Fully automatic palletizing method and system based on machine vision and path planning

Abstract

The present invention relates to the field of logistics automation technology, specifically to a fully automatic palletizing method and system based on machine vision and path planning, comprising the following steps: based on three-dimensional point cloud data of goods, extracting point cloud coordinates, normal vectors and curvature values ​​of the surface of goods, calculating the geometric segmentation range of point cloud edge features, and calculating the fitting relationship between boundary size and curvature distribution. In the present invention, surface point cloud coordinates, normal vectors and curvature values ​​are extracted based on three-dimensional point cloud data to enhance the accuracy of the description of geometric features of the surface of complex goods. Through spatial distribution analysis and feature set classification, the geometric center and variance range are reasonably determined to improve the accuracy and stability of shape classification. Through the calculation of center of gravity and size deviation, the contact point position and pressure distribution are optimized. Combined with the spatial priority analysis of grabbing points and obstacle points, the path planning and obstacle point elimination strategy are optimized to reduce path conflicts and improve the flexibility and efficiency of logistics operations.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics automation, and particularly to a full-automatic palletizing method and system based on machine vision and path planning. Background Art

[0002] The technical field of logistics automation includes related methods and systems for automating the management of logistics activities based on computer and intelligent technologies. Its core content is to utilize means such as information technology, artificial intelligence, and mechanical automation to optimize and control the entire process of logistics transportation, warehousing management, goods sorting, and distribution, thereby improving logistics efficiency. This technical field systematically covers the full-process automation from goods storage, sorting, loading and unloading to transportation, mainly including key technical modules such as path planning, robotic arm operation, visual recognition, and dynamic scheduling, and realizes precise control of logistics operations by combining various sensor data collection and real-time feedback technologies.

[0003] Among them, the full-automatic palletizing method refers to an operation method that realizes automatic identification, classification, grasping, and neat stacking of goods by combining machine vision technology and path planning technology. This patent theme aims at the demand for palletizing operations in the logistics or industrial production process. By using machine vision technology to identify the shape, size, and position of goods, combined with path planning algorithms to determine the optimal movement trajectory of the robotic arm, and using the robotic arm or other execution devices to complete the grasping and placement of goods, finally forming a stable and regular stack.

[0004] The existing technologies are not fine enough in the geometric description of complex goods boundaries and surface features, and it is difficult to accurately extract the boundary characteristics of complex-shaped goods. The analysis of the center-of-gravity offset and size deviation of goods lacks the ability of dynamic adjustment, resulting in poor stability of stacked goods. The analysis of the spatial influence range of obstacle points is insufficient, and it is difficult for path planning to effectively avoid low-priority obstacles, increasing the handling time and complexity. The real-time data processing response ability is insufficient, and it is difficult to achieve precise control of path optimization and operation process in a dynamic environment, affecting the logistics operation efficiency and reliability. Summary of the Invention

[0005] The purpose of the present invention is to solve the drawbacks existing in the prior art, and to propose a full-automatic palletizing method and system based on machine vision and path planning.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solutions: A full-automatic palletizing method based on machine vision and path planning, including the following steps:

[0007] S1: Based on the three-dimensional point cloud data of the goods, extract the point cloud coordinates, normal vectors, and curvature values on the surface of the goods, calculate the geometric segmentation range of the point cloud edge features, calculate the fitting relationship between the boundary size and the curvature distribution, and use the multinomial distribution of the relationship between the boundary size and the curvature for verification to generate the geometric feature description value of the goods;

[0008] S2: Based on the geometric feature description value of the goods, extract the curvature distribution and the boundary feature position of the goods, calculate the spatial range of the feature points of the Euclidean distance, judge the geometric center position and the distribution variance range, classify and process the feature set that meets the center distance and variance range, and generate the shape classification coefficient of the goods;

[0009] S3: Based on the shape classification coefficient of the goods, extract the center of gravity coordinates and the three-dimensional size range of the goods, calculate the relationship between the offset of the center of gravity coordinates of the goods and the size deviation, judge the distribution area of the stacking center of gravity offset and the stress balance state, adjust the contact point position and the pressure value range, and generate the stress distribution of the goods stacking;

[0010] S4: Based on the stress distribution of the goods stacking, extract the contact point pressure value and the center of gravity offset range, calculate the center of gravity offset adjustment parameter and the contact point stacking matrix, accumulate all the stacked energy relationship values, establish the stacked energy optimization matrix relationship of all the goods, and generate the stacked energy optimization value of the goods;

[0011] S5: Based on the stacked energy optimization value of the goods, extract the three-dimensional position relationship between the grasping point and the placing point, calculate the spatial priority value of the grasping point and the obstacle point, judge the spatial influence range of the lowest priority obstacle point, and optimize the exclusion parameter of the obstacle point in the grasping path to generate the goods path planning parameter table.

[0012] As a further solution of the present invention, the geometric feature description value of the goods includes the point cloud coordinates, normal vectors, curvature values, boundary sizes, and curvature distribution fitting relationships on the surface of the goods. The shape classification coefficient of the goods includes the curvature distribution, boundary feature position, spatial range of feature points, geometric center position, and distribution variance range. The stress distribution of the goods stacking includes the contact point pressure value, center of gravity offset range, stress balance state, contact point position, and pressure value range. The stacked energy optimization value of the goods includes the center of gravity offset adjustment parameter, contact point stacking matrix, stacked energy relationship value, and stacked energy optimization matrix relationship. The goods path planning parameter table includes the grasping point position, placing point position, spatial priority value of the grasping point, spatial priority value of the obstacle point, spatial influence range of the lowest priority obstacle point, and obstacle point exclusion parameter.

[0013] As a further solution of the present invention, the specific steps for obtaining the geometric feature description value of the goods are as follows:

[0014] S111: Extract the three-dimensional coordinates of each point in the point cloud data on the surface of the goods, vectorize the point cloud normal vector, analyze the change trend of the curvature values at each point by calculating the curvature value of each point, eliminate the abnormal points with too large curvature changes, extract the set of points with continuous distribution and smooth curvature in the point cloud, and generate the point cloud coordinates, normal vectors, and curvature values on the surface of the goods;

[0015] S112: Based on the point cloud coordinates and normal vectors on the surface of the goods, calculate the geometric angle difference between each point in the point cloud and its surrounding points, determine whether the geometric angle difference exceeds the set threshold, mark the points that meet the conditions as edge points, eliminate the points that do not meet the geometric conditions, and re-analyze the boundary range reorganization of the remaining edge points to generate the geometric segmentation range of the point cloud edge features;

[0016] S113: Based on the boundary points in the geometric segmentation range of the point cloud edge features, calculate the size distribution of the boundary points and fit the relationship of their curvature values. The formula is as follows:

[0017]

[0018] Through this formula, calculate the fitting relationship between the boundary size and the curvature value, and generate the multinomial distribution of the boundary size and curvature relationship;

[0019] Among them, C i represents the curvature value of the i-th boundary point, S i represents the size value of the i-th boundary point, N i represents the number of points in the curvature distribution range, W i represents the weight factor of each point in the curvature distribution, P represents the smoothing parameter, and f(x) represents the fitting function of the boundary size and curvature value;

[0020] S114: According to the multinomial distribution of the boundary size and curvature relationship, calculate the deviation value of the curvature distribution, screen out the abnormal data with the deviation value exceeding the fitting range, smooth the curvature distribution deviation value, and re-verify the geometric consistency to generate the geometric feature description value of the goods.

[0021] As a further solution of the present invention, the specific steps for obtaining the goods shape classification coefficient are as follows:

[0022] S211: Based on the geometric feature description value of the goods, extract the curvature distribution and boundary feature positions of the goods. By analyzing the change trend of the curvature values on the surface of the goods and combining the geometric distribution information of the boundary feature points, screen and extract the key points defining the boundary of the goods shape to generate the curvature distribution and boundary feature set;

[0023] S212: Using the curvature distribution and the boundary feature set, calculate the Euclidean distance from each feature point in the set to the geometric center of the goods. By constructing a geometric space model, analyze the spatial central tendency of the feature point distribution and the positions of outliers, and calibrate the spatial range of the feature points that meet the geometric conditions;

[0024] S213: According to the spatial range of the feature points, determine whether each feature point is within a reasonable geometric center position and distribution variance range. The formula is as follows:

[0025]

[0026] Through this formula, calculate the geometric distance from the point to the center and evaluate the distribution deviation, classify and process the set of feature points that meet the conditions, and obtain the classified set of feature points;

[0027] where d(x) represents the Euclidean distance from the point to the geometric center, x i represents the coordinates of the feature point, μ represents the geometric center position of the feature point set, σ represents the variance range of the feature point distribution, and n represents the total number of the feature point set;

[0028] S214: According to the classified set of feature points, analyze the statistical distribution characteristics, combine with the geometric distribution model, calculate the spatial distribution parameters of the feature point set category by category, establish the classification standard of the goods shape, and generate the goods shape classification coefficient.

[0029] As a further solution of the present invention, the steps for obtaining the stress distribution of the stacked goods are specifically as follows:

[0030] S311: Based on the goods shape classification coefficient, extract the centroid coordinates and the three-dimensional size range of the goods. By analyzing the shape characteristics of the goods unit item by item, calculate the geometric center position and the corresponding volume range, and eliminate the abnormally offset shape characteristic data to obtain the centroid coordinate and size data set;

[0031] S312: According to the centroid coordinate and size data set, calculate the centroid coordinate offset and size deviation of the goods unit. By comparing the offset data with the standard distribution of its corresponding geometric center, screen the offset characteristic points that match the center position, and eliminate the data with deviation values exceeding the reasonable range to generate the relationship data between the centroid coordinate offset and the size deviation;

[0032] S313: According to the relationship data between the centroid coordinate offset and the size deviation, evaluate the centroid offset distribution area of the goods in the stacked state. The formula is as follows:

[0033]

[0034] Calculate the offset of the cargo unit under the force balance state, judge the balance stability of the stack through the offset characteristic points, and adjust the contact point position and pressure value range;

[0035] Among them, d balance represents the offset measurement under the force balance state, x i , y i , z i respectively represent the centroid coordinates of the i-th cargo unit on the x, y, and z axes, represents the average value of the centroid coordinates of all cargo units, σ x , σ y , σ z respectively represent the standard deviations of the x, y, and z coordinates, which are used to describe the dispersion degree of the coordinate distribution, and n is the total number of cargo units involved in the calculation;

[0036] S314: Combine the contact point position and pressure value range, adjust the force position and contact area ratio of the contact point by measuring the pressure distribution multiple times, evaluate the stack stability after adjustment, and finally determine the force distribution model of the cargo stack and generate the force distribution of the cargo stack.

[0037] As a further solution of the present invention, the steps for obtaining the cargo stack energy optimization value are specifically as follows:

[0038] S411: Based on the force distribution of the cargo stack, extract the contact point pressure value and the centroid offset range, calculate the distribution characteristics and offset trend of the pressure value in the spatial coordinates by analyzing the correlation between the contact point pressure data and the offset value, and generate the contact point pressure and centroid offset data;

[0039] S412: According to the contact point pressure and centroid offset data, calculate the centroid offset adjustment parameter, calculate the relationship between the pressure change range and the offset value of each contact point based on the adjustment parameter, and establish a contact point stack matrix in combination with the geometric correlation between the contact points;

[0040] S413: Combine the contact point stack matrix, accumulate the stack energy relationship values of each contact point, calculate the total stack energy through iterative optimization based on the energy relationship values of each contact point, integrate the energy accumulation relationship and perform matrix optimization operations to generate a stack energy optimization matrix relationship;

[0041] S414: According to the stack energy optimization matrix relationship, calculate the total stack energy value of all cargoes by analyzing the distribution of the optimization values of all contact points in the stack energy matrix, and accumulate the optimized energy results. The formula is as follows:

[0042]

[0043] Calculate the total stack energy value of all cargoes and generate the cargo stack energy optimization value;

[0044] Among them, E optimized represents the final optimized energy value of the cargo stacking, and E i represents the initial energy value of the i-th contact point, and x i represents the centroid offset value of the i-th contact point, and x mean is the average value of the centroid offsets of all contact points. k is the adjustment intensity parameter that affects the weight of the offset on energy optimization, and n is the total number of all contact points.

[0045] As a further solution of the present invention, the obtaining steps of the cargo path planning parameter table are specifically as follows:

[0046] S511: Based on the optimized energy value of the cargo stacking, extract the three-dimensional position relationship between the grasping point and the placing point. By analyzing the spatial coordinates of the cargo units item by item, calculate the distance and angle from the grasping point to the placing point, and screen the coordinate characteristic points that affect the spatial connection to generate the position relationship data between the grasping point and the placing point;

[0047] S512: Use the position relationship data between the grasping point and the placing point to calculate the spatial priority value between the grasping point and the obstacle point. By analyzing the spatial distribution of each obstacle point, screen the key obstacle points according to the geometric characteristics of the influence of the obstacle point on the path, and eliminate the non-key obstacle points with low influence to generate the spatial priority sorting of the obstacle points;

[0048] S513: According to the spatial priority sorting of the obstacle points, judge the spatial influence range of the obstacle point with the lowest priority, and adopt the improved formula:

[0049]

[0050] Calculate the influence value of each obstacle point on the path to generate the obstacle point exclusion parameter for the optimized grasping path;

[0051] Among them, P impact represents the influence value of the obstacle point on the path, and w i represents the importance degree of the i-th obstacle point, and assigns a differentiated weight value according to the importance of the path. d i is the distance value between the obstacle point and the path, which is used to reflect the direct spatial influence degree of the obstacle point on the path. ∈ is a constant parameter used to avoid the denominator being zero when the obstacle point coincides with the path. k j represents the indirect influence adjustment coefficient of the secondary obstacle point on the path, which is used to measure the superimposed influence intensity of the associated obstacle points. d ij is the spatial distance between the i-th main obstacle point and the j-th secondary obstacle point, which is used to evaluate the indirect spatial effect of the relevant obstacle points. n is the total number of all main obstacle points in the path, representing the quantity range of the main obstacle points, and m is the sum of the quantities of the secondary obstacle points;

[0052] S514: Adjust the position arrangement of the obstacle points in the grasping path in combination with the optimized obstacle point exclusion parameters for the grasping path, calculate the connection path characteristics between the grasping point and the placing point, and summarize and generate a cargo path planning parameter table.

[0053] A fully automatic palletizing system based on machine vision and path planning, which is used to execute the above-mentioned fully automatic palletizing method based on machine vision and path planning. The system includes:

[0054] The cargo feature extraction module extracts the point cloud coordinates on the surface of the cargo based on the three-dimensional point cloud data of the cargo, calculates the normal vector and curvature value of the point cloud coordinates, determines the geometric segmentation range of the edge features, analyzes the fitting relationship between the boundary size and the curvature distribution, and generates a cargo geometric feature description value.

[0055] The cargo shape classification module extracts the curvature distribution and the position of the boundary features based on the cargo geometric feature description value, calculates the Euclidean distance space range of the feature points, judges the position of the geometric center, analyzes the distribution variance range, classifies the feature set that meets the center distance and variance range, and generates a cargo shape classification coefficient.

[0056] The stacking force analysis module extracts the center of gravity coordinates and the three-dimensional size range of the cargo based on the cargo shape classification coefficient, calculates the center of gravity offset value and the relationship between the size deviation, judges the distribution area of the center of gravity offset and the force balance state, adjusts the position and the pressure value range of the contact point, analyzes the mechanical distribution range of the contact point, and generates a cargo stacking force distribution.

[0057] The stacking energy optimization module extracts the pressure value of the contact point and the center of gravity offset range based on the cargo stacking force distribution, calculates the center of gravity offset adjustment parameter and the contact point stacking matrix, accumulates the energy relationship values after stacking, establishes a stacking energy optimization matrix relationship, and generates a cargo stacking energy optimization value.

[0058] The path planning module extracts the three-dimensional position relationship between the grasping point and the placing point based on the cargo stacking energy optimization value, calculates the spatial priority value between the grasping point and the obstacle point, analyzes the spatial influence range of the lowest priority obstacle point, optimizes the obstacle point exclusion parameters in the grasping path, and generates a cargo path planning parameter table.

[0059] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0060] In the present invention, surface point cloud coordinates, normal vectors, and curvature values are extracted based on three-dimensional point cloud data, edge feature segmentation and boundary size fitting analysis are refined, and the accuracy of geometric feature description of complex goods surfaces is enhanced. Through spatial distribution analysis and feature set classification, the geometric center and variance range are reasonably determined to improve the accuracy and stability of shape classification. Through the calculation of the centroid and size deviation, the contact point position and pressure distribution are optimized to ensure stacking stability and balanced force. By combining the contact point stacking matrix and the energy optimization matrix, the stacking operation efficiency and resource utilization rate are improved. By combining the spatial priority analysis of the grasping point and the obstacle point, the path planning and obstacle point exclusion strategy are optimized to reduce path conflicts and improve the flexibility and efficiency of logistics operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 is a schematic diagram of the workflow of the present invention;

[0062] Figure 2 is a flowchart of the steps for obtaining the geometric feature description values of the goods of the present invention;

[0063] Figure 3 is a flowchart of the steps for obtaining the shape classification coefficient of the goods of the present invention;

[0064] Figure 4 is a flowchart of the steps for obtaining the stacking force distribution of the goods of the present invention;

[0065] Figure 5 is a flowchart of the steps for obtaining the stacking energy optimization value of the goods of the present invention;

[0066] Figure 6 is a flowchart of the steps for obtaining the path planning parameter table of the goods of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0067] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0068] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the present invention. In addition, in the description of the present invention, "a plurality of" means two or more, unless otherwise specifically defined.

[0069] Example 1

[0070] Please refer to Figure 1 , the present invention provides a technical solution: a full-automatic palletizing method based on machine vision and path planning, including the following steps:

[0071] S1: Based on the three-dimensional point cloud data of the goods, extract the point cloud coordinates, normal vectors and curvature values on the surface of the goods, calculate the geometric segmentation range of the point cloud edge features, calculate the fitting relationship between the boundary size and the curvature distribution, and use the multinomial distribution of the relationship between the boundary size and the curvature for verification to generate the geometric feature description value of the goods;

[0072] S2: Based on the geometric feature description value of the goods, extract the curvature distribution and the boundary feature position of the goods, calculate the spatial range of the feature points of the Euclidean distance, judge the geometric center position and the distribution variance range, classify and process the feature sets that meet the center distance and variance range, and generate the shape classification coefficient of the goods;

[0073] S3: Based on the shape classification coefficient of the goods, extract the center of gravity coordinates and the three-dimensional size range of the goods, calculate the relationship between the offset of the center of gravity coordinates of the goods and the size deviation, judge the distribution area of the stacking center of gravity offset and the stress balance state, adjust the contact point position and the pressure value range, and generate the stacking stress distribution of the goods;

[0074] S4: Based on the stacking stress distribution of the goods, extract the contact point pressure value and the center of gravity offset range, calculate the center of gravity offset adjustment parameter and the contact point stacking matrix, accumulate all the energy relationship values after stacking, establish the stacking energy optimization matrix relationship of all goods, and generate the stacking energy optimization value of the goods;

[0075] S5: Based on the stacking energy optimization value of the goods, extract the three-dimensional position relationship between the grasping point and the placing point, calculate the spatial priority value of the grasping point and the obstacle point, judge the spatial influence range of the lowest priority obstacle point, and optimize the exclusion parameter of the obstacle point in the grasping path to generate the goods path planning parameter table.

[0076] The geometric feature description value of the goods includes the point cloud coordinates, normal vector, curvature value, boundary size, curvature distribution fitting relationship on the surface of the goods. The shape classification coefficient of the goods includes the curvature distribution, boundary feature position, spatial range of feature points, geometric center position, distribution variance range. The stacking stress distribution of the goods includes the contact point pressure value, center of gravity offset range, stress balance state, contact point position, pressure value range. The stacking energy optimization value of the goods includes the center of gravity offset adjustment parameter, contact point stacking matrix, energy relationship value after stacking, stacking energy optimization matrix relationship. The goods path planning parameter table includes the grasping point position, placing point position, spatial priority value of the grasping point, spatial priority value of the obstacle point, spatial influence range of the lowest priority obstacle point, obstacle point exclusion parameter.

[0077] Please refer toFigure 2 , the steps for obtaining the geometric feature description values of the goods are specifically as follows:

[0078] S111: Extract the three-dimensional coordinates of each point in the point cloud data on the surface of the goods, vectorize the point cloud normal vector, calculate the curvature value of each point, analyze the change trend of the curvature values in the point cloud point by point, eliminate the abnormal points with too large curvature changes, extract the set of points with continuous distribution and smooth curvature in the point cloud, and generate the point cloud coordinates, normal vectors and curvature values on the surface of the goods;

[0079] Extract the three-dimensional coordinates of each point in the point cloud data on the surface of the goods, vectorize the point cloud normal vector. First, it is necessary to obtain the original point cloud data set of the goods through a scanner or three-dimensional acquisition device, including the three-dimensional coordinates and normal vector information on the surface of the goods, clean the data through the point cloud preprocessing method, eliminate the abnormal points caused by noise or environmental interference, then calculate the normal direction of each point in the point cloud using the normal vector, calculate the curvature value of each point, and the curvature value can be calculated through the geometric relationship between the neighboring points around the point. Specifically, by selecting the set of neighboring points within a fixed radius range, using the least squares method to fit the surface equation of the points in the neighborhood, and then calculating the curvature value based on the second derivative of the surface equation. Store the calculated curvature values point by point into the curvature distribution data set, analyze the distribution characteristics of the curvature values and eliminate the points with abnormal curvature changes. For example, when the curvature value of a certain point deviates from the curvature values of the neighboring points by more than the set threshold, this point can be considered an abnormal point and eliminated. Re-analyze the remaining curvature values to ensure the continuity and smoothness of their distribution. Finally, extract the set of points with continuous distribution and smooth curvature in the point cloud, and generate the point cloud coordinates, normal vectors and curvature values on the surface of the goods.

[0080] S112: Based on the point cloud coordinates and normal vectors on the surface of the goods, calculate the geometric angle difference between each point in the point cloud and its surrounding points, determine whether the geometric angle difference exceeds the set threshold, mark the points that meet the conditions as edge points, eliminate the points that do not meet the geometric conditions, and re-analyze the boundary range recombination of the remaining edge points to generate the geometric segmentation range of the point cloud edge features;

[0081] Based on the point cloud coordinates and normal vectors of the cargo surface, the geometric angle difference between each point in the point cloud and the surrounding points is calculated. First, the normal vector data of the point cloud is used to calculate the normal angle difference between each point and the surrounding points. The normal angle difference can be calculated by the dot product formula of the point normal vector. When the normal angle difference is greater than the set threshold, the point is marked as an edge point. In order to screen the geometric distribution range of edge points, it is necessary to eliminate points that do not meet the geometric conditions, such as calculating the geometric relationship between the edge point and its neighboring points, including the distance and curvature difference. When the distance from the point to the neighboring point exceeds a certain set range or the curvature difference exceeds the set range, the point is not considered to be a valid edge point and is eliminated. Among the retained edge points, the boundary range reorganization process is further analyzed. By clustering the distribution of edge points or dividing the regions, the densely distributed edge points are classified into the same region to obtain a clearer geometric segmentation range, and finally generate the geometric segmentation range of the point cloud edge feature.

[0082] S113: Based on the boundary points in the geometric segmentation range of the point cloud edge features, calculate the size distribution of the boundary points and fit their curvature value relationship. The formula is as follows:

[0083]

[0084] Through this formula, the fitting relationship between the boundary size and the curvature value is calculated to generate a multinomial distribution of the relationship between the boundary size and the curvature;

[0085] Among them, C i represents the curvature value of the i-th boundary point, S i represents the size value of the i-th boundary point, N i represents the number of points in the curvature distribution range, W i represents the weight factor of each point in the curvature distribution, P represents the smoothing parameter, and f(x) represents the fitting function of the boundary size and curvature value;

[0086] Formula:

[0087]

[0088] The benefit of the formula is that by introducing the weight parameter W i and smoothing parameter P, optimize the fitting relationship between boundary size and curvature distribution, and improve the stability and accuracy of curvature calculation in the boundary area.

[0089] Detailed explanation of the formula and the process of formula calculation and derivation:

[0090] Extract the curvature value C of the boundary point i , size value S i , the number of neighboring points N i , calculate N by aggregating neighborhood points based on point cloud data i , using the surface fitting method to get C​​i , calculate S using the Euclidean distance or density clustering method i , weight parameter W i is set according to the importance of the region to which the boundary point belongs, and the smoothing parameter P is used to adjust the stability of the boundary point.

[0091] Substitute data example: Suppose there are five boundary points, curvature value C 1 = 0.4, C 2 = 0.5, C 3 = 0.6, C 4 = 0.3, C 5 = 0.7, size value is S 1 = 1.2, S 2 = 1.5, S 3 = 1.1, S 4 = 1.0, S 5 = 1.3, number of neighborhood points is N 1 = 10, N 2 = 12, N 3 = 15, N 4 = 8, N 5 = 11, weight parameter is W 1 = 0.9, W 2 = 1.0, W 3 = 0.8, W 4 = 1.1, W 5 = 1.2, smoothing parameter P = 5.

[0092] Calculate step by step:

[0093] First point calculation:

[0094] Second point calculation:

[0095] Third point calculation:

[0096] Fourth point calculation:

[0097] Fifth point calculation:

[0098] Sum to calculate the fitting value:

[0099]

[0100] The result shows that the fitting curvature distribution value is 0.7761, indicating that the distribution relationship between curvature and boundary size tends to be smooth and reasonable, and can be used for the generation and verification calculations of subsequent multiple distributions.

[0101] S114: Calculate the deviation value of the curvature distribution according to the multinomial distribution of the relationship between the boundary size and the curvature, screen out the abnormal data whose deviation value exceeds the fitting range, smooth the deviation value of the curvature distribution, re-verify the geometric consistency, and generate the geometric feature description value of the goods.

[0102] According to the multinomial distribution of the relationship between the boundary size and the curvature, calculate the deviation value of the curvature distribution. First, extract the curvature value data from the boundary point set, and calculate its deviation value in combination with the multinomial distribution. The calculation of the deviation value can be carried out by summing and normalizing the absolute deviation between the fitted curvature value and the actual curvature value. Specifically, it is calculated by the formula Obtained, screen out the abnormal data whose deviation value exceeds the fitting range. The screening condition for abnormal data is that the deviation value is greater than the set threshold or a certain fixed ratio value. Then, smooth the deviation value of the curvature distribution. For example, use the moving average method to smooth the deviation value sequence, and take the mean value within the moving window as the adjusted value of the deviation value sequence. When re-verifying the geometric consistency, by reconstructing the curvature distribution data of the geometric boundary and comparing it with the original curvature distribution point by point, ensure that the geometric boundary consistency meets the fitting accuracy requirements, and finally generate the geometric feature description value of the goods.

[0103] Please refer to Figure 3 , and the steps for obtaining the goods shape classification coefficient are specifically as follows:

[0104] S211: Based on the geometric feature description value of the goods, extract the curvature distribution and boundary feature positions of the goods. By analyzing the change trend of the curvature values on the surface of the goods and combining the geometric distribution information of the boundary feature points, screen and extract the key points that define the goods shape boundary, and generate the curvature distribution and boundary feature set;

[0105] Based on the geometric feature description value of the goods, analyze the curvature distribution of the goods. By discretizing the acquisition of curvature values, discretize the curvature values of continuous points into curvature amplitude intervals, then extract the curvature segment with the smallest change within the amplitude interval and match it with the geometric position of the boundary feature points. By comparing the normal vector of the boundary feature points with the gradient direction of the curvature segment, screen out the set of boundary feature points that conform to the curvature gradient change direction, and finally combine the spatial distribution pattern of the geometric positions to extract the key points used to define the goods shape boundary, and generate the curvature distribution and boundary feature set;

[0106] S212: Use the curvature distribution and boundary feature set to calculate the Euclidean distance from each feature point in the set to the geometric center of the goods. By constructing a geometric space model, analyze the spatial central tendency of the feature point distribution and the position of outliers, and calibrate the spatial range of the feature points that meet the geometric conditions;

[0107] Using the curvature distribution and boundary feature set, first, through the initial selection of the geometric center, calculate the Euclidean distance from each feature point in the curvature distribution to the geometric center, and express it with the formula Calculate it, and thus generate a preliminary spatial distribution model of all feature points. Then, count the set of points whose distance values are higher than the third quartile, determine whether these points form outliers, and then recalculate the geometric center coordinates of the overall point set and adjust the initial point set. By introducing a central offset correction factor, gradually decrease the spatial influence weight of the outliers, and correct the set according to the adjusted distribution weight, calibrate the range of feature points that meet the geometric conditions, and generate the spatial range of feature points;

[0108] S213: According to the spatial range of feature points, judge whether each feature point is within the reasonable geometric center position and distribution variance range. The formula is as follows:

[0109]

[0110] Through this formula, calculate the geometric distance from the point to the center and evaluate the distribution deviation, classify and process the set of feature points that meet the conditions, and obtain the classified set of feature points;

[0111] Among them, d(x) represents the Euclidean distance from the point to the geometric center, x i represents the coordinates of the feature point, μ represents the geometric center position of the feature point set, σ represents the variance range of the feature point distribution, and n represents the total number of the feature point set;

[0112] Formula:

[0113]

[0114] The benefit of the formula is that by combining the geometric center and the distribution variance parameter, it can not only analyze the spatial offset situation of the feature points, but also comprehensively evaluate the dispersion of the feature points in the set by introducing the distribution variance parameter, avoiding the offset misjudgment caused by the simple Euclidean distance;

[0115] Detailed explanation of the formula and the derivation process of formula calculation:

[0116] According to the spatial range of feature points, determine the geometric center coordinates of the feature point set Calculate the distance value from each point to the center. For example, assume n = 5, and the coordinates of the feature point set are x 1 = 1, x 2 = 3, x 3 = 5, x 4 = 2, x 5 = 4, and the geometric center coordinate is μ = (1 + 3 + 5 + 2 + 4) / 5 = 3. Then substitute it into the formula for calculation:

[0117]

[0118] Expand calculation:

[0119]

[0120] The results show that the average offset value of the feature point set to the center is 2.02. By introducing the distribution variance parameter, the geometric distribution state of the feature points can be analyzed more accurately, providing a basis for classification processing;

[0121] S214: According to the classified feature point set, analyze the statistical distribution characteristics, combine the geometric distribution model, calculate the spatial distribution parameters of the feature point set by category, establish the classification standard of the cargo shape, and generate the cargo shape classification coefficient.

[0122] Based on the classified feature point set, by analyzing the spatial distribution pattern of each type of feature point in the set, the point set is spatially enveloping and calculated by the minimum envelope volume. For the distribution center of gravity in the set, the core points that affect the distribution shape are separated by the weight distribution strategy, and the geometric arrangement of the core points is spatially analyzed. By comparing the results of the analysis with the curvature range standard, the classification standards for different cargo shapes are determined, and finally the cargo shape classification coefficient is generated.

[0123] Please refer to Figure 4 , the specific steps for obtaining the force distribution of cargo stacking are:

[0124] S311: Based on the cargo shape classification coefficient, extract the center of gravity coordinates and three-dimensional size range of the cargo, analyze the shape characteristics of the cargo unit item by item, calculate the geometric center position and corresponding volume range, eliminate the abnormally offset shape characteristic data, and obtain the center of gravity coordinates and size data set;

[0125] Based on the cargo shape classification coefficient, the center of gravity coordinates and three-dimensional size range of the cargo are extracted. By analyzing the shape characteristics of the cargo unit item by item, the cargo is divided into multiple structural units. The external contour points of each unit are extracted to determine the boundary coordinates. The maximum and minimum values ​​of the three-dimensional range of the cargo are calculated through the boundary coordinate data to obtain the cargo size range. Combined with the geometric center calculation formula of the boundary coordinate points, the center of gravity coordinates of each cargo unit are obtained by calculating the average coordinate value. Based on the center of gravity coordinate data, the deviation of each geometric point in each unit is judged, and abnormal data with an offset exceeding three times the standard deviation is eliminated. Finally, the center of gravity coordinates and size range of each unit are integrated into a set to obtain the center of gravity coordinate and size data set.

[0126] S312: Calculate the centroid coordinate offset and dimensional deviation of the cargo unit based on the centroid coordinates and the dimensional data set. By comparing the offset data with the standard distribution of its corresponding geometric center, screen the offset characteristic points that match the center position, eliminate the data with deviation values exceeding the reasonable range, and generate the relationship data between the centroid coordinate offset and the dimensional deviation.

[0127] Based on the centroid coordinates and the dimensional data set, calculate the centroid coordinate offset and dimensional deviation of the cargo unit. Calculate the position offset of each unit of the cargo relative to the overall geometric center. Use the offset formula (x i -μ x , y i -μ y , z i -μ z ) to calculate the three-dimensional offset of each unit. Combine the actual dimensions with the ideal dimension range to analyze the possible dimensional errors during the assembly or stacking process of each cargo unit. Map all offset values to relative ratio values to standardize the data. For the calculation results, screen and eliminate the units with dimensional deviation values exceeding twice the standard deviation. Integrate the offset and dimensional data of the remaining units to form a complete offset and dimensional deviation distribution, and finally generate the relationship data between the centroid coordinate offset and the dimensional deviation.

[0128] S313: Based on the relationship data between the centroid coordinate offset and the dimensional deviation, evaluate the distribution area of the centroid offset of the cargo in the stacked state. The formula is as follows:

[0129]

[0130] Calculate the offset of the cargo unit under the force balance state. Judge the balance stability of the stack through the offset characteristic points, and adjust the contact point position and the pressure value range.

[0131] Among them, d balance represents the offset measurement under the force balance state, x i , y i , z i respectively represent the centroid coordinates of the i-th cargo unit on the x, y, and z axes. represents the average value of the centroid coordinates of all cargo units, σ x , σ y , σ z respectively represent the standard deviations of the x, y, and z coordinates, which are used to describe the dispersion degree of the coordinate distribution. n is the total number of cargo units involved in the calculation.

[0132] Formula:

[0133]

[0134] The advantage of the formula is that by comprehensively calculating the normalized value of the three-dimensional offset of the cargo unit and its distribution variance, the stacking stability and force balance state of the cargo can be effectively analyzed to guide the adjustment of the contact point position and the optimization of the pressure range;

[0135] Detailed explanation of the formula and the derivation process of the formula calculation:

[0136] Suppose there are 5 cargo units, and the x, y, and z coordinate values are as follows: Unit 1 (10, 20, 30), Unit 2 (12, 18, 29), Unit 3 (11, 19, 31), Unit 4 (13, 22, 28), Unit 5 (9, 21, 30); First, calculate the mean value of each dimension

[0137]

[0138] Then calculate the standard deviation of each dimension:

[0139]

[0140] σ z = 1.20;

[0141] Then substitute into the formula to calculate the normalized offset of each cargo unit:

[0142]

[0143] For Unit 1:

[0144]

[0145] Calculate successively for Unit 2 to Unit 5, and finally obtain d balance = 2.73;

[0146] This result indicates that the overall offset of the cargo unit meets the stacking stability requirements, the offset distribution is small, and it is suitable for contact point adjustment and pressure distribution optimization.

[0147] S314: Combine the contact point position and the pressure value range, measure the pressure distribution multiple times, adjust the force position and contact area ratio of the contact point, evaluate the stacking stability after adjustment, and finally determine the force distribution model of the cargo stack to generate the force distribution of the cargo stack.

[0148] Combine the contact point position and the pressure value range, adjust the contact point position of the cargo unit successively, minimize the distance between its contact point and the bottom support point to ensure stability, calculate the pressure distribution for the adjusted contact point position to solve the reasonable range of the pressure value, and use the pressure formula Analyze the forces acting on each contact point, calculate the supporting force F based on the actual contact area of the cargo unit, and verify whether the pressure range is within the safe range. Adjust the contact points where the pressure values exceed the range, and finally re-evaluate the force distribution of the cargo stacking state to generate the force distribution of the cargo stacking.

[0149] Please refer to Figure 5 , the steps for obtaining the optimized value of the cargo stacking energy are specifically as follows:

[0150] S411: Based on the force distribution of the cargo stacking, extract the pressure values of the contact points and the range of the center-of-gravity offset. By analyzing the correlation between the pressure data of the contact points and the offset values, calculate the distribution characteristics of the pressure values in the spatial coordinates and their offset trends, and generate the contact point pressure and center-of-gravity offset data;

[0151] Based on the force distribution of the cargo stacking, first, it is necessary to obtain the pressure value data of each contact point during the stacking process of the cargo. This can be achieved by monitoring the pressure distribution of the contact points using stress sensors, collecting the range of the force changes at the contact points using equipment, and performing data cleaning and statistics through software tools. The range of the center-of-gravity offset is obtained by setting up a three-dimensional coordinate system and combining measuring instruments to obtain the offset position range of the center of gravity of the cargo in the horizontal and vertical directions. These offset data are input into the statistical model for analysis through a calculation program, correlating the pressure values of the contact points and the offset values of the cargo, and obtaining the relationship model between the two through data fitting analysis. For example, a multiple linear regression formula y = a·P + b·ΔG + c (where y is the offset trend, P is the pressure value of the contact point, and ΔG is the offset value) can be used. Through regression analysis, the correlation between the pressure value and the offset value is obtained, and then a data graph of the pressure distribution characteristics is generated using software, and the distribution trend in the spatial coordinates is calculated through a calculation formula, and finally the contact point pressure and center-of-gravity offset data are generated.

[0152] S412: According to the contact point pressure and center-of-gravity offset data, calculate the center-of-gravity offset adjustment parameters, calculate the relationship between the pressure change range and the offset value of each contact point based on the adjustment parameters, and establish a contact point stacking matrix in combination with the geometric correlation between the contact points;

[0153] According to the contact point pressure and center-of-gravity offset data, first, it is necessary to calculate the center-of-gravity offset adjustment parameters. The center-of-gravity offset adjustment parameters can be calculated by analyzing the offset degree of the cargo between each stacking layer. This process can use a matrix analysis tool to extract the geometric position relationship of the contact points and calculate the displacement amount between the contact points, and further analyze the offset trend in combination with the pressure change range. For example, the range of the offset amount change of each contact point can be calculated through the formula (where x 1 , x 2 , y 1 , y 2(i.e., the coordinate values of the two contact points), an association matrix is established between the offset value of each contact point and the pressure change range. The geometric correlation between the contact points is analyzed through software, and the geometric correlation matrix is optimized into a contact point stacking matrix that can be used for the stacking relationship. The stacking matrix reflects the mutual force and displacement trend of each contact point, providing basic data for optimizing the stacking energy in the next step.

[0154] S413: Combining the contact point stacking matrix, accumulate the stacking energy relationship values of each contact point. Based on the energy relationship values of each contact point, calculate the total stacking energy through iterative optimization, integrate the energy accumulation relationship, and perform matrix optimization operations to generate a stacking energy optimization matrix relationship.

[0155] Combined with the contact point stacking matrix, it is necessary to perform an accumulation analysis on the stacking energy relationship values of each contact point. First, according to the mechanical parameters in the contact point stacking matrix, use the iterative calculation method to optimize the total stacking energy. The calculation of the stacking energy analyzes the force state of each contact point through the mechanical model formula, and uses the matrix accumulation method to integrate the energy relationship values of each contact point into a stacking energy optimization matrix relationship. The optimization operation of this matrix is completed through the matrix sparse optimization algorithm. For example, the sparse algorithm can optimize the energy value distribution through the L1 regularization constraint term. The matrix relationship can reflect the overall trend of the stacking force between the contact points, and further accumulate the stacking energy to obtain the optimization result.

[0156] S414: According to the stacking energy optimization matrix relationship, by analyzing the optimized value distribution of all contact points in the stacking energy matrix, accumulate the optimized energy results. The formula is as follows:

[0157]

[0158] Calculate the total stacking energy value of all goods to generate the optimized value of the goods stacking energy.

[0159] Among them, E optimized represents the final optimized energy value of the goods stacking, E i represents the initial energy value of the i-th contact point, x i represents the centroid offset value of the i-th contact point, x mean is the mean value of the centroid offsets of all contact points, k is the adjustment intensity parameter affecting the weight of the offset on the energy optimization, and n is the total number of all contact points.

[0160] Formula:

[0161]

[0162] The advantage of the formula is that by introducing the distribution characteristics of the centroid offset value and the weight influence of the offset mean value, it can effectively adjust the optimization effect of the offset trend on the stacking energy, thus realizing a more reasonable calculation of the energy distribution.

[0163] Detailed Explanation of the Formula and the Deduction Process of Formula Calculation:

[0164] Parameter Description: E optimized represents the total optimized stacking energy value, E i represents the initial energy value of the i-th contact point, which is collected by the sensor and calculated through mechanical formulas, x i represents the centroid offset value of the i-th contact point, which comes from the offset measuring instrument, x mean represents the average value of the centroid offsets of all contact points, which can be calculated through the formula The calculation is as follows. k represents the adjustment intensity parameter, which is used to reflect the weight influence of the offset value on the optimized energy. n represents the total number of contact points, which is determined by the dimension of the stacking matrix. e represents the base of the natural logarithm.

[0165] Parameter Acquisition:

[0166] E i : The data is obtained by monitoring the pressure distribution of the contact points and calculated by combining the mechanical formula E = F·d (where F is the force and d is the displacement).

[0167] x i : The centroid offset value of each contact point is directly measured through a three-dimensional coordinate system.

[0168] x mean : The average value is obtained by statistically calculating the offset values of each contact point.

[0169] k: It is determined by the adjustment requirements of the actual cargo stacking. The reasonable range can be evaluated through adjustment simulation experiments, for example, between 0.5 and 2.0.

[0170] n: It is directly determined according to the number of contact points in the stacking matrix.

[0171] Actual Example:

[0172] Suppose the cargo stacking contains 4 contact points, and the initial energy value E i = [10, 15, 12, 8], the offset value x i = [1.2, 1.5, 1.0, 0.8], and the adjustment coefficient k = 1.2.

[0173] Calculate the average offset:

[0174] Substitute into the formula and calculate term by term:

[0175]

[0176] The total optimized energy value: E optimized = 9.51 + 14.43 + 11.22 + 7.25 = 42.41;

[0177] The result shows that the optimized stacking energy value calculated by the optimized formula is 42.41, which is more balanced after weight adjustment compared with the sum of the initial energy values (45). The energy distribution tends to be reasonable, improving the stacking stability.

[0178] Please refer to Figure 6 , and the specific steps for obtaining the cargo path planning parameter table are as follows:

[0179] S511: Based on the optimized stacking energy value of the cargo, extract the three-dimensional position relationship between the grasping point and the placing point. By analyzing the spatial coordinates of each cargo unit item by item, calculate the distance and angle from the grasping point to the placing point, screen the coordinate characteristic points affecting the spatial connection, and generate the position relationship data between the grasping point and the placing point;

[0180] Based on the optimized stacking energy value of the cargo, extract the three-dimensional position relationship between the grasping point and the placing point. By analyzing the spatial coordinates of each cargo unit item by item, split the three-dimensional coordinates of each cargo unit into the grasping point coordinates (x g , y g , z g ) and the placing point coordinates (x p , y p , z p ). Combining the offset of the center of gravity of the cargo, calculate the Euclidean distance between the grasping point and the placing point. The calculation formula is

[0181] where each parameter is collected by the cargo unit positioning system. For example, if the grasping point coordinates of a certain cargo unit are (1.5, 2.0, 1.0) and the placing point coordinates are (3.0, 5.0, 2.5), substituting into the formula gives By calculating the relative spatial angle of the cargo unit and combining the center of gravity adjustment parameters (such as the angle offset of the stacked cargo), further screen the cargo units that do not meet the smooth grasping conditions. For example, when the grasping point coordinates deviate from the geometric center of the cargo by more than a specific threshold (such as a 3° deviation range), mark and eliminate the data that does not meet the conditions, and screen the coordinates of the grasping points and placing points that meet the requirements to generate the position relationship data between the grasping point and the placing point;

[0182] S512: Using the position relationship data between the grasping point and the placing point, calculate the spatial priority value between the grasping point and the obstacle point. By analyzing the spatial distribution of each obstacle point, screen the key obstacle points according to the geometric characteristics of the influence of the obstacle points on the path, eliminate the non-key obstacle points with low influence, and generate the spatial priority ranking of the obstacle points;

[0183] Using the position relationship data of the grasping point and the placing point, calculate the spatial priority value between the grasping point and the obstacle point, and mark the three-dimensional coordinates of the obstacle point as (x o , y o , z o ). By analyzing the spatial distance between each obstacle point and the grasping path, and combining the geometric characteristics of the influence of the obstacle point on the path, calculate its minimum distance to the grasping path. The formula is where a, b, c, d are the plane equation parameters of the grasping path. For example, if the plane equation of the grasping path is 2x + 3y - z + 4 = 0 and the obstacle point coordinates are (1, 1, 1), then By performing a priority sorting on the minimum distance values of all obstacle points, define the obstacle points with larger distance values as lower-priority points, and define the points with smaller and densely distributed distance values as higher-priority points. Through the gradual screening of the degree of influence of the obstacle points on the path, eliminate the data of the obstacle points with lower priority and smaller influence, and generate the spatial priority sorting of the obstacle points;

[0184] S513: According to the spatial priority sorting of the obstacle points, judge the spatial influence range of the lowest-priority obstacle point, and adopt the improved formula:

[0185]

[0186] Calculate the influence value of each obstacle point on the path, and generate the obstacle point exclusion parameter for the optimized grasping path;

[0187] where P impact represents the influence value of the obstacle point on the path, w i represents the importance degree of the i-th obstacle point, and different weight values are assigned according to the importance to the path. d i is the distance value between the obstacle point and the path, which is used to reflect the direct spatial influence degree of the obstacle point on the path. ∈ is a constant parameter used to avoid the denominator being zero when the obstacle point coincides with the path. k j represents the indirect influence adjustment coefficient of the secondary obstacle point on the path, which is used to measure the superposition influence intensity of the associated obstacle points. d ij is the spatial distance between the i-th main obstacle point and the j-th secondary obstacle point, which is used to evaluate the indirect spatial effect of the relevant obstacle points. n is the total number of all main obstacle points in the path, representing the number range of the main obstacle points, and m is the sum of the number of secondary obstacle points, representing the complexity of the association network between the main obstacle points and the secondary obstacle points;

[0188] Formula:

[0189]

[0190] The advantage of the formula is that by introducing the influence parameters of the main obstacle points and secondary obstacle points, combining the non-linear calculation relationship between the weight value and the distance value, it comprehensively evaluates the direct and indirect influences of the obstacle points on the path, which helps to more accurately optimize the obstacle exclusion strategy in path planning;

[0191] Detailed explanation of the formula and the derivation process of formula calculation:

[0192] Select the data of the obstacle points (x o , y o , z o ) and the data of the grasping path plane. According to the minimum distance value in the previous formula, calculate the distance d from each obstacle point to the path i , for example, the minimum distance of the main obstacle point is d i = 2.5, the weight value w i is an important parameter for path influence, which is obtained through evaluation based on the spatial distribution characteristics of the obstacle points and the path. The influence coefficient k j of the secondary obstacle point is calculated from the superposition characteristics of the influence of the secondary obstacle point on the path, and the specific value is determined by monitoring the interaction parameters between the obstacle points. For example, the weight value w i of the main obstacle point = 1.2, the influence coefficient k j of the secondary obstacle point = 0.8, the distance d ij from the secondary obstacle point to the main obstacle point = 3.0, and the smoothing parameter ∈ = 0.01. Substituting these values into the formula, we get:

[0193]

[0194] The result shows that the influence value of the obstacle point on the path is 1.226, which belongs to a relatively high range of influence degree. Therefore, it is necessary to further optimize the obstacle exclusion parameters to reduce its obstruction to the grasping path;

[0195] S514: Combine the optimized grasping path obstacle exclusion parameters, adjust the position arrangement of the obstacle points in the grasping path, calculate the connection path characteristics between the grasping point and the placement point, and summarize and generate the cargo path planning parameter table.

[0196] Combine the optimized grasping path obstacle exclusion parameters, adjust the position arrangement of the obstacle points in the grasping path, calculate the optimal angle between the distribution coordinates of the obstacle points and the connection of the path through the optimization model, and calculate the connection angle between the projection point of the obstacle point on the path and the grasping point to the placement point. The formula is where is the vector from the grasping point to the projection point, is the vector from the grasping point to the placement point. For example, if the grasping point is (0, 0, 0), the obstacle point projection is (2, 2, 2), and the placement point is (4, 4, 4), then Substitute into the formula By adjusting the position of the projection point, the angle value approaches the optimal angle range (e.g., within 5°), and finally a cargo path planning parameter table is generated.

[0197] A full-automatic palletizing system based on machine vision and path planning, which is used to execute the above-mentioned full-automatic palletizing method based on machine vision and path planning. The system includes:

[0198] The cargo feature extraction module extracts the surface point cloud coordinates of the cargo based on the three-dimensional point cloud data of the cargo, calculates the normal vector and curvature value of the point cloud coordinates, determines the geometric segmentation range of the edge features, analyzes the fitting relationship between the boundary dimensions and curvature distribution, and generates the cargo geometric feature description value;

[0199] The cargo shape classification module extracts the curvature distribution and the position of the boundary features based on the cargo geometric feature description value, calculates the Euclidean distance space range of the feature points, judges the position of the geometric center, analyzes the distribution variance range, classifies the feature set that meets the center distance and variance range, and generates the cargo shape classification coefficient;

[0200] The stacking force analysis module extracts the center of gravity coordinates and the three-dimensional size range of the cargo based on the cargo shape classification coefficient, calculates the center of gravity offset value and the relationship between the size deviation, judges the distribution area of the center of gravity offset and the force balance state, adjusts the position of the contact point and the pressure value range, analyzes the mechanical distribution range of the contact point, and generates the cargo stacking force distribution;

[0201] The stacking energy optimization module extracts the pressure value of the contact point and the center of gravity offset range based on the cargo stacking force distribution, calculates the center of gravity offset adjustment parameter and the contact point stacking matrix, accumulates the energy relationship values after stacking, establishes the stacking energy optimization matrix relationship, and generates the cargo stacking energy optimization value;

[0202] The path planning module extracts the three-dimensional position relationship between the grasping point and the placing point based on the cargo stacking energy optimization value, calculates the spatial priority value between the grasping point and the obstacle point, analyzes the spatial influence range of the lowest priority obstacle point, optimizes the obstacle point exclusion parameter in the grasping path, and generates the cargo path planning parameter table.

[0203] The above is only the preferred embodiment of the present invention, and it is not intended to limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A fully automatic palletizing method based on machine vision and path planning, characterized in that: The following steps are involved: S1: Based on the three-dimensional point cloud data of the cargo, the point cloud coordinates, normal vector and curvature value of the cargo surface are extracted, the geometric segmentation range of the point cloud edge features is calculated, the fitting relationship between the boundary size and the curvature distribution is calculated, and the multinomial distribution of the relationship between the boundary size and the curvature is used for verification to generate the cargo geometric feature description value; S2: Based on the cargo geometric feature description value, extract the cargo curvature distribution and boundary feature position, calculate the feature point spatial range of the Euclidean distance, determine the geometric center position and distribution variance range, classify and process the feature set that meets the center distance and variance range, and generate the cargo shape classification coefficient; S3: Based on the cargo shape classification coefficient, the cargo center of gravity coordinates and three-dimensional size range are extracted, the relationship between the cargo center of gravity coordinate offset and the size deviation is calculated, the distribution area of ​​the stacking center of gravity offset and the force balance state are determined, the contact point position and the pressure value range are adjusted, and the cargo stacking force distribution is generated; S4: based on the cargo stacking force distribution, extract the contact point pressure value and the center of gravity offset range, calculate the center of gravity offset adjustment parameter and the contact point stacking matrix, accumulate all the energy relationship values ​​after stacking, establish the stacking energy optimization matrix relationship of all the cargoes, and generate the cargo stacking energy optimization value; S5: Based on the cargo stacking energy optimization value, extract the three-dimensional position relationship between the grabbing point and the placement point, calculate the spatial priority value of the grabbing point and the obstacle point, determine the spatial influence range of the lowest priority obstacle point, optimize the obstacle point exclusion parameters in the grabbing path, and generate a cargo path planning parameter table; The cargo geometric feature description value includes the cargo surface point cloud coordinates, normal vector, curvature value, boundary size, and curvature distribution fitting relationship; the cargo shape classification coefficient includes curvature distribution, boundary feature position, feature point spatial range, geometric center position, and distribution variance range; the cargo stacking force distribution includes contact point pressure value, center of gravity offset range, force balance state, contact point position, and pressure value range; the cargo stacking energy optimization value includes center of gravity offset adjustment parameter, contact point stacking matrix, energy relationship value after stacking, and stacking energy optimization matrix relationship; the cargo path planning parameter table includes grabbing point position, placement point position, grabbing point spatial priority value, obstacle point spatial priority value, lowest priority obstacle point spatial influence range, and obstacle point exclusion parameter.

2. The fully automatic palletizing method based on machine vision and path planning according to claim 1 is characterized in that: The steps for obtaining the cargo geometric feature description value are specifically as follows: S111: extracting the three-dimensional coordinates of each point in the cargo surface point cloud data, performing vector processing on the point cloud normal vector, calculating the curvature value of each point, analyzing the change trend of the curvature value in the point cloud point by point, removing abnormal points with excessive curvature changes, extracting a set of points with continuous distribution and smooth curvature in the point cloud, and generating cargo surface point cloud coordinates, normal vectors and curvature values; S112: Based on the point cloud coordinates and normal vector of the cargo surface, calculate the geometric angle difference between each point in the point cloud and the surrounding points, determine whether the geometric angle difference exceeds a set threshold, mark the points that meet the conditions as edge points, remove the points that do not meet the geometric conditions, re-analyze the boundary range of the remaining edge points, and generate a geometric segmentation range of the point cloud edge features; S113: based on the boundary points in the geometric segmentation range of the edge features of the point cloud, calculating the size distribution of the boundary points and fitting the curvature value relationship thereof to generate a multinomial distribution of the relationship between the boundary size and the curvature; S114: Calculate the deviation value of the curvature distribution according to the multinomial distribution of the relationship between the boundary size and the curvature, filter out abnormal data whose deviation value exceeds the fitting range, smooth the curvature distribution deviation value, re-verify the geometric consistency, and generate the cargo geometric feature description value.

3. The fully automatic palletizing method based on machine vision and path planning according to claim 2 is characterized in that: The steps for obtaining the cargo shape classification coefficient are specifically as follows: S211: extracting the curvature distribution and boundary feature positions of the cargo based on the cargo geometric feature description values, screening and extracting key points defining the cargo shape boundary by analyzing the change trend of the cargo surface curvature values ​​and combining the geometric distribution information of the boundary feature points, and generating a curvature distribution and boundary feature set; S212: using the curvature distribution and the boundary feature set, calculating the Euclidean distance from each feature point in the set to the geometric center of the cargo, constructing a geometric space model, analyzing the spatial concentration trend and outlier positions of the feature point distribution, and calibrating the spatial range of the feature points that meet the geometric conditions; S213: judging whether each feature point is within a reasonable geometric center position and distribution variance range according to the feature point spatial range, calculating the geometric distance from the point to the center and evaluating the distribution deviation, classifying and processing a set of feature points that meet the conditions, and obtaining a classified set of feature points; S214: Analyze the statistical distribution characteristics according to the classified feature point set, combine with the geometric distribution model, calculate the spatial distribution parameters of the feature point set class by class, establish the classification standard of the cargo shape, and generate the cargo shape classification coefficient.

4. The fully automatic palletizing method based on machine vision and path planning according to claim 3 is characterized in that: The steps for obtaining the cargo stacking force distribution are specifically as follows: S311: extracting the center of gravity coordinates and three-dimensional size range of the cargo based on the cargo shape classification coefficient, calculating the geometric center position and the corresponding volume range by analyzing the shape characteristics of the cargo unit item by item, eliminating the abnormally offset shape characteristic data, and obtaining the center of gravity coordinates and size data set; S312: Calculate the center of gravity coordinate offset and size deviation of the cargo unit according to the center of gravity coordinate and size data set, compare the offset data with the standard distribution of its corresponding geometric center, select the offset characteristic points that match the center position, remove the data with deviation values ​​exceeding a reasonable range, and generate center of gravity coordinate offset and size deviation relationship data; S313: According to the data on the relationship between the center of gravity coordinate offset and the size deviation, the distribution area of ​​the center of gravity offset of the goods in the stacked state is evaluated, the offset of the goods unit in the force balance state is calculated, the balance stability of the stack is determined by the offset characteristic point, and the contact point position and the pressure value range are adjusted; S314: In combination with the contact point position and the pressure value range, the force position of the contact point and the contact area ratio are adjusted by measuring the pressure distribution multiple times, the adjusted stacking stability is evaluated, and finally a force distribution model of the cargo stacking is determined to generate the cargo stacking force distribution.

5. The fully automatic palletizing method based on machine vision and path planning according to claim 4 is characterized in that: The steps for obtaining the cargo stacking energy optimization value are specifically as follows: S411: Based on the force distribution of the cargo stack, the contact point pressure value and the center of gravity offset range are extracted, and the distribution characteristics of the pressure value in the spatial coordinates and its offset trend are calculated by analyzing the correlation between the contact point pressure data and the offset value, so as to generate the contact point pressure and center of gravity offset data; S412: calculating the center of gravity offset adjustment parameter according to the contact point pressure and center of gravity offset data, calculating the relationship between the pressure variation range and the offset value of each contact point based on the adjustment parameter, and establishing a contact point stacking matrix in combination with the geometric correlation between the contact points; S413: combining the contact point stacking matrix, accumulating the stacking energy relationship value of each contact point, calculating the sum of the stacking energies through iterative optimization based on the energy relationship values ​​of each contact point, integrating the energy accumulation relationship and performing matrix optimization operations to generate a stacking energy optimization matrix relationship; S414: According to the stacking energy optimization matrix relationship, by analyzing the optimization value distribution of all contact points in the stacking energy matrix, accumulating the optimized energy results, calculating the total stacking energy value of all goods, and generating the goods stacking energy optimization value.

6. The fully automatic palletizing method based on machine vision and path planning according to claim 5 is characterized in that: The steps for obtaining the cargo path planning parameter table are specifically as follows: S511: Based on the cargo stacking energy optimization value, extract the three-dimensional positional relationship between the grabbing point and the placement point, analyze the spatial coordinates of the cargo unit item by item, calculate the distance and angle from the grabbing point to the placement point, select the coordinate characteristic points that affect the spatial connection, and generate the positional relationship data between the grabbing point and the placement point; S512: using the positional relationship data between the grabbing point and the placement point, calculating the spatial priority value between the grabbing point and the obstacle point, analyzing the spatial distribution of each obstacle point, screening the key obstacle points according to the geometric characteristics of the obstacle point's impact on the path, eliminating the non-key obstacle points with low impact, and generating a spatial priority ranking of the obstacle points; S513: According to the spatial priority ranking of the obstacle points, determine the spatial influence range of the lowest priority obstacle point, calculate the influence value of each obstacle point on the path, and generate optimized grasping path obstacle point exclusion parameters; S514: In combination with the optimized obstacle elimination parameters of the grasping path, the position arrangement of the obstacle points in the grasping path is adjusted, the connection path characteristics between the grasping points and the placement points are calculated, and a cargo path planning parameter table is generated by summarizing.

7. The fully automatic palletizing system based on machine vision and path planning is characterized by: For implementing the fully automatic palletizing method based on machine vision and path planning as described in any one of claims 1 to 6, the system comprises: The cargo feature extraction module extracts the cargo surface point cloud coordinates based on the cargo 3D point cloud data, calculates the normal vector and curvature value of the point cloud coordinates, determines the geometric segmentation range of the edge features, analyzes the fitting relationship between the boundary size and the curvature distribution, and generates the cargo geometric feature description value; The cargo shape classification module extracts the curvature distribution and boundary feature position based on the cargo geometric feature description value, calculates the Euclidean distance space range of the feature point, determines the geometric center position, analyzes the distribution variance range, classifies the feature set that meets the center distance and variance range, and generates the cargo shape classification coefficient; The stacking force analysis module extracts the coordinates of the center of gravity and the three-dimensional size range of the cargo based on the cargo shape classification coefficient, calculates the relationship between the center of gravity offset value and the size deviation, determines the distribution area of ​​the center of gravity offset and the force balance state, adjusts the position and pressure value range of the contact point, analyzes the mechanical distribution range of the contact point, and generates the cargo stacking force distribution; The stacking energy optimization module extracts the pressure value of the contact point and the center of gravity offset range based on the cargo stacking force distribution, calculates the center of gravity offset adjustment parameter and the contact point stacking matrix, accumulates the energy relationship value after stacking, establishes the stacking energy optimization matrix relationship, and generates the cargo stacking energy optimization value; The path planning module extracts the three-dimensional position relationship between the grabbing point and the placement point based on the cargo stacking energy optimization value, calculates the spatial priority value of the grabbing point and the obstacle point, analyzes the spatial influence range of the lowest priority obstacle point, optimizes the obstacle point exclusion parameters in the grabbing path, and generates a cargo path planning parameter table.

Images