Method for intelligent recommendation of fruit tree pruning based on point cloud skeleton analysis

Abstract

The present application relates to a kind of based on point cloud skeleton analysis's fruit tree pruning intelligent recommendation method, belong to agricultural information technology field.The method includes: S1: point cloud preprocessing: by statistical filtering and uniform downsampling technique removes noise and reduces data sampling rate, and is converted into triangular mesh model;S2: point cloud skeleton extraction: utilize skeleton extraction algorithm SEMC to point cloud is skeletonized processing, extracts the skeleton structure of fruit tree;S3: skeleton topological structure analysis: the skeleton extracted is decomposed into several unbranched segments, and segmented linked list structure is generated;Each segment in linked list represents an independent branch, and the node and branch information of linked list are used to describe the topological relationship of skeleton;S4: pruning intelligent recommendation: according to the branch information classified, combines pruning rule, identifies and marks the branch to be cut, and according to the geometric characteristics of branch generates pruning scheme;Meanwhile, through the visual display of pruning scheme, assist orchard management personnel to make decisions.

Description

Technical Field

[0001] The invention belongs to the field of agricultural information technology and relates to an intelligent recommendation method for fruit tree pruning based on point cloud skeleton analysis. Background Art

[0002] With the rapid development of modern agriculture, fruit tree pruning, a key step in orchard management, is crucial for optimizing tree structure and improving fruit yield and quality. However, traditional fruit tree pruning methods, which rely primarily on manual experience, suffer from low efficiency, high labor intensity, and inconsistent pruning quality, making them difficult to meet the refined management needs of modern orchards.

[0003] In recent years, the rise of three-dimensional point cloud technology has provided a new technical means for the refined analysis of fruit tree structure. Through point cloud modeling and skeleton extraction technology, the spatial structure of fruit trees can be clearly presented and gradually applied to the field of fruit tree management. Existing studies have attempted to use point cloud data to extract the skeleton of fruit trees and perform topological structure analysis for the automatic identification and geometric feature analysis of trunks and branches. Some studies have proposed pruning strategies based on classification results to assist manual decision-making. However, these studies generally focus on basic geometric analysis and ignore the need for accurate identification of various branches in complex tree structures, failing to achieve efficient and automated pruning recommendations.

[0004] The existing technologies for fruit tree point cloud skeleton analysis and pruning applications mainly have the following problems:

[0005] 1) High branch classification error: It is difficult to accurately distinguish the main trunk and different types of branches in complex tree structures, especially the identification of drooping branches, upright branches and competing branches, which has large errors.

[0006] 2) Imperfect pruning rules: The existing methods lack a unified standard for determining branches, resulting in poor applicability of pruning strategies among different fruit trees and making them difficult to promote and apply.

[0007] 3) The level of intelligent pruning is low: the geometric characteristics of the branches (such as length, angle, etc.) and their growth status are not fully integrated, and the accuracy and automation of pruning decisions are insufficient. Summary of the Invention

[0008] In light of this, the present invention aims to provide an intelligent recommendation method for fruit tree pruning based on point cloud skeleton analysis, enabling accurate identification and intelligent pruning recommendations for upright, competing, and drooping branches. This method effectively combines topological structure analysis of the point cloud skeleton with geometric characteristics such as branch length and angle to formulate scientific pruning rules and generate visual pruning plans. This method addresses existing issues such as large branch classification errors, inconsistent pruning rules, and insufficient intelligence, thereby improving the efficiency and accuracy of fruit tree pruning and promoting the development of automated and intelligent fruit tree pruning.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] An intelligent recommendation method for fruit tree pruning based on point cloud skeleton analysis specifically includes the following steps:

[0011] S1: Point cloud preprocessing:

[0012] De-noising is performed on the input fruit tree point cloud data, specifically including: using statistical filtering methods to remove isolated points, error points, and noise points; reducing the point cloud density through uniform sampling methods while retaining the overall geometric characteristics of the fruit tree; and converting the point cloud data into a triangular mesh model as needed;

[0013] S2: Point cloud skeleton extraction:

[0014] The skeleton extraction algorithm SEMC is used to perform skeleton processing on the fruit tree point cloud after point cloud preprocessing to extract the skeleton structure of the fruit tree;

[0015] S3: Skeleton topology analysis:

[0016] The extracted skeleton is decomposed into several non-branched segments and a segmented linked list structure is generated. Each segment in the linked list represents an independent branch. The node and branch information of the linked list is used to describe the topological relationship of the skeleton, which facilitates the classification and hierarchical processing of the branches.

[0017] S4: Intelligent pruning recommendation:

[0018] Based on the classified branch information and combined with pruning rules (such as removing vertical branches, dense branches, competing branches and drooping branches), the branches to be pruned are identified and marked, and a pruning plan is generated based on the geometric characteristics of the branches (such as length, diameter, bifurcation angle and spatial distribution); the pruning rules are set in combination with the branch length, diameter, bifurcation angle and spatial distribution characteristics, and can be adapted to different fruit tree types and pruning needs; at the same time, the visual display of the pruning plan assists orchard managers in making decisions.

[0019] Furthermore, in step S2, the skeleton extraction algorithm SEMC is a skeleton extraction algorithm based on mesh shrinkage, which specifically includes: first using the Meshlab tool to preprocess the fruit tree point cloud, including normal smoothing and spherical rotation surface reconstruction, to convert the point cloud into a triangular mesh model that meets the requirements; then calculating the mesh curvature and shrinking the mesh according to the curvature information to extract the skeleton structure.

[0020] Other alternative skeleton extraction algorithms include VOXEL-based shrinkage algorithms or L1-medial algorithms. Furthermore, in step S3, the skeleton topology analysis specifically includes: using a directed acyclic graph (DAG) to describe the topological structure of the point cloud data; organizing the paths between key nodes into a linked list structure through feature extraction and a depth-first search (DFS)-based path tracing algorithm; then, using a KDTree radius search to screen out potential broken branch connection points, and using an angle metric to screen out the only correctly connected nodes; finally, completing the skeleton restoration by calculating the growth direction of the branch linked list.

[0021] Furthermore, in step S3, in the skeleton topology analysis, the geometric features of the skeleton are extracted through the local geometric shapes of the tree point cloud, including the direction and curvature of the branches and the connection relationship between the nodes;

[0022]

[0023]

[0024] in, represents the direction vector of branch i, x i+1 、y i+1 、z i+1 Represents the next node position p i+1 The three-dimensional space coordinates of κ i is the curvature of the branch; through curvature analysis, the degree of bending of the branch can be identified, which is helpful for further topological analysis.

[0025] Skeleton repair specifically involves: first, dividing the skeleton into two parts: the correct skeleton and broken branches. The correct skeleton refers to all skeleton nodes on the path from any leaf node, traversing along the parent node to the root node; the remaining nodes that do not belong to the correct skeleton are considered broken branches. The task of skeleton repair is to reconnect and correct the broken branches. This requires calculating the distance and angle difference between the broken branch nodes and the correct skeleton nodes, screening potential broken branch connection points through KDTree radius search, and using angle measurement to determine the only correct connection node.

[0026] KDTree radius search: Through KDTree radius search, filter out all candidate connection points pj within the radius r from the broken branch node pi:

[0027] d(p i ,p d )≤r (3)

[0028] Among them, p i is a broken branch node, p j is a candidate connection point, d(p i ,pj ) is p i and p j the distance between them;

[0029] Angle metric: For all candidate connection points p j , calculate its relationship with the broken branch node p i The angle θ(p i ,p j ), select the connection point with the smallest angle as the correct connection node p connect :

[0030]

[0031] in, It is p i The direction vector to the previous node; It is p j to p i The direction vector.

[0032] Furthermore, in step S4, the branches to be pruned are identified, specifically including: formulating scientific pruning rules based on the experts' years of experience and judgment; extracting the geometric features of the branches (such as angle, length, etc.) by analyzing the topological structure of the fruit tree, and preliminarily realizing the automatic identification of upright branches, drooping branches, competing branches and dense branches; then, in order to improve the recognition accuracy, the angle threshold and other related parameters are dynamically adjusted according to the growth characteristics of different fruit trees, thereby optimizing the recognition effect of the pruning target; among them, the priority display order of the branches is drooping branches> upright branches> competing branches> dense branches (that is, some branches may be both drooping branches and dense branches, in which case they are preferentially displayed as drooping branches).

[0033] Furthermore, in step S4, the identification of upright branches is specifically performed by calculating the angle between two tree branches to determine whether it is an upright branch, thereby deciding whether to prune it;

[0034] The angle between branches is calculated as follows:

[0035]

[0036] Among them, θ ij represents the angle between two branches i and j, p i 、p j are the starting points of the two branches, p′ i , p′ j is their end point; x i 、y i 、z i They represent the starting point p of branch i respectively i The three-dimensional space coordinates, x′ i , y′ i 、z′i They represent the endpoint p′ of branch i respectively i The three-dimensional space coordinates, x′ j , y′ j 、z′ j They represent the endpoint p′ of branch j respectively j ; if the angle is in the interval [80°, 100°], the higher-level branch is identified as an upright branch.

[0037] Furthermore, in step S4, drooping branches generally refer to branches that grow toward the ground and exhibit a drooping phenomenon; drooping branches can be identified by calculating the angle between each branch and the Y axis;

[0038] Growth direction vector Defined as:

[0039]

[0040] in, and are the coordinates of two adjacent points on the branch, and n is the total number of branch point clouds;

[0041] The Y axis vector is defined as:

[0042] The angle θ is calculated as:

[0043]

[0044] If θ>90°, the branch is a drooping branch.

[0045] Furthermore, in step S4, competing branches are upright and vigorous branches that sprout from the second and third buds below the cut. They compete with the extension branches for growth space, affecting the development of the main trunk or main branches and need to be pruned. This usually occurs at the end of the branch. All the terminal branches can be traversed first, and then the competing branches can be identified based on the branch length and growth direction.

[0046] (1) Determination of branch length: If The shorter branch is selected as the competing branch; otherwise, the growth direction is used for further judgment; where L1 and L2 are the lengths of the two terminal branches, and D min is the length of the shortest branch among the branches;

[0047] The calculation formula of branch length D is:

[0048]

[0049] Where n is the total number of branch point clouds, x i 、y i 、z iRepresents the three-dimensional space coordinates of the i-th point cloud on the branch, x j 、y j 、z j Respectively represent the three-dimensional space coordinates of the j-th point cloud;

[0050] (2) Growth direction judgment: Formula (6) is used to calculate the angles between the two terminal branches and their parent branches. The smaller the angle, the more similar the growth direction of the terminal branch is to the main trunk, and the more it should be retained. Therefore, the branch with a larger angle value is identified as a competing branch.

[0051] Furthermore, in step S4, dense branches refer to branches that are too dense in the same area, resulting in branches blocking each other or growing crosswise, usually causing the crown to appear irregular and crowded. These branches usually grow at very close angles from the trunk or main branches, forming narrow angles, making the branches almost parallel or overlapping, resulting in affected lighting and ventilation. To identify these branches, the angle between branches and the branch density in the area can be combined to determine.

[0052] The identification of dense branches specifically includes: first, preliminary screening through density analysis to identify possible dense areas; then, by analyzing the angles between branches and the degree of matching between the branches and the growth direction of the parent branches, further screening out the dense branches that need to be pruned; this method combines the geometric characteristics of spatial distribution density and branch growth direction, and can accurately identify and mark dense branches.

[0053] (1) Analysis of branch density within the region

[0054] Counting the number of branches in a specific area of ​​a tree, where branches are too dense and may obscure or overlap each other, can be used to assess the density of branches and serve as a preliminary screening criterion.

[0055] Divide the point cloud space into grid cells: First, use a 3D space partitioning method to divide the tree point cloud data into several smaller grid cells. The number of branches in each grid cell can reflect the branch density in the area. The specific steps are as follows:

[0056] 1) Define the grid size: Set an appropriate grid size based on the scale of the trees and the density of the point cloud. Assume that the space is divided into a cubic grid with a side length of δ.

[0057] 2) Grid division: For each point p i =(x i ,y i ,z i ), calculate the grid cell where it is located; set the coordinates of the grid cell to:

[0058]

[0059] in, Indicates a round-down operation;

[0060] 3) Count the number of branches in each grid: Calculate the number of point clouds N belonging to different branches in each grid grid , that is, the number of different types of point clouds represents the number of branches. If the number of branches in a grid is N grid Exceeds the set threshold N thresh =3, it is considered that there are dense branches in the area;

[0061]

[0062] Among them, I(p i ∈grid k ) is the indicator function, m represents the total number of branches; when point p i The value is 1 if it falls within the grid k, otherwise it is 0;

[0063] (2) Branch angle analysis

[0064] After initially screening out possible dense branch areas through density analysis, we can further analyze the angles between branches to ultimately identify dense branches. In dense areas, if the angles between branches are too small and the growth direction of these branches is inconsistent with the direction of the parent branches, they may be dense branches that need to be pruned.

[0065] 1) Parent branch: If branch b grows from branch a, then branch a is the parent branch of branch b;

[0066] 2) Calculate the angle between branches: For each pair of branches i and j, the angle θ between them can be calculated ij :

[0067]

[0068] If the angle θ between the branches ij Less than the set threshold θ thresh =30°, it is considered that the growth directions of these branches are too close and may be dense branches;

[0069] 3) For each branch i, calculate the angle between its growth direction and that of the parent branch; if the angle is small, it indicates that its growth direction is relatively consistent with that of the parent branch and should be retained; if the angle is large, it indicates that its growth direction is inconsistent with that of the parent branch and should be considered a dense branch;

[0070]

[0071] in, is the growth direction vector of branch i, Is the growth direction vector of the parent branch; if θ parent >θ thresh , then the branch is considered to be a dense branch.

[0072] The beneficial effects of the present invention are as follows: the present invention significantly improves the intelligent level of fruit tree pruning management through the intelligent recommendation method of fruit tree pruning based on point cloud skeleton analysis, and has the following significant effects compared with the existing technology:

[0073] (1) Performance and efficiency improvement: This paper uses an improved SEMC algorithm to extract the skeleton of fruit tree point clouds, significantly improving both extraction efficiency and accuracy, and can better adapt to the topological structure of complex fruit trees. At the same time, by analyzing the geometric characteristics of branches (such as length, angle, and spatial distribution) and combining scientific pruning rules, this paper achieves accurate identification of unfavorable branches such as vertical branches, competing branches, and drooping branches. The accuracy of pruning recommendations reaches over 90%, significantly improving the scientific nature of pruning decisions.

[0074] (2) Reduction of labor and cost: The present invention can automatically generate pruning suggestions through point cloud data, reducing the reliance on human participation and lowering the labor cost of orchard management, especially in large-scale orchard management.

[0075] (3) This invention effectively addresses existing issues such as high branch classification errors, difficulty in generalizing and adapting pruning rules, and insufficient intelligence in pruning decision-making. By adopting a unified branch grading standard and intelligent recommendation scheme, the efficiency and accuracy of fruit tree pruning are comprehensively improved, addressing the shortcomings of existing technologies.

[0076] (4) This invention achieves the goal of accurate branch classification and intelligent pruning recommendations, which not only optimizes the fruit tree pruning process but also ensures the scientific and universal nature of the pruning scheme. Compared with traditional methods, pruning efficiency is significantly improved, and the growth quality and fruit yield of fruit trees after pruning are significantly improved, providing strong technical support for smart orchard management.

[0077] (5) Ecological and economic benefits: By reducing the labor and material input costs during pruning, the present invention maximizes the economic benefits of orchard management. At the same time, the method can effectively improve the growth environment of fruit trees, increase fruit yield and quality, and help modern agriculture develop in the direction of high efficiency and intelligence.

[0078] In summary, the present invention is significantly superior to existing technologies in terms of performance, efficiency, cost, and intelligence level, and has important application value and broad promotion prospects.

[0079] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0081] Figure 1 This is a tree branch classification roadmap based on skeleton extraction and topological structure analysis of the present invention;

[0082] Figure 2 Diagram of pruning recognition effect during pruning of fruit trees with different topological structures. DETAILED DESCRIPTION

[0083] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0084] Among them, the accompanying drawings are only for illustrative purposes and represent only schematic diagrams rather than actual pictures, and should not be understood as limiting the present invention. In order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the dimensions of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the accompanying drawings.

[0085] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "back", etc. indicating directions or positional relationships, they are based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0086] See also Figures 1 and 2 The present invention provides an intelligent recommendation method for fruit tree pruning based on point cloud skeleton analysis, which specifically includes the following steps:

[0087] (1) Point cloud preprocessing:

[0088] The input fruit tree point cloud data is denoised using statistical filtering to remove isolated points, error points, and noise points, thereby improving data quality. To optimize computational efficiency, uniform sampling is used to reduce point cloud density while preserving the overall geometric characteristics of the fruit tree. Furthermore, the point cloud data can be converted into a triangular mesh model as needed, providing a smoother geometric representation for subsequent skeleton extraction.

[0089] (2) Point cloud skeleton extraction:

[0090] The SEMC algorithm is used to skeletonize the fruit tree point cloud and extract the tree's skeleton structure. The SEMC algorithm can retain the main topological information of the fruit tree, providing a basis for subsequent branch classification and pruning recommendations.

[0091] (3) Skeleton topology analysis:

[0092] The extracted skeleton is topologically analyzed, decomposing it into several unbranched segments and generating a segmented linked list structure. Each segment in the linked list represents an independent branch. The node and branch information in the linked list is used to describe the topological relationship of the skeleton, facilitating the classification and hierarchical processing of the branches.

[0093] (4) Intelligent pruning recommendation:

[0094] Based on the classified branch information and pruning rules (such as removing vertical branches, competing branches, and drooping branches), branches requiring pruning are marked and a pruning plan is generated. Pruning rules are set based on branch length, diameter, bifurcation angle, and spatial distribution characteristics to adapt to different fruit tree species and pruning requirements. Furthermore, the visual display of pruning plans assists orchard managers in decision-making.

[0095] The system structure corresponding to the method of the present invention is:

[0096] (1) Point cloud preprocessing module: The input is the original fruit tree point cloud data. After denoising and density reduction, the quality and computational efficiency of the point cloud data are optimized and converted into a triangular mesh model as needed. The processed point cloud data will serve as the input of the subsequent point cloud skeleton extraction module.

[0097] (2) Point cloud skeleton extraction module: Based on the processed point cloud data, the SEMC algorithm is used to extract the skeleton structure of the fruit tree. The extraction results will serve as the input of the skeleton topology analysis module, providing the main topological information and geometric shape for further analysis.

[0098] (3) Skeleton topology analysis module: Using the skeleton data obtained from the point cloud skeleton extraction module, we perform topological analysis and decompose the skeleton into several non-branched segments. The topological relationship of the skeleton is stored in a segmented linked list structure. This structure will serve as the input of the intelligent pruning recommendation module to help achieve the classification and accurate identification of pruned branches.

[0099] (4) Intelligent pruning recommendation module: Based on the real-world pruning identification experience of experts and the set pruning rules (such as removing vertical branches, competing branches, and drooping branches), branches that need to be pruned are marked and a pruning plan is generated based on the geometric characteristics of the branches (such as length, diameter, bifurcation angle, and spatial distribution). The final output pruning plan is visualized to help orchard managers make decisions.

[0100] Alternatives: The skeleton extraction module can be replaced with other skeletonization algorithms adapted to point cloud data (such as a VOXEL-based shrinkage algorithm or an L1-medial algorithm). The branch classification algorithm can also optimize geometric feature extraction methods based on specific fruit tree structures. Pruning rules can be flexibly adjusted to suit the pruning needs of different fruit trees.

[0101] Figure 1 A roadmap for tree branch classification based on skeleton extraction and topological structure analysis, e.g. Figure 1 As shown, the present invention provides a branch classification method based on skeleton extraction and topological analysis, which mainly includes three stages: point cloud preprocessing, skeleton topology analysis, and pruning identification. In the point cloud preprocessing stage, statistical filtering and uniform downsampling techniques are used to remove noise and reduce the data sampling rate to improve the quality and consistency of the point cloud data. The processed point cloud data is converted into a triangular mesh using MeshLab for subsequent analysis. In the skeleton topology analysis stage, a directed acyclic graph (DAG) is used to describe the topological structure of the point cloud data. Through feature extraction and a path tracing algorithm based on depth-first search (DFS), the paths between key nodes are organized into a linked list structure. Subsequently, a KDTree radius search is performed to screen potential broken branch connection points, and an angle metric is used to screen the only correctly connected nodes. Finally, the skeleton is restored by calculating the growth direction of the branch linked list. In the pruning identification stage, scientific pruning rules are formulated based on the years of experience and judgment of experts. By analyzing the topological structure of the fruit tree and extracting the geometric features of the branches (such as angle and length), the automatic identification of upright branches, drooping branches, and competing branches is initially achieved. To improve recognition accuracy, the angle threshold and other related parameters are dynamically adjusted according to the growth characteristics of different fruit trees, thereby optimizing the recognition effect of pruning targets.

[0102] Figure 2This figure shows the branch recognition effect during pruning of fruit trees with different topological structures, where blue represents drooping branches, green represents upright branches, purple represents competing branches, and red represents dense branches. The specific rules for identifying branches during fruit pruning are as follows: During fruit tree pruning, drooping branches (blue) are branches in the middle and lower parts of the crown with an angle exceeding 90° with the vertical plane. They need to be pruned because they affect fruit picking or light exposure; upright branches (green) are upright branches in the middle part of the crown that grow vigorously, with an angle of 80° to 100° and a diameter greater than 1 cm. They need to be pruned because they compete for nutrients and affect fruit quality; competing branches (purple) are upright and vigorous branches that sprout from the second and third buds below the cut. They need to be pruned because they compete with extension branches for growth space, affecting the development of the main trunk or main branches. Dense branches (red) refer to branches that are too dense in the same area, resulting in branches blocking each other or growing crosswise, usually making the crown appear irregular and crowded. These branches often grow at very close angles from the trunk or main branches, forming narrow angles that make the branches almost parallel or overlapping, which affects light and ventilation. To identify these branches, you can combine the angle between the branches and the branch density in the area.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent recommendation method for fruit tree pruning based on point cloud skeleton analysis, characterized in that: The method specifically comprises the following steps: S1: Point cloud preprocessing: De-noising is performed on the input fruit tree point cloud data, specifically including: using statistical filtering methods to remove isolated points, error points, and noise points; reducing the point cloud density through uniform sampling methods while retaining the overall geometric characteristics of the fruit tree; and converting the point cloud data into a triangular mesh model as needed; S2: Point cloud skeleton extraction: The skeleton extraction algorithm SEMC is used to perform skeleton processing on the fruit tree point cloud after point cloud preprocessing to extract the skeleton structure of the fruit tree; S3: Skeleton topology analysis: The extracted skeleton is decomposed into several non-branched segments and a segmented linked list structure is generated; each segment in the linked list represents an independent branch, and the node and branch information of the linked list is used to describe the topological relationship of the skeleton; The skeleton topology analysis specifically involves describing the topology of point cloud data using a directed acyclic graph (DAG). The paths between key nodes are organized into a linked list structure through feature extraction and a depth-first search-based path tracing algorithm. Subsequently, a KDTree radius search is used to screen out potential broken branch connection points, and angle metrics are used to identify the only correctly connected nodes. Finally, the skeleton is repaired by calculating the growth direction of the branch linked list. Skeleton repair specifically involves dividing the skeleton into two parts: the correct skeleton and broken branches. The correct skeleton refers to all skeleton nodes along the path from any leaf node to the root node, along the parent node. The remaining nodes that do not belong to the correct skeleton are considered broken branches. The task of skeleton repair is to reconnect and correct the broken branches. This requires calculating the distance and angle difference between the broken branch nodes and the correct skeleton nodes, screening potential broken branch connection points through KDTree radius search, and using angle measurement to determine the only correct connection node. KDTree radius search: filter out nodes with a distance to a broken branch by searching with a KDTree radius In radius r All candidate join points in range : in, It is a broken branch node. is a candidate connection point, yes and the distance between them; Angle metric: For all candidate connection points , calculate its relationship with the broken branch node Angle between , select the connection point with the smallest angle as the correct connection node : in, yes The direction vector to the previous node; yes arrive direction vector; S4: Intelligent pruning recommendation: Based on the classified branch information and combined with pruning rules, the branches to be pruned are identified and marked, and a pruning plan is generated based on the geometric characteristics of the branches. The pruning rules are set in combination with branch length, diameter, bifurcation angle and spatial distribution characteristics to adapt to different fruit tree types and pruning needs. At the same time, the visual display of the pruning plan assists orchard managers in making decisions.

2. The fruit tree pruning intelligent recommendation method according to claim 1, characterized in that: In step S2, the skeleton extraction algorithm SEMC is a skeleton extraction algorithm based on mesh shrinkage, which specifically includes: first using the Meshlab tool to preprocess the fruit tree point cloud, including normal smoothing and spherical rotation surface reconstruction, to convert the point cloud into a triangular mesh model that meets the requirements; then calculating the mesh curvature and shrinking the mesh according to the curvature information to extract the skeleton structure.

3. The fruit tree pruning intelligent recommendation method according to claim 1, characterized in that: In step S3, in the skeleton topology analysis, the geometric features of the skeleton are extracted through the local geometric shapes of the tree point cloud, including the direction and curvature of the branches and the connection relationship between the nodes; in, Indicates branch The direction vector, 、 、 Respectively represent the current node position The three-dimensional space coordinates of 、 、 Respectively represent the next node position The three-dimensional space coordinates of is the curvature of the branch; through curvature analysis, the degree of bending of the branch can be identified.

4. The method for intelligent recommendation of fruit tree pruning according to claim 1, characterized in that: In step S4, the branches to be pruned are identified, specifically including: formulating pruning rules based on the experts' years of experience and judgment; extracting the geometric features of the branches by analyzing the topological structure of the fruit tree, and preliminarily realizing the automatic identification of upright branches, drooping branches, competing branches and dense branches; then dynamically adjusting the angle threshold and other related parameters according to the growth characteristics of different fruit trees to optimize the identification effect of pruning targets; among them, the priority display order of branches is drooping branches > upright branches > competing branches > dense branches.

5. The method for intelligent recommendation of fruit tree pruning according to claim 4, characterized in that: In step S4, the identification of upright branches is specifically performed by calculating the angle between two tree branches to determine whether it is an upright branch, thereby deciding whether to prune it; The angle between branches is calculated as follows: in, Represents two branches i and j The angle between 、 are the starting points of the two branches. 、 is their end; 、 、 Represents branches respectively i starting point The three-dimensional space coordinates of 、 、 Represents branches respectively j starting point The three-dimensional space coordinates of 、 、 Represents branches respectively i end The three-dimensional space coordinates of 、 、 Represents branches respectively j end ; if the angle is in the interval [80°, 100°], the higher-level branch is identified as an upright branch.

6. The method for intelligent recommendation of fruit tree pruning according to claim 4, characterized in that: In step S4, drooping branches refer to branches that grow toward the ground and show a drooping phenomenon; drooping branches are identified by calculating the angle between each branch and the Y axis; Growth direction vector Defined as: in, and are the coordinates of two adjacent points on the branch, n is the total number of branch point clouds; The Y axis vector is defined as: ; Angle Calculated: if >90°, the branch is a drooping branch.

7. The method for intelligent recommendation of fruit tree pruning according to claim 5, characterized in that: In step S4, the competing branches are upright and vigorous branches that sprout from the second and third buds below the cut, and occur at the end of the branches; first traverse all the end branches, and then identify the competing branches based on the branch length and growth direction; (1) Determination of branch length: If |L1- L2| > , then the shorter branch is selected as the competing branch; otherwise, the growth direction is used for further judgment; where L1 and L2 are the lengths of the two terminal branches, is the length of the shortest branch among the branches; Branch length The calculation formula is: in, n is the total number of branch point clouds, 、 、 Respectively represent the first i The three-dimensional space coordinates of the point cloud, 、 、 Respectively represent j The three-dimensional space coordinates of the point cloud; (2) Growth direction judgment: Formula (6) is used to calculate the angles between the two terminal branches and their parent branches. The smaller the angle, the more similar the growth direction of the terminal branch is to the main trunk, and the more it should be retained. Therefore, the branch with a larger angle value is identified as a competing branch.

8. The method for intelligent recommendation of fruit tree pruning according to claim 5, characterized in that: In step S4, the identification of dense branches specifically includes: firstly, performing preliminary screening through density analysis to identify dense areas; then, further screening dense branches that need to be pruned by analyzing the angles between branches and the degree of matching between the growth directions of branches and parent branches; (1) Analysis of branch density within the region Divide the point cloud space into grid cells: First, use a 3D space partitioning method to divide the tree point cloud data into several small grid cells. The number of branches in each grid cell reflects the branch density in the area. The specific steps are as follows: 1) Define the grid size: Set the grid size according to the scale of the trees and the density of the point cloud; assume that the space is divided into A cube grid; 2) Grid division: For each point , calculate the grid cell where it is located; set the coordinates of the grid cell to: in, Indicates a round-down operation; 3) Count the number of branches in each grid: Calculate the number of point clouds belonging to different branches in each grid , that is, the number of different types of point clouds represents the number of branches. If the number of branches in a grid is Exceeding the set threshold , it is considered that there are dense branches in the area; in, is the indicator function, m Indicates the total number of branches; when point Falling on the grid k The internal time value is 1, otherwise it is 0; (2) Branch angle analysis 1) Parent branch: If branch b grows from branch a, then branch a is the parent branch of branch b; 2) Calculate the angle between branches: For each pair of branches i and j, calculate the angle between them : If the angle between branches Less than the set threshold , it is considered that the growth directions of these branches are too close and they may be dense branches; 3) For each branch i , calculate the angle between its growth direction and the parent branch; if the angle is small, it indicates that its growth direction is relatively consistent with that of the parent branch and should be retained; if the angle is large, it indicates that its growth direction is inconsistent with that of the parent branch and should be regarded as a dense branch; in, It's a branch i The growth direction vector, Is the growth direction vector of the parent branch; if , then the branch is considered to be a dense branch.

Images