Litchi tree shape digitization reconstruction method, device and equipment and storage medium

Abstract

The present application relates to artificial intelligence technology, and discloses a litchi tree shape digital reconstruction method, device, equipment and storage medium. The method comprises: scanning the target litchi orchard environment to obtain point cloud information, and constructing an orchard point cloud map according to the point cloud information; using a litchi tree segmentation model to identify the litchi trees in the orchard point cloud map to obtain target litchi tree point cloud distribution; performing object recognition operation on the target litchi tree point cloud distribution based on the trunk, crown and whole tree to obtain an object recognition result, and marking the object recognition result with a bounding box to obtain a bounding box set, wherein the bounding box set includes a litchi trunk bounding box, a litchi crown layer bounding box and a litchi whole tree bounding box; using a pre-constructed point cloud slicing algorithm to extract information from the bounding box set to obtain a litchi tree shape digital information set. The present application can improve the efficiency and accuracy of litchi tree shape digitization.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and computer-readable storage medium for digitally reconstructing the appearance of a litchi tree. Background Art

[0002] Lychee is an important cash crop in my country's tropical and subtropical regions. Its high economic value makes it a primary source of income for many farmers. By measuring the appearance of lychee trees and digitizing their data, their growth can be monitored and their health assessed. Indicators such as crown size and shape reflect the tree's growth potential and health. In smart agriculture, equipment such as robots and drones are widely used in field management. Measuring lychee tree appearance data provides navigation and operational basis for these devices, improving the accuracy and efficiency of automated management. Accurate lychee tree appearance data can guide mechanized operations such as automated pruning and fruit picking, reducing labor costs and improving efficiency. Combining historical data with real-time monitoring data can improve the accuracy of yield forecasts and help farmers develop informed planting and marketing plans.

[0003] Traditionally, lychee tree shape measurement relies on manual measurement, which is time-consuming and inaccurate. Furthermore, lychee trees are generally tall, placing surveyors at risk. To address these challenges and achieve accurate lychee tree shape measurement, most currently use ground scanners. However, these scanners present several challenges: They require a long time to inspect an entire lychee orchard, hinder data exchange with other equipment, and are difficult to maneuver. Furthermore, ground scanners require high surface flatness, making them difficult to move and unable to meet the diverse needs of digitizing lychee tree shapes. Summary of the Invention

[0004] The present invention provides a litchi tree appearance digital reconstruction method, device, equipment and storage medium, the main purpose of which is to improve the efficiency and accuracy of litchi tree appearance digital reconstruction.

[0005] To achieve the above-mentioned object, the present invention provides a method for digitally reconstructing the appearance of a litchi tree, comprising:

[0006] Use a pre-built point cloud acquisition device to scan the target litchi orchard to obtain point cloud information, and use a pre-built SLAM algorithm to construct a point cloud map of the orchard based on the point cloud information;

[0007] Using a pre-trained litchi tree segmentation model to identify litchi trees on the orchard point cloud map, and obtaining a target litchi tree point cloud distribution;

[0008] Performing an object recognition operation based on the trunk, crown, and entire tree on the target litchi tree point cloud distribution to obtain an object recognition result, and performing bounding box selection marking on the object recognition result to obtain a bounding box set, wherein the bounding box set includes a litchi trunk bounding box, a litchi crown bounding box, and an entire litchi tree bounding box;

[0009] A pre-built point cloud slicing algorithm is used to extract information from the bounding box set to obtain a digital information set of the litchi tree's appearance.

[0010] Optionally, the step of constructing an orchard point cloud map based on the point cloud information using a pre-built SLAM algorithm includes:

[0011] Sorting the point cloud information according to the position of the point cloud acquisition device to obtain a single-frame point cloud sequence;

[0012] Using a pre-built inertial measurement unit to obtain the position and attitude of the point cloud acquisition device, and recording the position and attitude change sequence within a preset time period;

[0013] Acquire a high-definition garden image at a target node position of a motion trajectory of the point cloud acquisition device during the point cloud acquisition process;

[0014] According to the pre-built SLAM algorithm, based on the position and posture change sequence, the target node position and the high-definition garden image, each single-frame point cloud in the single-frame point cloud sequence is subjected to a point cloud deployment operation based on the world coordinate system to obtain a primary orchard point cloud map;

[0015] According to the pre-built loop detection algorithm, the primary orchard point cloud map is updated to obtain an orchard point cloud map.

[0016] Optionally, the inertial measurement unit is expressed as:

[0017]

[0018] Where, represents the angular velocity of the inertial measurement unit at time t in the body coordinate system B of the point cloud acquisition device, represents the acceleration of the inertial measurement unit at time t in the body coordinate system B of the point cloud acquisition device, ω t represents the actual measured angular velocity, represents the angular velocity deviation of the point cloud acquisition device at time t, represents the angular velocity white noise of the point cloud acquisition device at time t, represents the rotation matrix from the world coordinate system to the body coordinate system B of the point cloud acquisition device, α t represents the actual measured acceleration, g represents the gravity vector, represents the acceleration deviation of the point cloud acquisition device in the body coordinate system B at time t, represents the acceleration white noise of the point cloud acquisition device at time t.

[0019] Optionally, the position and posture of the point cloud acquisition device is expressed as:

[0020]

[0021] In the formula, the R t Represents the rotation matrix Δt represents a period of time, v t+Δt represents the speed of the point cloud acquisition device at time t+Δt, p t+Δt represents the position of the point cloud acquisition device at time t+Δt, R t+Δt Represents the rotation matrix of the point cloud acquisition device at time t+Δt.

[0022] Optionally, the step of using a pre-trained litchi tree segmentation model to identify litchi trees on the orchard point cloud map to obtain a target litchi tree point cloud distribution includes:

[0023] Obtaining point cloud sample data of similar litchi trees of the target litchi tree in the target litchi orchard;

[0024] Obtain a pre-built litchi tree segmentation model, and train the litchi tree segmentation model using the similar litchi tree point cloud sample data to obtain a pre-trained litchi tree segmentation model;

[0025] Obtaining artificially identified samples corresponding to the orchard point cloud map of the target litchi orchard, and fine-tuning the pre-trained litchi tree segmentation model using the artificially identified samples to obtain a trained litchi tree segmentation model;

[0026] The trained litchi tree segmentation model is used to identify litchi trees on the orchard point cloud map to obtain the target litchi tree point cloud distribution.

[0027] Optionally, performing an object recognition operation based on the trunk, crown, and entire tree on the target litchi tree point cloud distribution to obtain an object recognition result, and performing bounding box selection marking on the object recognition result to obtain a bounding box set includes:

[0028] performing outlier removal on the target litchi tree point cloud distribution to obtain a converged litchi tree point cloud distribution;

[0029] Using a pre-built clustering algorithm, clustering and grouping operations based on trunks and crowns are performed on the convergent litchi tree point cloud distribution to obtain trunk point cloud distribution and crown point cloud distribution;

[0030] Using a principal component analysis algorithm, the numerical features of the trunk point cloud distribution and the crown point cloud distribution based on the principal axis direction of the world coordinate system are obtained to obtain a numerical feature set, and a minimum bounding box set is constructed based on the numerical feature set;

[0031] The entire tree point cloud distribution of the trunk point cloud distribution and the crown point cloud distribution is obtained, and the entire tree point cloud distribution, the trunk point cloud distribution and the crown point cloud distribution are framed and marked using the minimum bounding box set.

[0032] Optionally, the method of extracting information from the bounding box set using a pre-built point cloud slicing algorithm to obtain a digitized information set of the litchi tree shape includes:

[0033] Using a pre-built point cloud slicing algorithm, the point cloud distribution of the target litchi tree in the bounding box set is sliced ​​to obtain a canopy plane sequence;

[0034] Performing a two-dimensional convex hull operation on the canopy plane sequence to obtain a convex hull canopy plane, and obtaining the area and volume of the convex hull canopy plane to obtain the maximum canopy area and the maximum canopy volume;

[0035] Integrating the height corresponding to each canopy plane in the canopy plane sequence, the maximum canopy area and the maximum canopy volume corresponding to each convex hull canopy plane, to obtain an integer point cloud distribution volume corresponding to the target litchi tree point cloud distribution;

[0036] The maximum canopy area, the maximum canopy volume and the integer point cloud distribution volume are outputted simultaneously to obtain a digital information set of the litchi tree appearance.

[0037] In order to solve the above problems, the present invention further provides a device for digitally reconstructing the appearance of a litchi tree, the device comprising:

[0038] A point cloud acquisition module is used to scan the target litchi orchard using a pre-built point cloud acquisition device to obtain point cloud information, and to construct a point cloud map of the orchard based on the point cloud information using a pre-built SLAM algorithm;

[0039] A litchi tree recognition module is used to identify litchi trees on the orchard point cloud map using a pre-trained litchi tree segmentation model to obtain a target litchi tree point cloud distribution;

[0040] An object feature segmentation module is configured to perform object recognition operations on the target litchi tree point cloud distribution based on the trunk, crown, and entire tree to obtain an object recognition result, and to perform bounding box selection and marking on the object recognition result to obtain a bounding box set, wherein the bounding box set includes a litchi trunk bounding box, a litchi crown bounding box, and a litchi tree overall bounding box;

[0041] The litchi tree shape digitization module is used to extract information from the bounding box set using a pre-built point cloud slicing algorithm to obtain a litchi tree shape digitization information set.

[0042] In order to solve the above problem, the present invention further provides an electronic device, comprising:

[0043] at least one processor; and,

[0044] a memory communicatively connected to the at least one processor; wherein,

[0045] The memory stores a computer program that can be executed by the at least one processor. The computer program is executed by the at least one processor so that the at least one processor can perform the above-mentioned method for digitally reconstructing the appearance of a litchi tree.

[0046] In order to solve the above problems, the present invention also provides a computer-readable storage medium, which stores at least one computer program. The at least one computer program is executed by a processor in an electronic device to implement the above-mentioned method for digitally reconstructing the appearance of a litchi tree.

[0047] The embodiment of the present invention first uses a point cloud acquisition device to perform an environmental scan of the target litchi orchard to initially obtain point cloud information. The point cloud information is then arranged using a SLAM algorithm to construct a point cloud map of the orchard. The SLAM algorithm can rearrange the serialized point cloud information according to the movement of the point cloud acquisition device to obtain a three-dimensional point cloud map of the orchard. The litchi trees in the orchard point cloud map are then identified using a pre-trained litchi tree segmentation model, wherein the litchi tree segmentation model is a neural network model that can improve the accuracy and efficiency of litchi tree recognition, obtaining a point cloud distribution of the target litchi tree. The point cloud distribution of the target litchi tree is then subjected to a box selection and marking operation based on the trunk, crown, and integer to obtain a bounding box set. Finally, information is extracted from the bounding box set using a point cloud slicing algorithm to obtain a digital information set of the litchi tree shape. Therefore, the embodiment of the present invention provides a method, device, equipment, and storage medium for digital reconstruction of the litchi tree shape, which can improve the efficiency and accuracy of the digitization of the litchi tree shape. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of a flow chart of a method for digitally reconstructing the appearance of a litchi tree provided in one embodiment of the present invention;

[0049] Figure 2 A schematic diagram of a detailed flow chart of the SLAM algorithm steps in the method for digitally reconstructing the appearance of a litchi tree provided in one embodiment of the present invention;

[0050] Figure 3A basic network structure schematic diagram of the litchi tree segmentation model in the litchi tree shape digital reconstruction method provided by an embodiment of the present application is shown in the figure.

[0051] Figure 4 A functional module diagram of the litchi tree shape digital reconstruction device provided by an embodiment of the present application is shown in the figure.

[0052] Figure 5 A structure schematic diagram of the electronic device for implementing the litchi tree shape digital reconstruction method provided by an embodiment of the present application is shown in the figure.

[0053] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0054] It should be understood that the specific embodiments described herein are merely intended to explain the present application and not to limit the present application.

[0055] The present application provides a litchi tree shape digital reconstruction method. In the present application, the execution subject of the litchi tree shape digital reconstruction method includes but is not limited to at least one of the electronic devices capable of being configured to execute the method provided by the present application, such as a server and a terminal. In other words, the litchi tree shape digital reconstruction method can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and big data and artificial intelligence platforms, etc. basic cloud computing services.

[0056] Referring to Figure 1 The figure shows a flowchart of the litchi tree shape digital reconstruction method provided by an embodiment of the present application. In the present embodiment, the litchi tree shape digital reconstruction method includes:

[0057] S1, using a pre-constructed point cloud acquisition device to scan the target litchi orchard environment to obtain point cloud information, and using a pre-constructed SLAM algorithm to construct an orchard point cloud map according to the point cloud information.

[0058] The point cloud acquisition device includes 16-line three-dimensional laser radar, 9-axis inertial measurement unit, GNSS receiver and embedded computer sensors.

[0059] Specifically, in an embodiment of the present invention, a laser radar, an inertial measurement unit, and a GNSS receiver are connected to an embedded computer to provide point cloud data, posture information, and GPS information for the SLAM algorithm. It should be noted that when assembling the acquisition device, it should be ensured that there are no obstacles within the vertical viewing angle (-15° to +15°) of the laser beam of the 16-line laser radar. In addition, in order to facilitate external parameter calibration with the laser radar, the inertial measurement unit is preferably installed directly below the laser radar, and then the lidar_align software is used for external parameter calibration. Finally, the acquisition personnel should hold the point cloud acquisition device to perform an environmental scan of the target litchi orchard to obtain point cloud information.

[0060] In detail, in an embodiment of the present invention, the method of constructing an orchard point cloud map based on the point cloud information using a pre-built SLAM algorithm includes:

[0061] Sorting the point cloud information according to the position of the point cloud acquisition device to obtain a single-frame point cloud sequence;

[0062] Using a pre-built inertial measurement unit to obtain the position and attitude of the point cloud acquisition device, and recording the position and attitude change sequence within a preset time period;

[0063] Acquire a high-definition garden image at a target node position of a motion trajectory of the point cloud acquisition device during the point cloud acquisition process;

[0064] According to the pre-built SLAM algorithm, based on the position and posture change sequence, the target node position and the high-definition garden image, each single-frame point cloud in the single-frame point cloud sequence is subjected to a point cloud deployment operation based on the world coordinate system to obtain a primary orchard point cloud map;

[0065] According to the pre-built loop detection algorithm, the primary orchard point cloud map is updated to obtain an orchard point cloud map.

[0066] In an embodiment of the present invention, when using the point cloud acquisition device, in order to make the collected point cloud of the litchi tree dense and accurate in shape, the acquisition personnel should hold the acquisition device and walk around the litchi tree to be collected to complete loop detection.

[0067] Specifically, in the embodiment of the present invention, during the execution of point cloud information collection, the laser radar, inertial measurement unit, and GNSS receiver transmit data in real time to the embedded computer. The SLAM algorithm based on the ROS operating system is deployed in advance on the embedded computer. The SLAM algorithm includes five modules: point cloud dedistortion, feature extraction, graph optimization, inertial measurement unit pre-integration, and map publishing. The working process of the five modules is as follows: Figure 2 shown.

[0068] In this example, a velocity model for the inertial measurement unit and a position and pose model for the point cloud acquisition device are constructed. These two models are used to estimate sensor motion during lidar scanning, providing a good initial pose estimate for the acquisition device. This improves the accuracy of inter-frame pose estimation of point clouds in orchard environments. This motion estimate is not only used for point cloud skew correction but also serves as the initial pose estimate during lidar odometry optimization. The lidar odometry method utilizes lidar for recursive pose state estimation.

[0069] Specifically, in the embodiment of the present invention, a specific implementation method of constructing an orchard point cloud map based on the point cloud information using a pre-built SLAM algorithm is as follows:

[0070] Step 1: First, based on the high-frequency inertial measurement unit data, calculate the lidar position of each point in a frame of point cloud at the imaging time to obtain a single-frame point cloud sequence. Then use this information to correct the coordinate system of all points to the starting point. Specifically, first use the lidar odometer to construct a primary orchard point cloud map based on the point cloud information collected in real time. Then determine the optimized pose from the graph optimization module, use the optimized pose as the starting pose of each frame, and use high-frequency inertial measurement unit data for integration to obtain a dense time-pose correspondence table. Using the time-pose correspondence table, perform a rigid body transformation on each point in each frame of point cloud collected in real time. The point cloud after rigid body transformation is the point cloud after dedistortion;

[0071] Step 2: Extract line and surface feature points from the dedistorted point cloud. These feature points are used for subsequent matching and mapping. Furthermore, high-definition garden images can be used for auxiliary pose alignment to improve the accuracy of the primary orchard point cloud map.

[0072] Step 3: Obtain relevant constraints on the pose from multiple aspects, including the relative position constraint obtained by the inertial measurement unit pre-integration between two consecutive point cloud frames in the single-frame point cloud sequence, the position constraint obtained by point cloud matching, the position constraint obtained by loop detection, the position constraint obtained by GPS, and the inertial measurement unit constraint. According to the above constraints, the optimal pose estimation is obtained through the graph optimization module. Specifically, if it is the first frame of point cloud, the position initialization is performed; if the time interval between the current frame point cloud and the previous frame point cloud reaches the threshold, the laser radar odometer is executed, otherwise it is skipped directly; the first few key frames of the point cloud data collected by the current acquisition device are selected to construct a local point cloud map; the current feature point cloud is downsampled; the downsampled feature point cloud is matched to the local orchard point cloud map; based on the relative position constraint obtained by the inertial measurement unit pre-integration between two consecutive point cloud frames, the position constraint obtained by point cloud matching, the position constraint obtained by loop detection, and the position constraint obtained by GPS, it is determined whether the point cloud data collected by the current acquisition device is a key frame. If it is a key frame, the pose estimation results of all key frames are updated;

[0073] Step 4: Estimate the accurate IMU pose from the raw IMU data. Specifically, the velocity and pose of the acquisition device are inferred based on the velocity model of the IMU and the velocity model of the point cloud acquisition device, and the IMU constraints are obtained. Return to step 3. The velocity model of the IMU is expressed as follows:

[0074]

[0075] Where, represents the angular velocity of the inertial measurement unit at time t in the body coordinate system B of the point cloud acquisition device, represents the acceleration of the inertial measurement unit at time t in the body coordinate system B of the point cloud acquisition device, ω t represents the actual measured angular velocity, represents the angular velocity deviation of the point cloud acquisition device at time t, represents the angular velocity white noise of the point cloud acquisition device at time t, represents the rotation matrix from the world coordinate system to the body coordinate system B of the point cloud acquisition device, α t represents the actual measured acceleration, g represents the gravity vector, represents the acceleration deviation of the point cloud acquisition device in the body coordinate system B at time t, represents the acceleration white noise of the point cloud acquisition device at time t.

[0076] The position and posture of the point cloud acquisition device is expressed as:

[0077]

[0078] In the formula, R t represents a rotation matrix Delta t represents a period of time, v t+Δt represents the speed of the point cloud collection device at t+Delta t, p t+Δt represents the position of the point cloud collection device at t+Delta t, R t+Δt represents the rotation matrix of the point cloud collection device at t+Delta t.

[0079] Step 5: The map publishing module updates the primary orchard point cloud map from the output of step 2 continuously, and finally forms a complete orchard point cloud map.

[0080] S2, the pre-trained litchi tree segmentation model is used for litchi tree identification on the orchard point cloud map, and target litchi tree point cloud distribution is obtained.

[0081] In the embodiment of the application, the litchi tree segmentation model is obtained by training a basic network model in a transfer learning manner, and reference Figure 3 As shown in the figure, the basic network model can be a pointnet++ network model.

[0082] The PointNet++ network consists of an input point cloud, a set abstraction layer, a grouping layer, a pointnet layer, a feature propagation layer, an output layer, a loss calculation layer, and an optimizer. The set abstraction layer includes Farthest Point Sampling (FPS), which selects a subset of points from the input point cloud as center points to ensure that the sampled points are evenly distributed within the point cloud. The number of points after sampling is usually less than the number of original points; the grouping layer includes grouping points within a fixed radius of K-nearest neighbors (KNN) or a fixed radius to form a local point cloud subset; the feature extraction layer includes a multi-layer perceptron (MLP): a series of fully connected layer transformations are performed on the features of each point, usually including ReLU activation function and batch normalization (Batch Normalization) and maximum pooling (Max Pooling): a global maximum pooling operation is performed on the point features of the local point cloud subset to obtain a global description of the local features; the feature propagation layer includes interpolation (Interpolating): an interpolation method (such as nearest neighbor interpolation or weighted interpolation) is used to propagate the features in the abstract layer back to the point cloud of the original resolution. The purpose of interpolation is to propagate fewer central point features to all points. Multilayer Perceptron (MLP): This further processes the interpolated features, transforming and enhancing them using a series of fully connected layers. Loss calculation selects an appropriate loss function based on the specific task, such as cross-entropy loss for classification tasks and smooth L1 loss for regression tasks. The optimizer uses an optimization algorithm to update network parameters to minimize the loss function.

[0083] like Figure 3As shown, the input is three-dimensional point cloud data, which can be represented as an N x D matrix, where N is the number of points and D is the dimension of the points (including three-dimensional space coordinates, normal vector information and color information). The PointNet++ network performs a Farthest Point Sampling (FPS) operation on the input point cloud data to select part of the points as center points, and then performs grouping operation on one branch (for the sake of distinction, we call this grouping operation the first grouping operation) and performs a multi-layer perceptron (MLP) operation on another branch (for the same reason, we call this MLP operation the first MLP operation). On the branch of the second grouping operation, for each center point, the points within a fixed radius range or K neighbors are grouped to form a local point cloud subset; on the branch of the second MLP operation, for each local point cloud subset, a small PointNet network is used for feature extraction, including MLP and max pooling operation, and the result after max pooling operation forms a higher level of local feature representation. On the result of the second MLP operation, FPS operation is performed again to select center points, and then grouping operation is performed on one branch (we call this grouping operation the third grouping operation) and multi-layer perceptron (MLP) operation is performed on another branch (we call this MLP operation the third MLP operation). On the branch of the third grouping operation, for each center point, the points within a fixed radius range or K neighbors are grouped to form a local point cloud subset; on the branch of the third MLP operation, for each local point cloud subset, a small PointNet network is used for feature extraction, including MLP and max pooling operation, and the result after max pooling operation forms a higher level of local feature representation. In the Feature Propagation Layer, the features in the SetAbstraction Layer 3 are interpolated to the point cloud in the SetAbstraction Layer 2, and processed through MLP, and the interpolated features are fused with the results of the second MLP operation; then the features in the Feature Propagation 1 are interpolated to the point cloud in the SetAbstraction Layer 1, and processed through MLP, and the interpolated features are fused with the results of the first MLP operation. The Output feature outputs the prediction result, and the prediction result is the segmentation result of the orchard point cloud map, such as obtaining the target litchi tree point cloud distribution and ground point cloud model.

[0084] In detail, in the embodiment of the present application, the litchi tree recognition of the orchard point cloud map by using the pre-trained litchi tree segmentation model to obtain the target litchi tree point cloud distribution comprises:

[0085] Obtaining point cloud sample data of similar litchi trees of the target litchi tree in the target litchi orchard;

[0086] Obtain a pre-built litchi tree segmentation model, and train the litchi tree segmentation model using the similar litchi tree point cloud sample data to obtain a pre-trained litchi tree segmentation model;

[0087] Obtaining artificially identified samples corresponding to the orchard point cloud map of the target litchi orchard, and fine-tuning the pre-trained litchi tree segmentation model using the artificially identified samples to obtain a trained litchi tree segmentation model;

[0088] The trained litchi tree segmentation model is used to identify litchi trees on the orchard point cloud map to obtain the target litchi tree point cloud distribution.

[0089] In one embodiment, the optimizer calculates the exponentially weighted moving average of the first-order moment (momentum) and second-order moment (unbiased variance) of the gradient to adjust the learning rate of each parameter and increase the accuracy of PointNet++ predictions. The specific formula is as follows:

[0090] m t =β1m t-1 +(1-β1)g t

[0091] v t =β2v t-1 +(1-β2)g t 2

[0092]

[0093] Among them, t is the current time; m t is the first-order moment estimate of the gradient, which is the exponentially weighted moving average of the gradient; v t is the second-order moment estimate of the gradient, which is the exponentially weighted moving average of the square of the gradient; β1 and β2 are the decay rates of the exponentially weighted moving average, usually set to 0.9 and 0.999 respectively; g t The gradient at the current time t represents the partial derivative of the loss function with respect to the parameter; and are bias-corrected estimates of the first and second moments. and are biased towards zero, so correction is needed; θ t is the parameter vector at the current time t; α is the learning rate, which controls the step size of each parameter update, and ∈ represents the deviation.

[0094] In the embodiment of the present application, the litchi tree segmentation model is trained through the above process to obtain a pre-trained litchi tree segmentation model, and then the trained litchi tree segmentation model is used to identify the target litchi tree point cloud distribution from the orchard point cloud map.

[0095] S3, performing object recognition operation on the target litchi tree point cloud distribution based on the trunk, the crown and the whole tree to obtain an object recognition result, and performing bounding box framing marking on the object recognition result to obtain a bounding box set, wherein the bounding box set includes a litchi trunk bounding box, a litchi crown layer bounding box and a litchi whole tree bounding box.

[0096] In detail, in the embodiment of the present application, the object recognition operation on the target litchi tree point cloud distribution based on the trunk, the crown and the whole tree to obtain an object recognition result, and performing bounding box framing marking on the object recognition result to obtain a bounding box set, includes:

[0097] Performing outlier removal on the target litchi tree point cloud distribution to obtain a converged litchi tree point cloud distribution;

[0098] Using a pre-constructed clustering algorithm, performing clustering grouping operation on the converged litchi tree point cloud distribution based on the trunk and the crown to obtain a trunk point cloud distribution and a crown point cloud distribution;

[0099] Using a principal component analysis algorithm, obtaining numerical features of the principal axis direction of the trunk point cloud distribution and the crown point cloud distribution based on the world coordinate system to obtain a numerical feature set, and constructing a minimum bounding box set according to the numerical feature set;

[0100] Obtaining a tree whole point cloud distribution of the trunk point cloud distribution and the crown point cloud distribution, and using the minimum bounding box set to frame mark the tree whole point cloud distribution, the trunk point cloud distribution and the crown point cloud distribution.

[0101] Specifically, after obtaining the target litchi tree point cloud distribution, the present application divides it by a clustering algorithm to divide out a litchi crown layer point cloud model and a litchi trunk point cloud model, uses a principal component analysis method to obtain feature vectors of three principal axis directions of the trunk model and the crown model of the target litchi tree point cloud distribution after removing outliers, finds the maximum and minimum value points of the model in the three axial directions to obtain the minimum bounding box of the model along the principal axis direction, wherein the length of the bounding box in the world coordinate system z axis is the model height.

[0102] The OBB algorithm was then used to construct the tree canopy bounding box, the trunk bounding box, and the overall bounding box, and the vertex coordinates of the bounding boxes were extracted. The height of the tree can be determined by calculating the overall bounding box's height; the maximum canopy width can be determined by calculating the canopy's canopy bounding box's width; the trunk length can be determined by calculating the trunk's trunk bounding box's height; and the canopy height can be determined by calculating the canopy's canopy bounding box's height.

[0103] S4. Using a pre-built point cloud slicing algorithm, extract information from the bounding box set to obtain a digital information set of the litchi tree's appearance.

[0104] Specifically, in an embodiment of the present invention, the method of extracting information from the bounding box set using a pre-built point cloud slicing algorithm to obtain a digitized information set of the litchi tree's shape includes:

[0105] Using a pre-built point cloud slicing algorithm, the point cloud distribution of the target litchi tree in the bounding box set is sliced ​​to obtain a canopy plane sequence;

[0106] Performing a two-dimensional convex hull operation on the canopy plane sequence to obtain a convex hull canopy plane, and obtaining the area and volume of the convex hull canopy plane to obtain the maximum canopy area and the maximum canopy volume;

[0107] Integrating the height corresponding to each canopy plane in the canopy plane sequence, the maximum canopy area and the maximum canopy volume corresponding to each convex hull canopy plane, to obtain an integer point cloud distribution volume corresponding to the target litchi tree point cloud distribution;

[0108] The maximum canopy area, the maximum canopy volume and the integer point cloud distribution volume are outputted simultaneously to obtain a digital information set of the litchi tree appearance.

[0109] Specifically, in an embodiment of the present invention, the target lychee tree point cloud distribution is cut using a preset slice spacing to obtain a canopy plane in slices, the canopy plane is two-dimensionally convex hulled, the area of ​​the canopy plane after the convex hull is calculated, and the maximum canopy area is obtained. Based on the area of ​​the canopy plane after the convex hull and the slice height, the volume of the corresponding height interval is obtained; according to the volume of each height interval, the volume of the target lychee tree point cloud distribution is obtained.

[0110] In one embodiment, the preset slice spacing is one centimeter, and the convex hull method used is the pcl:convexhull method in PCL (PointCloud Library) to form a two-dimensional convex hull, wherein the vertices of the convex hull are sorted in order of their polar angles to form a closed polygon. The area of ​​the polygon is the canopy plane area. After calculating the areas of each canopy plane, the largest canopy plane area is output as the maximum canopy area. The polygon area formula is as follows:

[0111]

[0112] Where A is the area, n is the number of vertices of the polygon, (x i ,y i ) are the coordinates of the i-th vertex, where i ranges from 1 to n.

[0113] Furthermore, the volume formula of the target litchi tree point cloud distribution is as follows:

[0114]

[0115] Among them, A i is the area of ​​the i-th canopy plane, V is the volume, n is the number of canopy planes, and h is the preset slice spacing height.

[0116] In one embodiment, the number of slices is dynamically adjusted based on the target litchi tree point cloud distribution density and height, rather than using a fixed number of slices. This allows more slices to be used in high-density areas and fewer slices to be used in low-density areas, thereby improving calculation accuracy.

[0117] The formula for calculating the number of points in the point cloud model at different height ranges is as follows:

[0118]

[0119] Among them, H (i) is the height interval of point i, and Δz is the height interval.

[0120] The formula for dynamically adjusting the number of slices in each height range based on height distribution and density is as follows:

[0121]

[0122] Among them, N s is the total number of canopy planes, K is the number of height intervals, d k is the number of points within the height interval k, D is the number of points in the target litchi tree point cloud distribution, N bThe number of basic slices in the canopy plane is an initial reference value used as a benchmark for dynamically adjusting the number of slices.

[0123] The point cloud model is segmented according to the dynamically adjusted number and height of slices. The formula is as follows:

[0124]

[0125] Among them, H max is the maximum height of the point cloud on the z-axis, H min is the minimum height of the point cloud on the z axis, N s is the total number of canopy planes.

[0126] The embodiment of the present invention obtains the point cloud data of the litchi orchard through the SLAM algorithm, and then performs point cloud segmentation through pointnet++ to segment the point cloud distribution of the target litchi tree in the orchard, and slices the segmented point cloud distribution of the target litchi tree and performs convex reconstruction, so as to quickly and conveniently obtain the digital information of the appearance of the litchi trees in the litchi orchard, providing data supplement for variable spray environment, yield prediction and quality control, improving agricultural efficiency, and providing technical support for the realization of precise pesticide spraying, crop health management, yield prediction and automated operations in smart agriculture.

[0127] The embodiment of the present invention first uses a point cloud acquisition device to perform an environmental scan of the target litchi orchard to initially obtain point cloud information. The point cloud information is then arranged using a SLAM algorithm to construct a point cloud map of the orchard. The SLAM algorithm can rearrange the serialized point cloud information according to the movement of the point cloud acquisition device to obtain a three-dimensional point cloud map of the orchard. The litchi trees in the orchard point cloud map are then identified using a pre-trained litchi tree segmentation model, wherein the litchi tree segmentation model is a neural network model that can improve the accuracy and efficiency of litchi tree recognition, obtain the target litchi tree point cloud distribution, and then perform a box selection and marking operation based on the trunk, crown, and integer on the target litchi tree point cloud distribution to obtain a bounding box set. Finally, information is extracted from the bounding box set using a point cloud slicing algorithm to obtain a digitized information set of the litchi tree shape. Therefore, the embodiment of the present invention provides a method for digitally reconstructing the shape of a litchi tree, which can improve the efficiency and accuracy of the digitization of the shape of the litchi tree.

[0128] like Figure 4 FIG. 1 is a functional module diagram of a device for digitally reconstructing the appearance of a litchi tree provided by an embodiment of the present invention.

[0129] The lychee tree shape digital reconstruction device 100 described in the present invention can be installed in an electronic device. Depending on the functionality to be implemented, the lychee tree shape digital reconstruction device 100 may include a point cloud acquisition module 101, a lychee tree recognition module 102, an object feature segmentation module 103, and a lychee tree shape digitization module 104. A module, also referred to as a unit, is a series of computer program segments that can be executed by a processor in an electronic device and perform a fixed function. These modules are stored in the memory of the electronic device.

[0130] In this embodiment, the functions of each module / unit are as follows:

[0131] The point cloud acquisition module 101 is used to scan the target litchi orchard using a pre-built point cloud acquisition device to obtain point cloud information, and to construct a point cloud map of the orchard based on the point cloud information using a pre-built SLAM algorithm;

[0132] The litchi tree identification module 102 is configured to identify litchi trees on the orchard point cloud map using a pre-trained litchi tree segmentation model to obtain a target litchi tree point cloud distribution;

[0133] The object feature segmentation module 103 is configured to perform object recognition operations on the target litchi tree point cloud distribution based on the trunk, crown, and entire tree to obtain an object recognition result, and perform bounding box selection and marking on the object recognition result to obtain a bounding box set, wherein the bounding box set includes a litchi trunk bounding box, a litchi crown bounding box, and a litchi tree overall bounding box;

[0134] The litchi tree shape digitization module 104 is configured to extract information from the bounding box set using a pre-built point cloud slicing algorithm to obtain a litchi tree shape digitization information set.

[0135] In detail, each module in the litchi tree appearance digital reconstruction device 100 described in the embodiment of the present application adopts the same method as above when in use. Figures 1 to 3 The same technical means are used as the method for digitally reconstructing the appearance of litchi trees described in , and can produce the same technical effects, so they will not be repeated here.

[0136] like Figure 5 FIG. 1 is a schematic structural diagram of an electronic device 1 for implementing a method for digitally reconstructing the appearance of a litchi tree provided by an embodiment of the present invention.

[0137] The electronic device 1 may include a processor 10, a memory 11, a communication bus 12, and a communication interface 13. It may also include a computer program stored in the memory 11 and executable on the processor 10, such as a program for digitally reconstructing the appearance of a litchi tree.

[0138] In some embodiments, the processor 10 may be composed of an integrated circuit, such as a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips. The processor 10 is the control core (Control Unit) of the electronic device 1, connecting the various components of the entire electronic device using various interfaces and circuits. It executes or runs programs or modules stored in the memory 11 (e.g., executing a program for digitally reconstructing the appearance of a lychee tree), and calls data stored in the memory 11 to perform various functions of the electronic device and process data.

[0139] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (for example, an SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of an electronic device, such as a mobile hard disk of the electronic device. In other embodiments, the memory 11 can also be an external storage device of an electronic device, such as a plug-in mobile hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card, etc. equipped on the electronic device. Furthermore, the memory 11 can also include both an internal storage unit of the electronic device and an external storage device. The memory 11 can not only be used to store application software and various types of data installed in the electronic device, such as the code of the lychee tree appearance digital reconstruction program, but can also be used to temporarily store data that has been output or is to be output.

[0140] The communication bus 12 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.

[0141] The communication interface 13 is used for communication between the above-mentioned electronic device 1 and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display (Display), an input unit (such as a keyboard (Keyboard), optionally, the user interface may also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device and to display a visual user interface.

[0142] Figure 5 Only the electronic device with components is shown, and it can be understood by those skilled in the art that Figure 5 The structure shown does not constitute a limitation on the electronic device 1 , and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0143] For example, although not shown, the electronic device 1 may further include a power source (such as a battery) for powering the various components. Preferably, the power source may be logically connected to the at least one processor 10 via a power management device, thereby implementing functions such as charging management, discharging management, and power consumption management through the power management device. The power source may further include any components such as one or more DC or AC power sources, a recharging device, a power failure detection circuit, a power converter or inverter, a power status indicator, etc. The electronic device 1 may further include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0144] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0145] The litchi tree shape digital reconstruction program stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When executed in the processor 10, it can achieve the following:

[0146] Use a pre-built point cloud acquisition device to scan the target litchi orchard to obtain point cloud information, and use a pre-built SLAM algorithm to construct a point cloud map of the orchard based on the point cloud information;

[0147] Using a pre-trained litchi tree segmentation model to identify litchi trees on the orchard point cloud map, and obtaining a target litchi tree point cloud distribution;

[0148] Performing an object recognition operation based on the trunk, crown, and entire tree on the target litchi tree point cloud distribution to obtain an object recognition result, and performing bounding box selection marking on the object recognition result to obtain a bounding box set, wherein the bounding box set includes a litchi trunk bounding box, a litchi crown bounding box, and an entire litchi tree bounding box;

[0149] A pre-built point cloud slicing algorithm is used to extract information from the bounding box set to obtain a digital information set of the litchi tree's appearance.

[0150] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, which will not be repeated here.

[0151] Furthermore, if the modules / units integrated in the electronic device 1 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0152] The present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed by a processor of an electronic device, the computer program can implement:

[0153] Use a pre-built point cloud acquisition device to scan the target litchi orchard to obtain point cloud information, and use a pre-built SLAM algorithm to construct a point cloud map of the orchard based on the point cloud information;

[0154] Using a pre-trained litchi tree segmentation model to identify litchi trees on the orchard point cloud map, and obtaining a target litchi tree point cloud distribution;

[0155] Performing an object recognition operation based on the trunk, crown, and entire tree on the target litchi tree point cloud distribution to obtain an object recognition result, and performing bounding box selection marking on the object recognition result to obtain a bounding box set, wherein the bounding box set includes a litchi trunk bounding box, a litchi crown bounding box, and an entire litchi tree bounding box;

[0156] A pre-built point cloud slicing algorithm is used to extract information from the bounding box set to obtain a digital information set of the litchi tree's appearance.

[0157] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the module division is merely a logical function division, and other division methods may be used in actual implementation.

[0158] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network elements. Some or all of the modules may be selected to achieve the purpose of the solution of this embodiment according to actual needs.

[0159] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or hardware plus software functional modules.

[0160] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0161] Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the claims are intended to be embraced therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0162] Blockchain, as used in this article, refers to a novel application model for computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Blockchain is essentially a decentralized database, a series of data blocks generated using cryptographic methods. Each block contains information about a batch of online transactions, used to verify the validity of this information (to prevent counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, the platform product service layer, and the application service layer.

[0163] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0164] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. Terms such as "first" and "second" are used to indicate names and do not imply any particular order.

[0165] 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 may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for digitally reconstructing the appearance of a litchi tree, characterized in that: The method comprises: Use a pre-built point cloud acquisition device to scan the target litchi orchard to obtain point cloud information, and use a pre-built SLAM algorithm to construct a point cloud map of the orchard based on the point cloud information; Using a pre-trained litchi tree segmentation model to identify litchi trees on the orchard point cloud map, and obtaining a target litchi tree point cloud distribution; Performing an object recognition operation based on the trunk, crown, and entire tree on the target litchi tree point cloud distribution to obtain an object recognition result, and performing bounding box selection marking on the object recognition result to obtain a bounding box set, wherein the bounding box set includes a litchi trunk bounding box, a litchi crown bounding box, and an entire litchi tree bounding box; Using a pre-built point cloud slicing algorithm, information is extracted from the bounding box set to obtain a digital information set of the litchi tree's shape; The method of using a pre-built SLAM algorithm to construct an orchard point cloud map based on the point cloud information includes: Sorting the point cloud information according to the position of the point cloud acquisition device to obtain a single-frame point cloud sequence; Using a pre-built inertial measurement unit to obtain the position and attitude of the point cloud acquisition device, and recording the position and attitude change sequence within a preset time period; Acquire a high-definition garden image at a target node position of a motion trajectory of the point cloud acquisition device during the point cloud acquisition process; According to the pre-built SLAM algorithm, based on the position and posture change sequence, the target node position and the high-definition garden image, each single-frame point cloud in the single-frame point cloud sequence is subjected to a point cloud deployment operation based on the world coordinate system to obtain a primary orchard point cloud map; According to the pre-built loop detection algorithm, the primary orchard point cloud map is updated to obtain an orchard point cloud map.

2. The digital reconstruction method for the appearance of a litchi tree as claimed in claim 1, wherein: The inertial measurement unit is expressed as: Where, Represents the inertial measurement unit The coordinate system of the point cloud acquisition device at the moment The angular velocity in Represents the inertial measurement unit The coordinate system of the point cloud acquisition device at the moment The acceleration in represents the actual measured angular velocity, Indicates that the point cloud acquisition device is The angular velocity bias at the moment, Indicates that the point cloud acquisition device is The angular velocity white noise at time , Represents the coordinate system from the world coordinate system to the body coordinate system of the point cloud acquisition device The rotation matrix of represents the actual measured acceleration, represents the gravity vector, Indicates that the point cloud acquisition device is The body coordinate system at the moment The acceleration bias in Indicates that the point cloud acquisition device is The acceleration white noise at the moment.

3. The digital reconstruction method for litchi tree appearance as claimed in claim 2, wherein: The position and posture of the point cloud acquisition device is expressed as: In the formula, Represents the rotation matrix , Indicates a period of time, Indicates that the point cloud acquisition device is The speed of time, Indicates that the point cloud acquisition device is The location at the moment, Indicates that the point cloud acquisition device is The rotation matrix at time .

4. The digital reconstruction method for the appearance of a litchi tree as claimed in claim 3, wherein: The method of using the pre-trained litchi tree segmentation model to identify litchi trees on the orchard point cloud map to obtain the target litchi tree point cloud distribution includes: Obtaining point cloud sample data of similar litchi trees of the target litchi tree in the target litchi orchard; Obtain a pre-built litchi tree segmentation model, and train the litchi tree segmentation model using the similar litchi tree point cloud sample data to obtain a pre-trained litchi tree segmentation model; Obtaining manually identified samples corresponding to the orchard point cloud map of the target litchi orchard, and fine-tuning the pre-trained litchi tree segmentation model using the manually identified samples to obtain a trained litchi tree segmentation model; The trained litchi tree segmentation model is used to identify litchi trees on the orchard point cloud map to obtain the target litchi tree point cloud distribution.

5. The digital reconstruction method for the appearance of a litchi tree as claimed in claim 4, wherein: The object recognition operation based on the trunk, crown and the entire tree is performed on the target litchi tree point cloud distribution to obtain an object recognition result, and the bounding box selection mark is performed on the object recognition result to obtain a bounding box set, including: performing outlier removal on the target litchi tree point cloud distribution to obtain a converged litchi tree point cloud distribution; Using a pre-built clustering algorithm, clustering and grouping operations based on trunks and crowns are performed on the convergent litchi tree point cloud distribution to obtain trunk point cloud distribution and crown point cloud distribution; Using a principal component analysis algorithm, the numerical features of the trunk point cloud distribution and the crown point cloud distribution based on the principal axis direction of the world coordinate system are obtained to obtain a numerical feature set, and a minimum bounding box set is constructed based on the numerical feature set; The entire tree point cloud distribution of the trunk point cloud distribution and the crown point cloud distribution is obtained, and the entire tree point cloud distribution, the trunk point cloud distribution and the crown point cloud distribution are framed and marked using the minimum bounding box set.

6. The method for digitally reconstructing the appearance of a litchi tree as claimed in claim 5, wherein: The pre-built point cloud slicing algorithm is used to extract information from the bounding box set to obtain a digitized information set of the litchi tree shape, including: Using a pre-built point cloud slicing algorithm, the point cloud distribution of the target litchi tree in the bounding box set is sliced ​​to obtain a canopy plane sequence; Performing a two-dimensional convex hull operation on the canopy plane sequence to obtain a convex hull canopy plane, and obtaining the area and volume of the convex hull canopy plane to obtain the maximum canopy area and the maximum canopy volume; Integrating the height corresponding to each canopy plane in the canopy plane sequence, the maximum canopy area and the maximum canopy volume corresponding to each convex hull canopy plane, to obtain an integer point cloud distribution volume corresponding to the target litchi tree point cloud distribution; The maximum canopy area, the maximum canopy volume and the integer point cloud distribution volume are outputted simultaneously to obtain a digital information set of the litchi tree appearance.

7. A device for digitally reconstructing the appearance of a litchi tree, characterized in that: The device comprises: The point cloud acquisition module is used to use a pre-built point cloud acquisition device to perform an environmental scan of the target litchi orchard to obtain point cloud information, and sort the point cloud information according to the position of the point cloud acquisition device to obtain a single-frame point cloud sequence; use a pre-built inertial measurement unit to obtain the position and posture of the point cloud acquisition device, and record the position and posture change sequence within a preset time period; obtain a high-definition garden image at the target node position of the motion trajectory of the point cloud acquisition device during the point cloud acquisition process; according to the pre-built SLAM algorithm, according to the position and posture change sequence, the target node position and the high-definition garden image, perform a point cloud deployment operation based on the world coordinate system on each single-frame point cloud in the single-frame point cloud sequence to obtain a primary orchard point cloud map; according to the pre-built loop detection algorithm, update the primary orchard point cloud map to obtain an orchard point cloud map; A litchi tree recognition module is used to identify litchi trees on the orchard point cloud map using a pre-trained litchi tree segmentation model to obtain a target litchi tree point cloud distribution; An object feature segmentation module is configured to perform object recognition operations on the target litchi tree point cloud distribution based on the trunk, crown, and entire tree to obtain an object recognition result, and to perform bounding box selection and marking on the object recognition result to obtain a bounding box set, wherein the bounding box set includes a litchi trunk bounding box, a litchi crown bounding box, and a litchi tree overall bounding box; The litchi tree shape digitization module is used to extract information from the bounding box set using a pre-built point cloud slicing algorithm to obtain a litchi tree shape digitization information set.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for digitally reconstructing the appearance of a litchi tree according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for digitally reconstructing the appearance of a litchi tree according to any one of claims 1 to 6 is implemented.

Images