PCDA-based automatic material leveling method for C-type stockyard

Through the automatic flattening method based on PCDA, the scanner obtains and processes point cloud data, and the automated operation of the C-type material field is achieved, which solves the problems of unclear silo status and unreliable detectors, and improves efficiency and accuracy. Adapting to multi-modal and multi-working conditions reduces the work intensity of personnel.

CN119006502APending Publication Date: 2024-11-22JIANGSU JINHENG INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411089544.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The automatic control of existing C-type material yard scrapers has problems such as unclear silo status, unreliable limit detectors, limited installation location, slow operation rhythm and inability to cope with multiple stack-tip stack types, resulting in inefficiency of the system and High work intensity of personnel.

Method used

Using the automatic flattening method based on PCDA, point cloud data is obtained through the scanner, data analysis, filtering, segmentation and edge extraction are performed, silo status and boundary data are obtained, and scraper automatic operation is realized, reducing dependence on manual intervention. .

Benefits of technology

It significantly improves the efficiency and accuracy of the material yard operation, solves the problems of unclear silo status and unreliable detectors, adapts to multi-modal and multi-working conditions, improves the scope of application and operation effect of the system, and reduces the intensity of personnel .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006502A_ABST
    Figure CN119006502A_ABST
Patent Text Reader

Abstract

The invention discloses a PCDA-based automatic material leveling method for a C-type material yard. The method comprises the following steps: (1) obtaining material pile state data of a C-type material bin; (2) analyzing the point cloud data; (3) filtering the point cloud data; (4) segmenting point cloud data; (5) point cloud edge extraction; (6) the scraper machine obtains the state of the stock bin; (7) the state of the stock bin is judged, if the stock bin is in the state that the materials are newly unloaded, the materials are leveled and go to the highest point, and if the stock bin is in the state that the material piles are horizontally taken, the materials are taken and go to the continuous taking point; (8) a horizontal material taking process is executed; and (9) horizontal material taking is finished. According to the method, scanner point cloud data and a related processing segmentation recognition algorithm are utilized, so that the efficiency and the accuracy of stock ground operation are remarkably improved, and meanwhile, the dependence on manual intervention is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of raw material storage and transportation, and in particular relates to an automatic material leveling method for a C-type material yard based on PCDA. Background Art

[0002] In the steelmaking process, the raw material yard area is responsible for the storage, processing, transportation and distribution of raw materials and auxiliary materials. It is the first process of steel production, the key node of the internal supply chain of steel enterprises and the production unit with the worst environment. Among them, the C-type material yard plays an indispensable role as the core of raw material storage and transportation.

[0003] At present, the production method of C-type material field scraper still relies on manual identification of the material bin status. Based on experience and visual observation of the material pile shape, it is judged whether it is manually controlled or automatically controlled by a limit monitor. The existing automatic control process is as follows: Figure 1 shown.

[0004] The existing technology uses a limit monitor or radar device to detect the high point of the material pile and the Y-axis limit position at the same time. If the stop condition is not met, the main arm will move down and determine whether to move to another area based on the existing flow system. This reciprocating process can achieve the effect of automatic material removal production.

[0005] The existing technology uses limit monitors or radar devices to guide the scraper to operate, which has many disadvantages:

[0006] First, the silo status is unclear and the system cannot know the silo status, thus affecting the automatic operation of the system;

[0007] Secondly, the limit detector is unreliable and usually requires the installation of multiple redundancies to meet the system requirements, which increases the construction and maintenance costs.

[0008] Third, the installation position of the limit detector is limited, which cannot meet the needs of all types of scrapers on the market and has a narrow scope of application;

[0009] Fourth, the existing technical guidance operation pace is slow and cannot meet the actual production needs on site;

[0010] Fifth, the existing technology cannot cope with the multi-pile type, cannot adapt to the multi-mode and multi-operating conditions, and has poor operating effects in actual application scenarios;

[0011] Sixth, the accuracy of pile boundary detection is poor. Summary of the invention

[0012] Purpose of the invention: In view of the problems in the prior art that the status of the intermediate material silo cannot be obtained in time and the detector equipment is unreliable and has limited installation positions, the present invention provides an automatic material leveling method for C-type material yard based on PCDA, which utilizes scanner point cloud data and related processing segmentation and recognition algorithms, thereby significantly improving the efficiency and accuracy of material yard operations and reducing dependence on manual intervention.

[0013] Technical solution: To solve the above technical problems, the present invention provides an automatic material leveling method for a C-type material yard based on PCDA, comprising the following steps:

[0014] Step 1: Obtain the data of the C-type silo stockpile status, make a comprehensive judgment on the data, and then write the data into the silo management system;

[0015] Step 2: Parse point cloud data;

[0016] Step 3: Point cloud data filtering;

[0017] Step 4: Point cloud data segmentation;

[0018] Step 5: Point cloud edge extraction;

[0019] Step 6: The scraper obtains the silo status, obtains the boundary data corresponding to the processed point cloud data, and obtains the material point data;

[0020] Step 7: Determine the state of the silo. If the silo is in a state where materials have been newly unloaded, level the material and proceed to step 8. If the silo is in a state where materials have been taken out of the silo, take out the materials and proceed to step 9.

[0021] Step 8: The system goes to the highest point;

[0022] Step 9: The system goes to the renewal point;

[0023] Step 10: Execute the horizontal material collection process. When the scraper reaches the predetermined starting position, the main arm starts and reciprocates according to the boundary data. Each time it reaches a boundary point, it descends a predetermined angle and turns back at the same time, operating layer by layer until it reaches the predetermined angle or the tonnage of materials transported meets the production requirements;

[0024] Step 11: The horizontal material collection is completed.

[0025] Furthermore, the specific steps of obtaining the C-type silo stockpile status data in step 1 are as follows:

[0026] Get the scanner's moving position and scanning status information; get the unloading vehicle information; get the upstream dock and upstream belt information; get the scraper's position, scraper's operating status, and storage location information.

[0027] Furthermore, the specific steps of parsing the point cloud data in step 2 are as follows:

[0028] The PCD file of point cloud data is a storage format for laser point cloud. The PCD file is mainly a list of Cartesian coordinates (x, y, z) and intensity values ​​i, that is, each point cloud is accompanied by a unique three-dimensional coordinate system and energy reflection intensity.

[0029] In this coordinate system, the x-axis points to the front of the scanner, and the y-axis points to the left of the scanner; because this coordinate system uses the right-hand rule, the z-axis of the coordinate system points to the top of the scanner;

[0030] In the point cloud data, the first two bytes, such as A and B, represent the rotation angle, the middle two bytes, such as C and D, represent the distance, and the last byte, such as E, represents the intensity;

[0031] Step 2.1, calculate the rotation angle value α of the laser line: invert the two bytes A and B of the rotation angle into hexadecimal, and then convert them into unsigned decimal, then the obtained value is the current rotation angle value α;

[0032] Step 2.2, calculate the distance R measured by the laser line: invert the two bytes C and D representing the distance into hexadecimal, then convert to unsigned decimal, and then multiply by the instrument w resolution. The value obtained is the measured distance R from the laser point to the scanner;

[0033] Step 2.3, obtain the timestamp and lidar model parameters of the frame;

[0034] Step 2.4, convert the angle and distance information into three-dimensional coordinate XYZ values;

[0035] The three-dimensional coordinates XYZ values ​​are obtained by comprehensive transformation of the rotation angle α, the vertical angle ω, which is the fixed value corresponding to each laser beam, and the trigonometric function of the distance value R:

[0036] X=R*cos(ω))*sin(α)

[0037] Y=R*cos(ω)*cos(α)

[0038] Z = R*sin(ω).

[0039] Furthermore, the specific steps of point cloud data filtering in step 3 are as follows:

[0040] Considering the characteristics of outliers, it is defined that if the point cloud is less than a certain density, the point cloud is invalid; the average distance from each point to its nearest k points is calculated, and the distances of all points in the point cloud form a Gaussian distribution; given the mean and variance, points outside 3∑ can be eliminated;

[0041]

[0042] Where A is the amplitude, x0, y0, z0 are the coordinates of the center point, is the variance.

[0043] Furthermore, the specific steps of point cloud data segmentation in step 4 are as follows:

[0044] The segmentation algorithm is used to separate the pile from the background. The DBSCAN algorithm based on Euclidean distance is used to distinguish the pile from the wall support of the C-type silo. Point cloud segmentation based on DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm that divides point cloud data into clusters with different densities and identifies noise points. The following are the main steps of point cloud segmentation based on DBSCAN:

[0045] Define the distance threshold ε and the minimum number of points MinPts: The core of the DBSCAN algorithm is to determine the core points, boundary points, and noise points by setting the distance threshold and the minimum number of points;

[0046] The distance threshold ε defines the neighborhood range of a point; points within the range of ε are considered to be adjacent to each other;

[0047] The minimum number of points MinPts defines the minimum number of points in the neighborhood required for a core point; if the number of points in the neighborhood of a point is greater than or equal to MinPts, then this point is a core point;

[0048] Initialize the label of each point: set the initial label of all points to UNCLASSIFIED;

[0049] Iterate over each point:

[0050] For each unclassified point, check the number of points in its ε neighborhood;

[0051] If the number of points in the neighborhood is less than MinPts, the point is marked as a noise point NOISE;

[0052] If the number of points in the neighborhood is greater than or equal to MinPts, the point is marked as the current cluster label, and the points in its neighborhood are also marked with the same cluster label;

[0053] For the points in the neighborhood of the core point, the neighborhood is recursively expanded to add the points in the neighborhood to the same cluster;

[0054] After traversing all points, all core points and boundary points are marked as a cluster, while noise points remain unclassified;

[0055] Return clustering results: Points with the same clustering label are classified into the same cluster and the final clustering results are returned.

[0056] Furthermore, the specific steps of point cloud edge extraction in step 5 are as follows:

[0057] Step 5.1, calculation of normal vector:

[0058] First, a normal vector is estimated for each point in the point cloud. The set of neighboring points of each point is found through the K-nearest neighbors search strategy, and the least squares plane is fitted to these points. The normal of the fitted plane is the estimated normal vector of the origin.

[0059] Step 5.2, normal vector difference analysis:

[0060] Points near edges or corners have normal vectors that are significantly different from those of neighboring points; this is measured by calculating the angular difference between the normal vectors of adjacent points.

[0061] Step 5.3: Edge extraction:

[0062] Based on the normal vector difference, the threshold method is further used to screen edge points; a suitable threshold k is the empirical value λ, and all points whose normal vector difference exceeds this threshold are regarded as edge points, thereby extracting the edge of the pile.

[0063] Furthermore, in step 6, the scraper conveyor obtains the silo status and the specific steps of the silo point cloud data are as follows:

[0064] Step 6.1, obtain the silo status corresponding to the target silo number from the upper PLC, including the location of the silo re-collection point and the last material collection direction information;

[0065] Step 6.2: Obtain the target bin highest point data and target bin boundary data from the server.

[0066] Furthermore, the specific steps of executing the horizontal material collection process in step 10 are as follows:

[0067] Step 10.1, when reaching the predetermined position, i.e., the highest point or the continuation point, start the process;

[0068] Step 10.2, the main arm of the scraper starts to rotate and moves according to the driving direction determined by the system;

[0069] Step 10.3: After reaching one side boundary, the scraper stops moving and the main arm drops to a predetermined angle;

[0070] Step 10.4: After the angle is lowered, the scraper moves to the other side until it reaches the boundary of the other side;

[0071] Step 10.5, repeat this process until the system stops when the manually set flattening end angle is reached in the flattening mode; and stops when the system receives a signal from the downstream equipment that the flattening is complete in the retrieving mode.

[0072] Compared with the prior art, the advantages of the present invention are:

[0073] The method solves the problem in the prior art that the state of the silo is unclear and the system cannot know the state of the silo through step 1; secondly, by adding a 3D scanner to the scraper system, the problem of unreliable limit detectors and limited installation positions of limit detectors is solved; finally, through steps 6 and 10, the problem of the prior art guiding operation rhythm is slow, it cannot cope with the pile type with multiple pile points, and the pile boundary detection accuracy is poor. At the same time, it also liberates manpower, greatly reduces the work intensity of personnel and improves the working environment of personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 A flow chart of the prior art;

[0075] Figure 2 is a flow chart of the present invention;

[0076] Figure 3 A set of point cloud data in a specific embodiment. DETAILED DESCRIPTION

[0077] The present invention is further explained below in conjunction with the accompanying drawings and specific implementation methods.

[0078] like Figure 1 As shown, the present invention provides an automatic material leveling method for a C-type material yard based on PCDA, comprising the following steps:

[0079] Step 1: Obtain the data of the C-type silo stockpile status, make a comprehensive judgment on the data, and then write the data into the silo management system;

[0080] Get the scanner's moving position and scanning status information; get the unloading vehicle information; get the upstream dock and upstream belt information; get the scraper's position, scraper's operating status, and storage location information.

[0081] Step 2: Parse point cloud data;

[0082] The PCD file of point cloud data is a storage format for laser point cloud. The PCD file is mainly a list of Cartesian coordinates (x, y, z) and intensity values ​​i, that is, each point cloud is accompanied by a unique three-dimensional coordinate system and energy reflection intensity.

[0083] In this coordinate system, the x-axis points to the front of the scanner, and the y-axis points to the left of the scanner; because this coordinate system uses the right-hand rule, the z-axis of the coordinate system points to the top of the scanner;

[0084] Taking data acquisition as an example, briefly describe the data processing steps.

[0085] In the point cloud data, the first two bytes, such as A and B, represent the rotation angle, the middle two bytes, such as C and D, represent the distance, and the last byte, such as E, represents the intensity;

[0086] like Figure 3 As shown, E0,63 represents the rotation angle, B6,07 represents the distance, and 2A represents the intensity.

[0087] Step 2.1, calculate the rotation angle value α of the laser line: invert the two bytes A and B of the rotation angle into hexadecimal, and then convert them into unsigned decimal, then the obtained value is the current rotation angle value α;

[0088] Step 2.2, calculate the distance R measured by the laser line: invert the two bytes C and D representing the distance into hexadecimal, then convert to unsigned decimal, and then multiply by the instrument w resolution. The value obtained is the measured distance R from the laser point to the scanner;

[0089] Step 2.3, obtain the timestamp and lidar model parameters of the frame;

[0090] Step 2.4, convert the angle and distance information into three-dimensional coordinate XYZ values;

[0091] The three-dimensional coordinates XYZ values ​​are obtained by comprehensive transformation of the rotation angle α, the vertical angle ω, which is the fixed value corresponding to each laser beam, and the trigonometric function of the distance value R:

[0092] X=R*cos(ω)*sin(α)

[0093] Y=R*cos(ω)*cos(α)

[0094] Z = R*sin(ω).

[0095] Step 3: Point cloud data filtering;

[0096] Considering the characteristics of outliers, it is defined that if the point cloud is less than a certain density, the point cloud is invalid; the average distance from each point to its nearest k points is calculated, and the distances of all points in the point cloud form a Gaussian distribution; given the mean and variance, points outside 3∑ can be eliminated;

[0097]

[0098] Where A is the amplitude, x0, y0, z0 are the coordinates of the center point, is the variance.

[0099] Step 4: Point cloud data segmentation;

[0100] The segmentation algorithm is used to separate the pile from the background. The DBSCAN algorithm based on Euclidean distance is used to distinguish the pile from the wall support of the C-type silo. Point cloud segmentation based on DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm that divides point cloud data into clusters with different densities and identifies noise points. The following are the main steps of point cloud segmentation based on DBSCAN:

[0101] Define the distance threshold ε and the minimum number of points MinPts: The core of the DBSCAN algorithm is to determine the core points, boundary points, and noise points by setting the distance threshold and the minimum number of points;

[0102] The distance threshold ε defines the neighborhood range of a point; points within the range of ε are considered to be adjacent to each other;

[0103] The minimum number of points MinPts defines the minimum number of points in the neighborhood required for a core point; if the number of points in the neighborhood of a point is greater than or equal to MinPts, then this point is a core point;

[0104] Initialize the label of each point: set the initial label of all points to UNCLASSIFIED;

[0105] Iterate over each point:

[0106] For each unclassified point, check the number of points in its ε neighborhood;

[0107] If the number of points in the neighborhood is less than MinPts, the point is marked as a noise point NOISE;

[0108] If the number of points in the neighborhood is greater than or equal to MinPts, the point is marked as the current cluster label, and the points in its neighborhood are also marked with the same cluster label;

[0109] For the points in the neighborhood of the core point, the neighborhood is recursively expanded to add the points in the neighborhood to the same cluster;

[0110] After traversing all points, all core points and boundary points are marked as a cluster, while noise points remain unclassified;

[0111] Return clustering results: Points with the same clustering label are classified into the same cluster and the final clustering results are returned.

[0112] Step 5: Point cloud edge extraction;

[0113] Step 5.1, calculation of normal vector:

[0114] First, a normal vector is estimated for each point in the point cloud. The set of neighboring points of each point is found through the K-nearest neighbors search strategy, and the least squares plane is fitted to these points. The normal of the fitted plane is the estimated normal vector of the origin.

[0115] Step 5.2, normal vector difference analysis:

[0116] Points near edges or corners have normal vectors that are significantly different from those of neighboring points; this is measured by calculating the angular difference between the normal vectors of adjacent points.

[0117] Step 5.3: Edge extraction:

[0118] Based on the normal vector difference, the threshold method is further used to screen edge points; a suitable threshold k is the empirical value λ, and all points whose normal vector difference exceeds this threshold are regarded as edge points, thereby extracting the edge of the pile.

[0119] Step 6: The scraper obtains the silo status, obtains the boundary data corresponding to the processed point cloud data, and obtains the material point data;

[0120] Step 6.1, obtain the silo status corresponding to the target silo number from the upper PLC, including the location of the silo re-collection point and the last material collection direction information;

[0121] Step 6.2: Obtain the target bin highest point data and target bin boundary data from the server.

[0122] Step 7: Determine the state of the silo. If the silo has just been unloaded, level the material and proceed to step 8. If the silo has been unloaded, remove the material and proceed to step 9.

[0123] Step 8: The system goes to the highest point;

[0124] Step 9: The system goes to the renewal point;

[0125] Step 10: Execute the horizontal material collection process. When the scraper reaches the predetermined starting position, the main arm starts and reciprocates according to the boundary data. Each time it reaches a boundary point, it descends a predetermined angle and turns back at the same time, operating layer by layer until it reaches the predetermined angle or the tonnage of materials transported meets the production requirements;

[0126] Step 10.1, when reaching the predetermined position, i.e., the highest point or the continuation point, start the process;

[0127] Step 10.2, the main arm of the scraper starts to rotate and moves according to the driving direction determined by the system;

[0128] Step 10.3: After reaching one side boundary, the scraper stops moving and the main arm drops to a predetermined angle;

[0129] Step 10.4: After the angle is lowered, the scraper moves to the other side until it reaches the boundary of the other side;

[0130] Step 10.5, repeat this process until the system stops when the manually set flattening end angle is reached in the flattening mode; and stops when the system receives a signal from the downstream equipment that the flattening is complete in the retrieving mode.

[0131] Step 11: The horizontal material collection is completed.

[0132] When implementing:

[0133] The operator issues a change-bin command, and the scraper executes the change-bin command, moving from the greenhouse 167M-16# bin to the 212M-20# bin. When the scraper performs the operation process or the bin change process on the track, the scanner collects point cloud data in real time, and after being processed by the data acquisition IPC program, it is uploaded to the server for data analysis.

[0134] 1. When the scraper system sends a walking feedback signal. Since in this specific embodiment, the X-axis of the laser scanner is perpendicular to the scraper track, taking the system starting position as an example, the data acquisition IPC program collects the current position information of the scraper as the X-axis coordinate position (167M) of the point cloud data. The point cloud analysis program takes the laser vertical angle ω (30°) to calculate the Y and Z axis data of the current point cloud data, and obtains the data of the current point (X-167, Y-25, Z-8). By analogy, when ω changes from 30° to 150°, the system will collect the point cloud data information of the surface where 167 meters is located. When X changes from 167 to 212, the system will collect the point cloud model data of all silos between 167 meters and 212 meters. At the same time, the system obtains the upstream belt and unloading vehicle system signals. According to the current position of the unloading vehicle in the 17# silo, which is in the unloading state, the system determines that the 17# silo model data has changed and sets the silo to the unscanning state. The data will be uploaded to the silo management system for unified record and storage.

[0135] 2. In the point cloud processing server, the data parsed by the IPC program of each scraper data collection are aggregated and filtered. In this specific embodiment, a statistical filtering method is used. 2.1 First calculate the average distance. For each point in the point cloud, calculate its distance to the nearest k points. In this embodiment, the value is 30. Calculate the average of these distances as the average neighboring distance of the point; 2.2 Construct a distance distribution. Collect the average neighboring distances of all points and construct a distance distribution; 2.3 Calculate the mean and variance. Calculate the mean and variance of the distance distribution; 2.4 Eliminate outliers. In this embodiment, eliminate those points whose average neighboring distance is outside 2σ to obtain the denoised point cloud data.

[0136] 3. The algorithm program first reads the integrated denoised silo point cloud data and executes the segmentation algorithm; 3.1 Select core points: For each point in the data set, check the number of points in its neighborhood (distance less than a certain threshold eps). If the number of points is greater than or equal to the minimum number of points min_points, the point is considered to be a core point; 3.2 Build clusters: Starting from any core point, add all points in its neighborhood (including core points and boundary points) to a new cluster. If there are other core points in the neighborhood, continue to expand the cluster until no new core points are added; 3.3 Mark noise points: Those points that are not included in any cluster are marked as noise points. In this specific embodiment, the eps value is 0.02 and the minpoints value is 10.

[0137] 4. The algorithm reads the denoised point cloud data and executes the edge extraction algorithm. 4.1 Calculate the point cloud normal vector, use the K nearest neighbor algorithm to find n neighboring points around each point in the point cloud to estimate the normal vector, and this embodiment finds 30 neighboring points; 4.2 Calculate the point cloud curvature, calculate the point cloud curvature based on the normal vector data obtained in 4.1, take the data with curvature greater than 0.05, and obtain the point cloud edge data. The point cloud edge data can calculate the data required for the current warehouse, including the highest point position 20# warehouse (224M, 25°), left boundary (234M), and right boundary (213M) data.

[0138] 5. When the scraper moves to the 20# bin, the system has obtained the current bin status as leveled. The leveled status means that the scraper system has performed leveling operations in this bin. When the leveling operation is completed, the position of the scraper system at the end, the direction and angle information of the continued material will be stored. After that, the system automatically goes to the continued material point (223M, left, 23.3°) to execute the material collection process. When the unloading weight of the downstream bin reaches the standard, the operator selects the stop process button and the process ends.

Claims

1. An automatic material leveling method for C-type material yard based on PCDA, characterized in that: The steps include: Step 1: Obtain the data of the C-type silo stockpile status, make a comprehensive judgment on the data, and then write the data into the silo management system; Step 2: Parse point cloud data; Step 3: Point cloud data filtering; Step 4: Point cloud data segmentation; Step 5: Point cloud edge extraction; Step 6: The scraper obtains the silo status, obtains the boundary data corresponding to the processed point cloud data, and obtains the material point data; Step 7: Determine the state of the silo. If the silo is in a state where materials have been newly unloaded, level the material and proceed to step 8. If the silo is in a state where materials have been taken out of the silo, take out the materials and proceed to step 9. Step 8: The system goes to the highest point; Step 9: The system goes to the renewal point; Step 10: Execute the horizontal material collection process. When the scraper reaches the predetermined starting position, the main arm starts and reciprocates according to the boundary data. Each time it reaches a boundary point, it descends a predetermined angle and turns back at the same time, operating layer by layer until it reaches the predetermined angle or the tonnage of materials transported meets the production requirements; Step 11: The horizontal material collection is completed.

2. The automatic material leveling method for C-type material yard based on PCDA according to claim 1 is characterized in that: The specific steps for obtaining the C-type silo stockpile status data in step 1 are as follows: Get the scanner's moving position and scanning status information; get the unloading vehicle information; get the upstream dock and upstream belt information; get the scraper's position, scraper's operating status, and storage location information.

3. The automatic material leveling method for C-type material yard based on PCDA according to claim 1 is characterized in that: The specific steps of parsing the point cloud data in step 2 are as follows: The PCD file of point cloud data is a storage format for laser point cloud. The PCD file is mainly a list of Cartesian coordinates (x, y, z) and intensity values ​​i, that is, each point cloud is accompanied by a unique three-dimensional coordinate system and energy reflection intensity. In this coordinate system, the x-axis points to the front of the scanner, and the y-axis points to the left of the scanner; because this coordinate system uses the right-hand rule, the z-axis of the coordinate system points to the top of the scanner; In the point cloud data, the first two bytes, such as A and B, represent the rotation angle, the middle two bytes, such as C and D, represent the distance, and the last byte, such as E, represents the intensity; Step 2.1, calculate the rotation angle value α of the laser line: invert the two bytes A and B of the rotation angle into hexadecimal, and then convert them into unsigned decimal, then the obtained value is the current rotation angle value α; Step 2.2, calculate the distance R measured by the laser line: invert the two bytes C and D representing the distance into hexadecimal, then convert to unsigned decimal, and then multiply by the instrument w resolution. The value obtained is the measured distance R from the laser point to the scanner; Step 2.3, obtain the timestamp and laser radar model parameters of the frame; Step 2.4, convert the angle and distance information into three-dimensional coordinate XYZ values; The three-dimensional coordinates XYZ values ​​are obtained by comprehensive transformation of the rotation angle α, the vertical angle ω, which is the fixed value corresponding to each laser beam, and the trigonometric function of the distance value R: X=R*cos(ω)*sin(α) Y=R*cos(ω)*cos(α) Z = R*sin(ω).

4. The automatic material leveling method for C-type material yard based on PCDA according to claim 1 is characterized in that: The specific steps of point cloud data filtering in step 3 are as follows: Considering the characteristics of outliers, it is defined that if the point cloud is less than a certain density, the point cloud is invalid; the average distance from each point to its nearest k points is calculated, and the distances of all points in the point cloud form a Gaussian distribution; given the mean and variance, points outside 3∑ can be eliminated; Where A is the amplitude, x0, y0, z0 are the coordinates of the center point, is the variance.

5. The automatic material leveling method for C-type material yard based on PCDA according to claim 1 is characterized in that: The specific steps of point cloud data segmentation in step 4 are as follows: Use the segmentation algorithm to separate the pile from the background; use the clustering DBSCAN algorithm based on Euclidean distance to distinguish the pile from the wall support of the C-type silo; DBSCAN-based point cloud segmentation is a density-based clustering algorithm that divides point cloud data into clusters with different densities and identifies noise points; the following are the main steps of DBSCAN-based point cloud segmentation: Define the distance threshold ε and the minimum number of points MinPts: The core of the DBSCAN algorithm is to determine the core points, boundary points, and noise points by setting the distance threshold and the minimum number of points; The distance threshold ε defines the neighborhood range of a point; points within the range of ε are considered to be adjacent to each other; The minimum number of points MinPts defines the minimum number of points in the neighborhood required for a core point; if the number of points in the neighborhood of a point is greater than or equal to MinPts, then this point is a core point; Initialize the label of each point: set the initial label of all points to UNCLASSIFIED; Iterate over each point: For each unclassified point, check the number of points in its ε neighborhood; If the number of points in the neighborhood is less than MinPts, the point is marked as a noise point NOISE; If the number of points in the neighborhood is greater than or equal to MinPts, the point is marked as the current cluster label, and the points in its neighborhood are also marked with the same cluster label; For the points in the neighborhood of the core point, the neighborhood is recursively expanded to add the points in the neighborhood to the same cluster; After traversing all points, all core points and boundary points are marked as a cluster, while noise points remain unclassified; Return clustering results: Points with the same clustering label are classified into the same cluster and the final clustering results are returned.

6. The automatic material leveling method for C-type material yard based on PCDA according to claim 1 is characterized in that: The specific steps of point cloud edge extraction in step 5 are as follows: Step 5.1, calculation of normal vector: First, a normal vector is estimated for each point in the point cloud. The set of neighboring points of each point is found through the K-nearest neighbor search strategy, and the least squares plane is fitted to these points. The normal of the fitted plane is the estimated normal vector of the origin. Step 5.2, normal vector difference analysis: Points near edges or corners have normal vectors that are significantly different from those of neighboring points; this is measured by calculating the angle difference between the normal vectors of adjacent points. Step 5.3: Edge extraction: Based on the normal vector difference, the threshold method is further used to screen edge points; a suitable threshold k is the empirical value λ, and all points whose normal vector difference exceeds this threshold are regarded as edge points, thereby extracting the edge of the pile.

7. The automatic material leveling method for C-type material yard based on PCDA according to claim 1 is characterized in that: In step 6, the scraper obtains the silo status, and the specific steps of the silo point cloud data are as follows: Step 6.1, obtain the silo status corresponding to the target silo number from the upper PLC, including the location of the silo re-collection point and the last material collection direction information; Step 6.2: Obtain the target bin highest point data and target bin boundary data from the server.

8. The automatic material leveling method for C-type material yard based on PCDA according to claim 1 is characterized in that: The specific steps of executing the horizontal material collection process in step 10 are as follows: Step 10.1, when reaching the predetermined position, i.e., the highest point or the continuation point, start the process; Step 10.2, the main arm of the scraper starts to rotate and moves according to the driving direction determined by the system; Step 10.3: After reaching one side boundary, the scraper stops moving and the main arm drops to a predetermined angle; Step 10.4: After the angle is lowered, the scraper moves to the other side until it reaches the boundary of the other side; Step 10.5, repeat this process until the system stops when the manually set flattening end angle is reached in the flattening mode; and stops when the system receives a signal from the downstream equipment that the flattening is complete in the retrieving mode.