point cloud data, if it is clustered at a low initial minimum number of points but disappears or is segmented at a high initial minimum number of points, it indicates that the region where the frozen coal point cloud data is located has a low density, so the required minimum number of points needs to be reduced; conversely, if it is clustered at a low initial minimum number of points but remains stable at a high initial minimum number of points, it indicates that the region where the frozen coal point cloud data is located has a high density, so a higher minimum number of points can be used for constraint to avoid over-segmentation.

[0059] Based on this, embodiments of the present invention can cluster frozen coal point cloud data according to the global minimum initial minimum number of points and the maximum initial minimum number of points, and dynamically adjust the initial minimum number of points for each frozen coal point cloud data according to the differences between the clustering results.

[0060] For example, in an embodiment of the present invention, calculating the overlap of frozen coal point cloud data in a first cluster region and a second cluster region includes: recording the number of frozen coal point cloud data that simultaneously belong to both the first and second cluster regions as a first index of the frozen coal point cloud data; recording the total number of frozen coal point cloud data in the first and second cluster regions as a second index of the frozen coal point cloud data; and recording the ratio of the first index to the second index as the overlap of the frozen coal point cloud data.

[0061] Specifically, the current frozen coal point cloud data is clustered using the minimum initial minimum number of points and the target neighborhood radius to obtain the first clustering region of the current frozen coal point cloud data; the current frozen coal point cloud data is clustered using the maximum initial minimum number of points and the target neighborhood radius to obtain the second clustering region of the current frozen coal point cloud data.

[0062] The higher the overlap, the more the large area identified by the first clustering region overlaps with the high-density core identified by the second clustering region. The higher the probability that the frozen coal point cloud data is located in a solid whole, the more necessary it is to increase the initial minimum number of points so that the core region needs more neighboring points to be connected in the final clustering, thereby forming a more compact cluster boundary and effectively preventing over-segmentation with neighboring, slightly lower-density frozen coal blocks.

[0063] Conversely, the lower the overlap, the greater the structural difference between the large area identified in the first clustering region and the high-density core identified in the second clustering region. In this case, the frozen coal point cloud data is more likely to be located in a low-density region and can only be clustered under relatively lenient conditions. If the conditions are strict, it will not meet the requirements of the second clustering region, i.e., the core point. Therefore, the initial minimum number of points should be reduced to lower the threshold for becoming a core point and ensure that these real frozen coal point cloud data are not missed in the final clustering, thereby avoiding undersegmentation.

[0064] For example, in an embodiment of the present invention, adjusting the initial minimum number of points of the frozen coal point cloud data according to the relationship between the overlap and a set threshold to obtain the target minimum number of points includes: if the overlap of the frozen coal point cloud data is greater than the set threshold, multiplying the initial minimum number of points of the frozen coal point cloud data by a scaling factor to obtain the target minimum number of points, wherein the scaling factor is greater than 1; if the overlap of the frozen coal point cloud data is less than the set threshold, multiplying the initial minimum number of points of the frozen coal point cloud data by a scaling factor to obtain the target minimum number of points, wherein the scaling factor is less than 1; if the overlap of the frozen coal point cloud data is equal to the set threshold, then using the initial minimum number of points of the frozen coal point cloud data as the target minimum number of points.

[0065] Specifically, the scaling factor is calculated when the overlap of the frozen coal point cloud data exceeds a set threshold. See the following formula for details:

[0066] ;

[0067] For frozen coal point cloud data scaling factor, For frozen coal point cloud data The degree of overlap.

[0068] The greater the overlap, the larger the initial minimum number of points needs to be, and therefore the larger the corresponding scaling factor, in order to obtain a larger target minimum number of points.

[0069] Specifically, the scaling factor is calculated when the overlap of the frozen coal point cloud data is less than a set threshold. See the following formula for details:

[0070] ;

[0071] For frozen coal point cloud data scaling factor, For frozen coal point cloud data The degree of overlap.

[0072] Similarly, the greater the overlap of frozen coal point cloud data, the larger the target minimum number of points needs to be set.

[0073] Thus, based on the local density characteristics of frozen coal point cloud data with different degrees of overlap, the embodiments of the present invention can perform targeted analysis according to different density distributions, making the obtained target minimum number of points more consistent with the actual point cloud data classification and improving the accuracy of the final clustering results.

[0074] Based on the above steps, the target neighborhood radius and target minimum number of points for each frozen coal point cloud data can be obtained. By using the target neighborhood radius and target minimum number of points for each frozen coal point cloud data in DBSCAN clustering, each frozen coal area can be accurately identified.

[0075] The path planning module 104 is used to cluster the target neighborhood radius and the minimum number of target points based on the frozen coal point cloud data to obtain the target cleaning area; to plan the cleaning path of the cleaning robot based on the target cleaning area, and to control the cleaning robot to perform cleaning operations along the cleaning path.

[0076] It should be noted that clustering is performed based on the target neighborhood radius and the minimum number of target points of each frozen coal point cloud data obtained from the above steps. The identified frozen coal area is the area that the robot needs to clean. Setting the cleaning path according to the cleaning area can avoid interference from background structures and improve work efficiency.

[0077] For example, in an embodiment of the present invention, clustering based on the target neighborhood radius and the target minimum number of points of frozen coal point cloud data to obtain a target cleaning region includes: for a frozen coal point cloud data, if the number of frozen coal point cloud data in the target neighborhood radius is greater than or equal to the target minimum number of points, then the frozen coal point cloud data is marked as a core point; if the number of frozen coal point cloud data in the target neighborhood radius is less than the target minimum number of points, and the frozen coal point cloud data is located in the target neighborhood radius of other frozen coal point cloud data, then the frozen coal point cloud data is marked as a boundary point; connecting adjacent core points to form clusters, and boundary points are assigned to the clusters of their respective core points to obtain the target cleaning region.

[0078] For example, in an embodiment of the present invention, planning the cleaning path of the cleaning robot based on the target cleaning area includes: obtaining the projected area of ​​all frozen coal point cloud data in the target cleaning area on the horizontal plane, and using each projected area as the coal block size corresponding to the target cleaning area; setting a cleaning priority for each target cleaning area according to the coal block size, and sequentially accessing each target cleaning area based on the cleaning priority to obtain the cleaning path of the cleaning robot.

[0079] The size of the coal blocks in the target cleaning area is positively correlated with its cleaning priority; that is, the larger the coal blocks in the target cleaning area, the higher the priority for cleaning.

[0080] Specifically, when the cleaning robot accesses each target cleaning area sequentially based on the cleaning priority and obtains the cleaning path, it can call the 3D model of the background point cloud data of the support rods, carriage walls and the bottom of the carriage identified in the data acquisition module 101 above as an obstacle map to ensure that the generated path maintains a safe distance from these background structures and avoids collisions.

[0081] Starting from the current position of the cleaning robot, each target cleaning area is visited in sequence according to the cleaning priority. For a single target cleaning area, a reciprocating path can be used to cover its entire point cloud projection area to ensure that the cleaning head can fully cover the frozen coal block.

[0082] After obtaining the obstacle avoidance path planning for the cleaning robot according to the above steps, the obstacle avoidance safety operation of the cleaning robot on the train carriage can be realized.

[0083] For example, in an embodiment of the present invention, controlling a cleaning robot to perform cleaning operations along a cleaning path includes: converting the cleaning path into a sequence of execution instructions for the cleaning robot, and controlling the cleaning robotic arm to carry high-pressure nozzles to sequentially complete the cleaning operations of each target cleaning area.

[0084] The step of converting the cleaning path into a sequence of execution instructions for the cleaning robot can be implemented using existing technologies, and will not be elaborated upon here in this embodiment of the invention.

[0085] As can be seen, in this embodiment of the invention, when implementing automatic obstacle avoidance path planning for unloading frozen coal in a train carriage with a complex internal structure, a data acquisition module can be set up: used to acquire frozen coal point cloud data inside the carriage; a target neighborhood radius module: used to increase the neighborhood radius along a set radius range, taking the number of frozen coal point cloud data contained in each neighborhood radius as the local density; determining the neighborhood radius at the point where the local density change shows a downward inflection point based on the local density difference between each neighborhood radius and its adjacent neighborhood radii, as the target neighborhood radius of the frozen coal point cloud data; and a target minimum number of points module: used to acquire the initial minimum number of points of the frozen coal point cloud data, the initial minimum number of points being related to the global average point density and the target neighborhood radius. The neighborhood volume is positively correlated; clustering is performed on all frozen coal point cloud data using the minimum and maximum initial minimum point counts respectively to obtain the first and second cluster regions; the overlap between the first and second cluster regions of the frozen coal point cloud data is calculated, and the initial minimum point count of the frozen coal point cloud data is adjusted according to the relationship between the overlap count and a set threshold to obtain the target minimum point count; the path planning module is used to cluster based on the target neighborhood radius and the target minimum point count of the frozen coal point cloud data to obtain the target cleaning area; the cleaning path of the cleaning robot is planned based on the target cleaning area, and the cleaning robot is controlled to perform cleaning operations along the cleaning path, which effectively improves the accuracy of obstacle avoidance path planning.

[0086] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. An automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure, characterized in that: Includes the following modules: Data Acquisition Module: Used to acquire frozen coal point cloud data inside the car; Target Neighborhood Radius Module: Used to increase the neighborhood radius along a set radius range, taking the number of frozen coal point cloud data contained in each neighborhood radius as the local density; Based on the local density difference between each neighborhood radius and its adjacent neighborhood radii, the neighborhood radius at the point where the local density change shows a decreasing inflection point is determined, which is taken as the target neighborhood radius of the frozen coal point cloud data; Target Minimum Point Count Module: Used to acquire the initial minimum number of points in the frozen coal point cloud data, which is positively correlated with the global average point density and the neighborhood volume under the target neighborhood radius; Clustering was performed on all frozen coal point cloud data using the minimum and maximum initial minimum point counts respectively, resulting in the first and second clustering regions. The overlap between the first and second cluster regions of the frozen coal point cloud data is calculated, and the initial minimum number of points of the frozen coal point cloud data is adjusted according to the relationship between the overlap and a set threshold to obtain the target minimum number of points; the path planning module is used to cluster the frozen coal point cloud data based on the target neighborhood radius and the target minimum number of points to obtain the target cleaning area; the cleaning path of the cleaning robot is planned based on the target cleaning area, and the cleaning robot is controlled to perform cleaning operations along the cleaning path.

2. The automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure according to claim 1, characterized in that, The acquisition of point cloud data of frozen coal inside the carriage includes: Raw point cloud data of the interior of the car is collected. Background point cloud data in the raw point cloud data is removed through preprocessing to obtain frozen coal point cloud data of the frozen coal area of ​​the car.

3. The automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure according to claim 2, characterized in that, The step of increasing the neighborhood radius along a set radius range, and using the number of frozen coal point cloud data contained in each neighborhood radius as the local density, includes: The system presets the radius range and the growth step size of the neighborhood radius; within the radius range, it obtains a sequence of neighborhood radii that increases from left to right based on the growth step size; it obtains the maximum Z-axis value in all original point cloud data, and uses the maximum Z-axis value as the height of the cylindrical neighborhood corresponding to each neighborhood radius of the frozen coal point cloud data, and uses each neighborhood radius of the frozen coal point cloud data as the radius of the cylindrical neighborhood; it uses the number of frozen coal point cloud data contained in the cylindrical neighborhood under each neighborhood radius as the local density of the frozen coal point cloud data under each neighborhood radius.

4. The automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure according to claim 1, characterized in that, The step of determining the neighborhood radius at the point where the local density change shows a decreasing inflection point based on the local density difference between each neighborhood radius and its adjacent neighborhood radii, and using this as the target neighborhood radius for the frozen coal point cloud data, includes: Based on a three-fold growth step size, the adjacent neighborhood radii are obtained on both sides of the neighborhood radius of the frozen coal point cloud data. The density change rate of the neighborhood radius is obtained based on the normalized value of the local density difference between the left and right adjacent neighborhood radii. The target neighborhood radius of the frozen coal point cloud data is defined as the first descending inflection point obtained by traversing from the smallest neighborhood radius if the density change rate of the current neighborhood radius is less than 0.5 times that of the density change rate of the previous neighborhood radius.

5. The automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure according to claim 1, characterized in that, The initial minimum number of points for obtaining frozen coal point cloud data includes: The global average point density is obtained based on the local density of the target neighborhood radius of all frozen coal point cloud data. The product of the neighborhood volume under the target neighborhood radius of the frozen coal point cloud data and the global average point density is rounded up to obtain the initial minimum number of points of the frozen coal point cloud data.

6. The automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure according to claim 1, characterized in that, The calculation of the overlap between the first and second cluster regions of the frozen coal point cloud data includes: The number of frozen coal point cloud data that simultaneously belong to both the first and second cluster regions within the first and second cluster regions is recorded as the first indicator of the frozen coal point cloud data. The total number of frozen coal point cloud data in the first and second cluster regions is recorded as the second index of the frozen coal point cloud data; the ratio of the first index to the second index is recorded as the overlap of the frozen coal point cloud data.

7. The automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure according to claim 1, characterized in that, The process of adjusting the initial minimum number of points in the frozen coal point cloud data based on the relationship between overlap and a set threshold to obtain the target minimum number of points includes: If the overlap of the frozen coal point cloud data is greater than a set threshold, the initial minimum number of points in the frozen coal point cloud data is multiplied by a scaling factor to obtain the target minimum number of points, where the scaling factor is greater than 1; if the overlap of the frozen coal point cloud data is less than a set threshold, the initial minimum number of points in the frozen coal point cloud data is multiplied by a scaling factor to obtain the target minimum number of points, where the scaling factor is less than 1; if the overlap of the frozen coal point cloud data is equal to a set threshold, the initial minimum number of points in the frozen coal point cloud data is used as the target minimum number of points.

8. The automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure according to claim 1, characterized in that, The target cleaning region is obtained by clustering based on the target neighborhood radius and the minimum number of target points of the frozen coal point cloud data, including: For a frozen coal point cloud data, if the number of frozen coal point cloud data in the target neighborhood radius is greater than or equal to the target minimum number of points, then the frozen coal point cloud data is marked as a core point; if the number of frozen coal point cloud data in the target neighborhood radius is less than the target minimum number of points, and the frozen coal point cloud data is located in the target neighborhood radius of other frozen coal point cloud data, then the frozen coal point cloud data is marked as a boundary point; connecting adjacent core points forms a cluster, and boundary points are assigned to the cluster of their respective core points to obtain the target cleaning area.

9. The automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure according to claim 1, characterized in that, The cleaning path of the cleaning robot based on the target cleaning area includes: Obtain the projected area of ​​all frozen coal point cloud data in the target cleaning area on the horizontal plane, and use each projected area as the size of the coal block corresponding to the target cleaning area; The cleaning priority is set for each target cleaning area based on the size of the coal block. The cleaning robot visits each target cleaning area in sequence based on the cleaning priority to obtain the cleaning path.

10. The automatic obstacle avoidance path planning system for unloading frozen coal from a train carriage with a complex internal structure according to claim 1, characterized in that, The control of the cleaning robot to perform cleaning operations along the cleaning path includes: The cleaning path is converted into a sequence of execution instructions for the cleaning robot, which controls the cleaning robotic arm to carry the high-pressure nozzle to complete the cleaning operation of each target cleaning area in sequence.

Images