Method and device for generating prescription map of orchard variable rate application

Abstract

The present application relates to the technical field of variable pesticide application, and provides a method and device for generating an orchard variable pesticide application prescription map, comprising: collecting point cloud information of an orchard environment and latitude and longitude information of a calibration ball in the orchard by using a scanning device; constructing a three-dimensional point cloud map of the orchard and filtering ground point clouds to obtain a three-dimensional point cloud model of fruit trees based on the point cloud information by using a SLAM algorithm and a SLAM graph optimization model during the collection process; segmenting the three-dimensional point cloud model of the fruit trees by using a clustering algorithm to construct a bounding box and determine latitude and longitude information of the fruit trees, and reconstructing the three-dimensional point cloud model of the fruit trees to obtain crown layer partition volume and crown layer partition leaf area density of the fruit trees; calculating pesticide application prescription values of each crown layer partition based on the crown layer partition volume and the crown layer partition leaf area density to obtain fruit tree variable pesticide application prescription values; and importing the latitude and longitude information of the calibration ball, the three-dimensional point cloud map of the orchard, the latitude and longitude information of the fruit trees, and the fruit tree variable pesticide application prescription values into a GIS system to generate an orchard variable pesticide application prescription map. The method can meet diversified pesticide application requirements.

Description

Technical Field

[0001] This invention belongs to the field of variable application technology, and particularly relates to a method, apparatus, computer equipment, and storage medium for generating variable application prescription maps in orchards. Background Technology

[0002] The fruit industry is one of the most important industries in many regions of my country, providing a significant driving force for agricultural economic development. Traditional orchard pesticide application methods generally adopt an extensive continuous spraying strategy. However, this method results in less than 30% effective pesticide deposition on the target fruit trees, leading to low pesticide utilization, serious droplet drift and residue, and easily causing environmental pollution, fruit quality decline, and ecosystem imbalance.

[0003] To address the aforementioned issues and achieve precision pesticide application in orchards, variable-rate application techniques are currently widely used. However, existing variable-rate application techniques often rely on real-time sensors to detect the location, volume, and leaf area density of the fruit tree canopy. This leads to several problems: high performance requirements for non-contact sensors, resulting in a certain delay in application response time. Furthermore, sprayer route deviations can prevent the acquisition of accurate canopy information, and the limited range of variable decision-making features fails to meet the diverse application needs of pesticide applications. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, and storage medium for orchard variable application prescription maps that can improve response speed and meet diverse application needs, in order to address the above-mentioned technical problems.

[0005] This invention provides a method for creating a variable-rate pesticide application prescription map in an orchard, comprising:

[0006] The scanning device was used to collect point cloud information of the orchard environment and latitude and longitude information of the calibration spheres in the orchard;

[0007] During the data collection process, the SLAM algorithm and SLAM graph optimization model are used to construct a three-dimensional point cloud map of the orchard based on the point cloud information, and the ground point cloud in the three-dimensional point cloud map of the orchard is filtered to obtain a three-dimensional point cloud model of the fruit trees.

[0008] Clustering algorithm is used to segment the three-dimensional point cloud model of the fruit tree to construct the bounding box of the three-dimensional point cloud model of the fruit tree. Based on the bounding box, the latitude and longitude information of the fruit tree is determined, and the canopy partition volume and canopy partition leaf area density of the fruit tree are obtained by reconstructing the three-dimensional point cloud model of the fruit tree.

[0009] The pesticide prescription value for each canopy zone is calculated based on the canopy zone volume and the canopy zone leaf area density to obtain the fruit tree variable pesticide prescription value;

[0010] The latitude and longitude information of the calibration ball, the three-dimensional point cloud map of the orchard, the latitude and longitude information of the fruit trees, and the variable application prescription values ​​of the fruit trees are imported into the GIS system to generate a variable application prescription map of the orchard.

[0011] In one embodiment, the step of constructing a 3D point cloud map of the orchard based on the point cloud information using the SLAM algorithm and SLAM graph optimization model during the acquisition process includes:

[0012] During the acquisition process, the acquired point cloud information is reprojected into a depth image, and the ground point cloud is filtered based on the angle between adjacent points in the depth image, and the point cloud with neighboring points is calculated by traversing and filtering out noisy point clouds.

[0013] The pose of the point cloud is estimated using a laser odometry model. The estimated pose is then combined with the voxel mesh method to store the point cloud in the world coordinate system, resulting in a 3D point cloud map of the orchard with geographical location information.

[0014] The orchard's 3D point cloud map was optimized using a backend graph optimization model.

[0015] In one embodiment, the laser odometry model is represented as follows:

[0016]

[0017]

[0018] Where X represents the pose; This represents the key point cloud extracted from the point cloud of the new frame; ρ(s) is the loss function; lerp represents the interpolation function; n i yes The normal of the neighborhood; These are lidar measurements; It is a local point cloud map; α i It is the linear interpolated pose between the start pose and the end pose, a i (s) is the neighborhood planarization function; C loc (X) is the location consistency function, expressed as: C vel (X) is the velocity uniformity function, expressed as: R b and t b It is the rotation and translation of the starting pose, R e and t e It is the rotation and translation that ends the pose;

[0019] The backend graph optimization model is represented as follows:

[0020]

[0021] in, and It is a tree trunk point cloud factor. C n It indicates that the tree trunks are clustered together.

[0022] In one embodiment, the tree trunk point cloud is obtained by extracting the feature point cloud at breast height (DBH) of the fruit trees in the orchard's 3D point cloud map using a Euclidean clustering algorithm, and then segmenting the feature point clouds belonging to the same fruit tree DBH using a judgment formula, which is as follows:

[0023]

[0024] Where, Δd ij It is the distance between the centers of feature point clouds i and j; Δd max Δt is the maximum distance between the centers of feature point clouds i and j; r is the radius of the feature circle; Δt ij is the observation time difference between characteristic circles i and j; a and b are constants.

[0025] In one embodiment, the step of filtering the ground point cloud in the orchard 3D point cloud map to obtain the fruit tree 3D point cloud model includes:

[0026] The ground point cloud in the orchard 3D point cloud map is filtered based on the rotation constraint method and the installation height of the lidar in the scanning device, or the ground point cloud in the orchard 3D point cloud map is fitted and filtered to obtain the fruit tree 3D point cloud model.

[0027] In one embodiment, the formula for calculating the canopy zoned drug application prescription value is as follows:

[0028]

[0029] in, ρ is the pesticide prescription value for canopy zone i of fruit tree n; leaf V represents the leaf area density of canopy zone i; v represents the moving speed of the spraying device; a and b represent the spraying flow rate coefficients; V canopy Let i be the volume of the fruit tree canopy.

[0030] In one embodiment, the step of importing the latitude and longitude information of the calibration ball, the orchard 3D point cloud map, the fruit tree latitude and longitude information, and the fruit tree variable application prescription values ​​into the GIS system to generate an orchard variable application prescription map includes:

[0031] Based on the location of the calibration sphere point cloud model in the orchard 3D point cloud map and the geographical location in the satellite map corresponding to the latitude and longitude information of the calibration sphere, the satellite map and the orchard 3D point cloud map are registered to obtain a geographic information layer and a land feature spatial layer.

[0032] XYZ / prescription value data is created based on the latitude and longitude information of the fruit tree and the prescription values ​​of each canopy partition in the variable prescription values ​​of the fruit tree. Based on the XYZ / prescription value data, a columnar volumetric raster is constructed using a spatial interpolation algorithm. The columnar volumetric raster is used to represent the fruit tree canopy model, and the height of the columnar volumetric raster is adjusted to obtain a volumetric raster layer.

[0033] Import the voxel raster layer, the geographic information layer, and the land feature spatial layer into the GIS workspace to obtain the orchard variable application prescription map.

[0034] A variable-rate pesticide application prescription map device for orchards, comprising:

[0035] The data acquisition module is used to collect point cloud information of the orchard environment and latitude and longitude information of the calibration sphere in the orchard using a scanning device;

[0036] The mapping module is used to construct a 3D point cloud map of the orchard based on the point cloud information using the SLAM algorithm and SLAM graph optimization model during the acquisition process, and to filter the ground point cloud in the 3D point cloud map of the orchard to obtain a 3D point cloud model of the fruit trees.

[0037] The reconstruction module is used to segment the three-dimensional point cloud model of the fruit tree using a clustering algorithm to construct the bounding box of the three-dimensional point cloud model of the fruit tree, determine the latitude and longitude information of the fruit tree based on the bounding box, and reconstruct the three-dimensional point cloud model of the fruit tree to obtain the canopy partition volume and canopy partition leaf area density of the fruit tree.

[0038] The pesticide application calculation module is used to calculate the pesticide application prescription value for each canopy zone based on the canopy zone volume and the canopy zone leaf area density, and obtain the fruit tree variable pesticide application prescription value;

[0039] The generation module is used to import the latitude and longitude information of the calibration ball, the three-dimensional point cloud map of the orchard, the latitude and longitude information of the fruit trees, and the variable pesticide application prescription values ​​of the fruit trees into the GIS system to generate a variable pesticide application prescription map of the orchard.

[0040] The present invention also provides a computer device, the computer device including a processor and a memory, the memory storing a computer program, the processor executing the computer program to implement the steps of the orchard variable application prescription map method described in any of the above claims.

[0041] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the orchard variable application prescription map method described in any of the preceding claims.

[0042] The aforementioned orchard variable-rate pesticide application map method, apparatus, computer equipment, and storage medium efficiently acquire large-scale fruit tree canopy information such as canopy volume and leaf area density using SLAM algorithms and improved SLAM map optimization models. This canopy information is then used to calculate the required pesticide dosage for the target fruit trees, thereby improving operational efficiency, meeting diverse pesticide application needs, and increasing response speed. Furthermore, using 3D GIS technology to create orchard variable-rate pesticide application maps can generate maps with geographical features, land cover features, and pesticide application characteristics, enabling precise targeted pesticide application. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating a variable application prescription method for orchards in one embodiment.

[0044] Figure 2 This is a schematic diagram of the scanning method and the placement of the calibration ball in one embodiment of the scanning device.

[0045] Figure 3 This is a schematic diagram of the graph optimization model of the SLAM algorithm in one embodiment.

[0046] Figure 4 This is a schematic diagram of the Bounding Box of a fruit tree point cloud model in one embodiment.

[0047] Figure 5 This is a schematic diagram of reconstructing a point cloud model of a fruit tree in one embodiment.

[0048] Figure 6 This is a schematic diagram of reconstructing a point cloud model of a fruit tree in one embodiment.

[0049] Figure 7 This is a structural block diagram of an orchard variable pesticide application prescription map device in one embodiment. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0051] In one embodiment, such as Figure 1 As shown, a method for generating a variable-rate pesticide application prescription map in an orchard is provided, including the following steps:

[0052] Step S101: Use a scanning device to collect point cloud information of the orchard environment and latitude and longitude information of the calibration ball in the orchard.

[0053] Specifically, a scanning device for scanning the orchard environment is first built on a self-propelled mobile robot platform. This device includes sensor hardware such as a 16-line 3D LiDAR, a 9-axis IMU, a GNSS receiver, and an NVIDIA Jeston embedded computer system. The GNSS receiver is connected to the 3D LiDAR junction box and uses network RTK differential services to compute GNSS signals, obtaining RTK-GNSS positioning information to provide GNSS factors for the SLAM graph optimization model. The IMU is connected to the embedded computer to measure the pose information of the mobile robot platform, providing IMU factors for the SLAM graph optimization model. It is important to ensure that there are no obstacles within the vertical field of view (-15° to +15°) of the 16-line LiDAR's laser beam during hardware installation. Furthermore, when installing the 9-axis IMU, to facilitate extrinsic parameter calibration with the LiDAR, the IMU is preferably installed directly below the LiDAR and calibrated using the `lidar_align` tool.

[0054] Then, the scanning device is used to collect point cloud information of the structured orchard environment. To ensure a uniform and dense distribution of the fruit tree point cloud, the scanning device moves in a circular motion as it passes over the fruit trees to collect data. The specific scanning method is as follows: Figure 2 As shown. The calibration balls are pre-set at the boundaries of the structured orchard plant protection operation area. The installation of the calibration balls can be found in [reference needed]. Figure 2 As shown, GNSS equipment is used to obtain the latitude and longitude information of a calibration sphere when scanning the orchard environment. In practical applications, laser reflective markers can also be used instead of calibration spheres, and obtaining the latitude and longitude information of the laser reflective markers can achieve the same effect.

[0055] Step S102: During the data acquisition process, a 3D point cloud map of the orchard is constructed based on point cloud information using the SLAM algorithm and SLAM graph optimization model. The ground point cloud in the 3D point cloud map of the orchard is filtered to obtain a 3D point cloud model of the fruit trees.

[0056] Specifically, during the data acquisition process, specifically step S101, a preliminary 3D orchard point cloud map can be obtained by using a front-end odometry model to construct a local point cloud map in real time based on the acquired point cloud information. After acquisition, the back-end graph optimization model is used to optimize the 3D orchard point cloud map. That is, the SLAM algorithm is run on the ROS operating system in the NVIDIA Jeston embedded computer system to construct a SLAM graph optimization model to obtain a globally consistent 3D orchard point cloud map with geographic location information. Then, the ground point cloud in the 3D orchard point cloud map is filtered out to obtain a 3D point cloud model of the fruit trees. In this embodiment, addressing the difficulties in extracting point cloud feature information and the poor mapping accuracy of the point cloud map in dynamic orchard scenarios using the LiDAR-SLAM algorithm, a laser odometry model with elastic features is constructed based on the principle of pose continuity. Laser odometry factors and tree trunk point cloud factors are introduced to construct a back-end graph optimization model to improve the estimation accuracy of point cloud pose between frames in dynamic environments, thereby improving the mapping robustness of the SLAM algorithm in structured orchard environments.

[0057] In one embodiment, during the data acquisition process, constructing a 3D point cloud map of the orchard based on point cloud information using the SLAM algorithm and SLAM graph optimization model includes the following steps:

[0058] Step 1: During the acquisition process, the acquired point cloud information is reprojected into a depth image. The ground point cloud is filtered based on the angle between adjacent points in the depth image, and the point cloud with neighboring points is calculated by traversing the point cloud to filter out noisy point clouds.

[0059] Specifically, in this step, the lidar scans at a rotation rate of 10Hz to acquire point cloud information of the orchard environment, and the 9-axis IMU and GNSS receiver sample at 10kHz and 200Hz respectively. After sampling, the point cloud information is preprocessed, which in this embodiment mainly includes segmenting the ground point cloud and filtering environmental noise. First, the original point cloud information is reprojected into a depth image, which can be represented by an array, where the coordinates are x and y, and z is converted into depth information. Then, it is determined whether it is a ground point cloud by judging whether the angle between adjacent points in the depth image is less than 10°. If the angle is less than 10°, it is determined to be a ground point cloud and filtered. After filtering the ground point cloud, a depth-first traversal is performed on the depth image, that is, starting from the coordinates [0, 0], a frame of point cloud is traversed. By traversing the four neighboring points of a certain point, the angle between it and the neighboring points is calculated. If the angle is greater than 60°, it is considered to be the same point cloud set, and if the number of point clouds in a point cloud set is less than 30, it is judged as a noise point cloud and filtered.

[0060] Step 2: Optimize the pose of the point cloud using the laser odometry model, and store the point cloud in the world coordinate system based on the estimated pose to obtain a 3D point cloud map of the orchard with geographical location information.

[0061] Specifically, this step constructs a laser odometry model with elastic characteristics based on the continuity of pose during a single frame of data acquisition and the discontinuity of pose during inter-frame data acquisition. Using the front-end laser odometry model, the first frame acquired by the lidar is used as the original local point cloud map, and the starting pose of a single frame of point cloud data is employed. and ending pose To estimate inter-frame pose, the pose at intermediate time points is processed using a linear interpolation algorithm. Assume the starting pose of the nth frame is... The final pose of frame n-1 is Theoretically, the two poses are identical, so the proximity-constraint algorithm is first used to process the proximity of the poses. Then, a sampling algorithm is used to extract the local point cloud map. And a new frame of point cloud Partial point cloud As a key point cloud, pose optimization is performed based on the key point cloud to estimate the pose change between the new frame point cloud and the local point cloud map. The laser odometry model in this embodiment is represented as follows:

[0062]

[0063]

[0064] Where X represents the pose; This represents the key point cloud extracted from the point cloud of the new frame; ρ(s) is the loss function; lerp represents the interpolation function; n i yes The normal of the neighborhood; These are lidar measurements; It is a local point cloud map; α i It is the linear interpolated pose between the start pose and the end pose, a i (s) is the neighborhood planarization function; C loc (X) is the location consistency function, expressed as: C vel (X) is the velocity uniformity function, expressed as: R b and t b It is the rotation and translation of the starting pose, R e and t e It is the rotation and translation of the final pose.

[0065] Then, a voxel grid method is used to store the world coordinate system point cloud map, with each grid storing a maximum of 20 points. This enables the construction of a globally consistent 3D point cloud map of the orchard with geographical location information based on inter-frame pose estimation. The process of constructing the 3D point cloud map of the orchard using the front-end odometry can be simply understood as follows: First, the first frame of point cloud acquired is used as the original local point cloud map. The next frame, i.e., the second frame, is used as the new frame of point cloud for pose estimation to obtain a point cloud map. This point cloud map is then used as the new local point cloud map. The next frame, i.e., the third frame, is used as the new frame of point cloud for pose estimation to obtain a new local point cloud map, and so on. Each subsequent frame of point cloud acquired is subjected to pose estimation in this way. The point cloud map obtained after the pose estimation of the last frame is completed is the preliminary 3D point cloud map of the orchard. At the same time, during the pose estimation process, it is also necessary to perform proximity processing between the start pose of each frame of point cloud and the end pose of the corresponding previous frame of point cloud.

[0066] Step 3: Optimize the orchard 3D point cloud map using the backend graph optimization model.

[0067] Specifically, after obtaining the preliminary 3D point cloud map of the orchard in step 2, a backend graph optimization model is established. In this embodiment, the SLAM graph optimization model is as follows: Figure 3 As shown, the pose estimation error is minimized through a backend graph optimization model to improve mapping accuracy. The graph optimization model in this embodiment mainly includes laser odometry factors, tree trunk point cloud factors, IMU factors, and GNSS factors. The backend graph optimization model is represented as follows:

[0068]

[0069] in, and It is a tree trunk point cloud factor. C n It indicates that the tree trunks are clustered together.

[0070] In one embodiment, the tree trunk point cloud C n The feature point cloud at breast height of the fruit trees in the 3D point cloud map of the orchard was extracted using the Euclidean clustering algorithm, and the distance Δd between the centers of the feature point clouds was calculated. ij The difference between the radius and the feature point cloud at breast height is used to determine whether they belong to the same fruit tree. If so, they are segmented into the same trunk point cloud set C. n .

[0071]

[0072] Where, Δd ij It is the distance between the centers of feature point clouds i and j; Δd max Δt is the maximum distance between the centers of feature point clouds i and j; r is the radius of the feature circle; Δt ijis the observation time difference between characteristic circles i and j; a and b are constants.

[0073] In one embodiment, filtering the ground point cloud in the orchard 3D point cloud map to obtain the fruit tree 3D point cloud model includes: filtering the ground point cloud in the orchard 3D point cloud map based on the rotation constraint method and the installation height of the lidar in the scanning device, or fitting and filtering the ground point cloud in the orchard 3D point cloud map to obtain the fruit tree 3D point cloud model.

[0074] Specifically, after obtaining a globally consistent 3D point cloud map of the orchard with geographic location information, if the orchard has a high degree of structure, the ground point cloud is filtered based on the rotation constraint method and the installation height of the LiDAR to obtain a 3D point cloud model of the fruit trees. If the orchard has a low degree of structure, algorithms such as PLS and RANSAC can be used to fit the ground point cloud and filter it. After filtering out the ground point cloud in the 3D point cloud map of the orchard, a large-scale 3D point cloud model of the fruit trees can be obtained.

[0075] Step S103: Use clustering algorithm to segment the three-dimensional point cloud model of the fruit tree and construct the bounding box of the three-dimensional point cloud model of the fruit tree. Determine the latitude and longitude information of the fruit tree based on the bounding box, and reconstruct the three-dimensional point cloud model of the fruit tree to obtain the canopy partition volume and canopy partition leaf area density of the fruit tree.

[0076] Specifically, after obtaining the 3D point cloud model of the fruit tree, it is segmented using a clustering algorithm. The OBB algorithm is then used to construct the bounding box of the fruit tree point cloud model. The vertex coordinates A1A2B1B2 of the bounding box are extracted as the latitude and longitude information of the fruit tree. Figure 4 As shown. Then, the point cloud model of the fruit tree is reconstructed using algorithms such as Alpha-Shape or convex hull. Figure 5 As shown, the fruit tree canopy is divided into three zones: upper, middle, and lower. The height of the fruit tree can be obtained by calculating the height of the fruit tree point cloud bounding box. The volume of the canopy zone is obtained based on the reconstructed fruit tree point cloud model. The leaf area density of the canopy zone can be obtained based on the functional relationship between the number of point clouds in the canopy zone and the leaf area density of the canopy, thus realizing the acquisition of fruit tree canopy information.

[0077] Step S104: Calculate the pesticide prescription value for each canopy zone based on the canopy zone volume and canopy zone leaf area density to obtain the variable pesticide prescription value for fruit trees.

[0078] Specifically, a pesticide application rate model for fruit trees is introduced. Based on the canopy zone volume and leaf area density, the pesticide application prescription value for each canopy zone is calculated to obtain the variable pesticide application prescription value for fruit trees. The pesticide application rate model for fruit trees is as follows:

[0079]

[0080] in, ρ represents the pesticide prescription value for canopy zone i of fruit tree n, in L; leaf Leaf area density of canopy i zone, in m³ 2 / m 3 v represents the moving speed of the spraying device, in m / s; a and b are the spraying flow coefficients, determined by the nozzle flow model; V canopy The volume of the fruit tree canopy in zone i is expressed in meters. 3 .

[0081] In addition, the fruit tree pesticide application model can also calculate the required pesticide application amount for fruit trees based on multiple types of decision information, including fruit tree height, fruit tree diameter at breast height, fruit tree soil fertility, and fruit tree diseases and pests.

[0082] Step S105: Import the latitude and longitude information of the calibration ball, the three-dimensional point cloud map of the orchard, the latitude and longitude information of the fruit trees, and the variable application prescription values ​​of the fruit trees into the GIS system to generate the variable application prescription map of the orchard.

[0083] Specifically, in this embodiment, the orchard variable application prescription map is generated by using SLAM technology to construct a three-dimensional point cloud map of the orchard environment. This allows for the large-scale acquisition of fruit tree canopy information as input for the fruit tree application rate model. After calculating the prescription value, the map is generated based on the prescription value combined with the latitude and longitude information of the fruit trees.

[0084] In one embodiment, step S105 includes: registering the satellite map with the orchard 3D point cloud map based on the location of the calibration sphere point cloud model in the orchard 3D point cloud map and the geographical location in the satellite map corresponding to the latitude and longitude information of the calibration sphere; creating XYZ / prescription value data based on the latitude and longitude information of the fruit trees and the prescription values ​​of each canopy partition in the variable prescription values ​​of the fruit trees; constructing columnar volumetric raster based on the XYZ / prescription value data using a spatial interpolation algorithm; using the columnar volumetric raster to represent the fruit tree canopy model and adjusting the height of the columnar volumetric raster to obtain a volumetric raster layer; and importing the volumetric raster layer, the geographic information layer, and the spatial feature layer into the GIS workspace to obtain the orchard variable prescription map.

[0085] Specifically, first, a GIS workspace is created, a new spherical scene is created, and the orchard 3D point cloud map is converted into a LAS or PLY file. Then, a point cloud cache is created, and the orchard 3D point cloud map file is set as the source file. The projection settings are set to a planar coordinate system to generate the point cloud cache, which is then loaded into the spherical scene as a 3D tile cache layer. Next, an online map data source is created, selecting the MapWorld data source type, and the satellite map is loaded into the spherical scene. The orchard 3D point cloud map is set as the registration layer, and the satellite map is set as the reference layer. A crosshair is used to mark the location of the calibration sphere point cloud model in the orchard 3D point cloud map, and the latitude and longitude information of the calibration sphere corresponds to the geographical location in the satellite map. The satellite map and the orchard 3D point cloud map are then geographically aligned to achieve registration between the orchard 3D point cloud map and the satellite map, obtaining the geographic information layer and the spatial information layer of the ground features.

[0086] Then, based on the latitude and longitude coordinates of all fruit tree bounding boxes within the orchard plant protection operation area and the pesticide prescription values ​​for each canopy zone in the variable pesticide prescription values, an xls data table is created and loaded into the workspace. Based on the prescription values ​​for different canopy zones, a spatial interpolation algorithm is used to construct a voxel raster, thus obtaining the corresponding voxel raster layer (prescription value information). The variable pesticide prescription values ​​of the fruit trees are used as feature values ​​of the voxel raster model, and a columnar voxel raster model is used to represent the fruit tree canopy model, adjusting the height of the voxel raster model. Alternatively, an xls data table can be created using the latitude and longitude coordinates of the fruit tree bounding boxes and the pesticide prescription values ​​for each canopy zone in the variable pesticide prescription values, loaded into the workspace, and a two-dimensional raster can be constructed. The two-dimensional raster is then stretched three-dimensionally to construct a columnar model, with minimum and maximum heights set. The pesticide prescription values ​​for each canopy zone are then rendered using feature values. Finally, all layers are imported into the GIS workspace to obtain the orchard variable pesticide prescription map, as shown below. Figure 6 As shown, the orchard variable application prescription map in this embodiment is a three-dimensional, three-dimensional voxel grid fruit tree variable application prescription map. It includes three layers of information: satellite map (geographic information), orchard three-dimensional point cloud map (geographic spatial information), and voxel grid map (prescription value information). It uses a voxel grid interpolation algorithm to describe the spatial differences of the fruit tree canopy, enabling precise variable application of pesticides to different zones of the fruit tree canopy.

[0087] The above-mentioned method for generating orchard variable application prescription maps has the following advantages compared with existing technologies:

[0088] 1. An improved SLAM algorithm based on structured orchard environments was developed. The SLAM graph optimization model was improved by employing a laser odometry factor with elastic characteristics. Based on the continuity of pose during a single frame of data acquisition and the discontinuity of pose during inter-frame data acquisition, two pose parameters were used to parameterize a single frame of point cloud data. Furthermore, a trunk point cloud factor was introduced to further enhance the modeling accuracy of the orchard's 3D point cloud map. This method can efficiently acquire information such as tree height, canopy volume, and canopy leaf area density on a large scale, constructing a multi-type pesticide dosage decision-making information calculation model for fruit trees. This calculates the prescription value for target fruit trees, improving operational efficiency and meeting diverse pesticide application needs. It also enhances the accuracy of crop information acquisition and improves the spray response speed. This addresses the problems of poor accuracy, limited spray response speed, and limited pesticide dosage decision-making features in current orchard variable pesticide application technologies.

[0089] 2. A 3D GIS technology was used to create a variable-rate pesticide application prescription map for the orchard. This involved establishing three layers of information: a satellite map, a 3D point cloud map of the orchard, and a voxel grid map. This generated a map with geographical, feature, and application characteristics. Furthermore, based on the orchard planting pattern, the OBB algorithm was used to locate the spraying area, enabling precise targeted pesticide application and reducing pesticide drift and ground runoff. A voxel grid interpolation algorithm was employed to process the variable-rate pesticide application prescription values, allowing for accurate description of the spatial differences within the fruit tree canopy using 3D features, and enabling precise variable-rate pesticide application to different zones of the canopy.

[0090] In one embodiment, such as Figure 7 As shown, an orchard variable application prescription map generation device is provided, comprising:

[0091] The acquisition module 701 is used to acquire point cloud information of the orchard environment and latitude and longitude information of the calibration sphere within the orchard using a scanning device.

[0092] The mapping module 702 is used to construct a 3D point cloud map of the orchard based on point cloud information during the acquisition process using the SLAM algorithm and SLAM graph optimization model, and to filter the ground point cloud in the 3D point cloud map of the orchard to obtain a 3D point cloud model of the fruit trees.

[0093] The reconstruction module 703 is used to construct the bounding box of the three-dimensional point cloud model of the fruit tree using a clustering algorithm, determine the latitude and longitude information of the fruit tree based on the bounding box, and obtain the canopy partition volume and canopy partition leaf area density of the fruit tree by reconstructing the three-dimensional point cloud model of the fruit tree.

[0094] The pesticide application calculation module 704 is used to calculate the pesticide application prescription value of each canopy zone based on the canopy zone volume and canopy zone leaf area density, and obtain the fruit tree variable pesticide application prescription value.

[0095] The generation module 705 is used to import the latitude and longitude information of the calibration ball, the three-dimensional point cloud map of the orchard, the latitude and longitude information of the fruit trees, and the variable pesticide application prescription values ​​of the fruit trees into the GIS system to generate the variable pesticide application prescription map of the orchard.

[0096] The specific limitations of the orchard variable application prescription map device can be found in the limitations of the orchard variable application prescription map method described above, and will not be repeated here. Each module in the above-mentioned orchard variable application prescription map device can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module. Based on this understanding, all or part of the processes in the above-described embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-described orchard variable application prescription map method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms.

[0097] In one embodiment, a computer device is provided, which may be a server, including a processor, memory, and a network interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements an orchard variable-rate pesticide prescription map method. Exemplarily, the computer program may be divided into one or more modules, one or more modules are stored in memory, and executed by the processor to complete the invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the computer device. The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the computer device, connecting various parts of the computer device via various interfaces and lines.

[0098] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0099] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0100] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for generating a variable-rate pesticide application prescription map for orchards, characterized in that, include: The scanning device was used to collect point cloud information of the orchard environment and latitude and longitude information of the calibration spheres in the orchard; During the data collection process, the SLAM algorithm and SLAM graph optimization model are used to construct a three-dimensional point cloud map of the orchard based on the point cloud information, and the ground point cloud in the three-dimensional point cloud map of the orchard is filtered to obtain a three-dimensional point cloud model of the fruit trees. Clustering algorithm is used to segment the three-dimensional point cloud model of the fruit tree to construct the bounding box of the three-dimensional point cloud model of the fruit tree. Based on the bounding box, the latitude and longitude information of the fruit tree is determined, and the canopy partition volume and canopy partition leaf area density of the fruit tree are obtained by reconstructing the three-dimensional point cloud model of the fruit tree. The pesticide prescription value for each canopy zone is calculated based on the canopy zone volume and the canopy zone leaf area density to obtain the fruit tree variable pesticide prescription value; Import the latitude and longitude information of the calibration ball, the three-dimensional point cloud map of the orchard, the latitude and longitude information of the fruit trees, and the variable pesticide application prescription values ​​of the fruit trees into the GIS system to generate a variable pesticide application prescription map of the orchard; The process of constructing a 3D point cloud map of the orchard based on the point cloud information using the SLAM algorithm and SLAM graph optimization model during the data acquisition includes: During the acquisition process, the acquired point cloud information is reprojected into a depth image, and the ground point cloud is filtered based on the angle between adjacent points in the depth image, and the point cloud with neighboring points is calculated by traversing and filtering out noisy point clouds. The pose of the point cloud is estimated using a laser odometry model. The estimated pose is then combined with the voxel mesh method to store the point cloud in the world coordinate system, resulting in a 3D point cloud map of the orchard with geographical location information. The orchard 3D point cloud map was optimized using a backend graph optimization model; The laser odometry model is represented as follows: in, Express posture; This indicates the key point cloud extracted from the point cloud of the new frame; It represents the loss function; lerp represents the interpolation function. yes The normal of the neighborhood; These are lidar measurements; It is a local point cloud map; It is the linear interpolated pose between the start pose and the end pose. It is a neighborhood planarization function; It is a positional consistency function, expressed as ; It is the speed consistency function, expressed as ; and It is the rotation and translation of the starting pose. and It is the rotation and translation that ends the pose; The backend graph optimization model is represented as follows: in, ; and It is a tree trunk point cloud factor. , ; It indicates that the tree trunks are clustered together; The tree trunk point cloud set is obtained by extracting the feature point cloud at breast height (DBH) of the fruit trees in the orchard's 3D point cloud map using a Euclidean clustering algorithm. Then, the feature point clouds belonging to the same fruit tree at DBH are segmented using a judgment formula, which is as follows: in, It is a characteristic dot cloud and The distance between the centers of the circles; It is a characteristic dot cloud and The maximum distance between the centers of the circles; It is the radius of the characteristic circle; It is a characteristic circle and The observation time difference; and It is a constant.

2. The method according to claim 1, characterized in that, The process of filtering the ground point cloud in the orchard 3D point cloud map to obtain the fruit tree 3D point cloud model includes: The ground point cloud in the orchard 3D point cloud map is filtered based on the rotation constraint method and the installation height of the lidar in the scanning device, or the ground point cloud in the orchard 3D point cloud map is fitted and filtered to obtain the fruit tree 3D point cloud model.

3. The method according to claim 1, characterized in that, The formula for calculating the prescription value for canopy zonal drug application is as follows: in, For the canopy of fruit tree n The prescription values ​​for drug administration in each zone; For the canopy Leaf area density of different zones; The speed at which the spraying device moves; , The spraying flow rate coefficient; For fruit tree canopy Partition volume.

4. The method according to claim 1, characterized in that, The process of importing the latitude and longitude information of the calibration ball, the three-dimensional point cloud map of the orchard, the latitude and longitude information of the fruit trees, and the variable pesticide application prescription values ​​of the fruit trees into the GIS system to generate a variable pesticide application prescription map of the orchard includes: Based on the location of the calibration sphere point cloud model in the orchard 3D point cloud map and the geographical location in the satellite map corresponding to the latitude and longitude information of the calibration sphere, the satellite map is registered with the orchard 3D point cloud map to obtain a geographic information layer and a land feature spatial layer. XYZ / prescription value data is created based on the latitude and longitude information of the fruit trees and the prescription values ​​of each canopy partition in the variable prescription values ​​of the fruit trees. Based on the XYZ / prescription value data, a columnar volumetric raster is constructed using a spatial interpolation algorithm. The columnar volumetric raster is used to represent the fruit tree canopy model, and the height of the columnar volumetric raster is adjusted to obtain a volumetric raster layer. Import the voxel raster layer, the geographic information layer, and the land feature spatial layer into the GIS workspace to obtain the orchard variable application prescription map.

5. A device for generating a variable-rate pesticide application prescription map for orchards, characterized in that, include: The data acquisition module is used to collect point cloud information of the orchard environment and latitude and longitude information of the calibration sphere in the orchard using a scanning device; The mapping module is used to construct a 3D point cloud map of the orchard based on the point cloud information using the SLAM algorithm and SLAM graph optimization model during the acquisition process, and to filter the ground point cloud in the 3D point cloud map of the orchard to obtain a 3D point cloud model of the fruit trees. The reconstruction module is used to segment the three-dimensional point cloud model of the fruit tree using a clustering algorithm to construct the bounding box of the three-dimensional point cloud model of the fruit tree, determine the latitude and longitude information of the fruit tree based on the bounding box, and reconstruct the three-dimensional point cloud model of the fruit tree to obtain the canopy partition volume and canopy partition leaf area density of the fruit tree. The pesticide application calculation module is used to calculate the pesticide application prescription value for each canopy zone based on the canopy zone volume and the canopy zone leaf area density, and obtain the fruit tree variable pesticide application prescription value; The generation module is used to import the latitude and longitude information of the calibration ball, the three-dimensional point cloud map of the orchard, the latitude and longitude information of the fruit trees, and the variable pesticide application prescription values ​​of the fruit trees into the GIS system to generate a variable pesticide application prescription map of the orchard; The process of constructing a 3D point cloud map of the orchard based on the point cloud information using the SLAM algorithm and SLAM graph optimization model during the acquisition includes: reprojecting the acquired point cloud information into a depth image during the acquisition process; filtering the ground point cloud based on the angle between adjacent points in the depth image and filtering noisy point clouds by traversing and calculating the angle between the point cloud and its neighboring points; estimating the pose of the point cloud using a laser odometry model; combining the estimated pose with the voxel mesh method to store the point cloud in the world coordinate system to obtain a 3D point cloud map of the orchard with geographical location information; and optimizing the 3D point cloud map of the orchard using a backend graph optimization model. The laser odometry model is represented as follows: in, Express posture; This indicates the key point cloud extracted from the point cloud of the new frame; It represents the loss function; lerp represents the interpolation function. yes The normal of the neighborhood; These are lidar measurements; It is a local point cloud map; It is the linear interpolated pose between the start pose and the end pose. It is a neighborhood planarization function; It is a positional consistency function, expressed as ; It is the speed consistency function, expressed as ; and It is the rotation and translation of the starting pose. and It is the rotation and translation that ends the pose; The backend graph optimization model is represented as follows: in, ; and It is a tree trunk point cloud factor. , ; It indicates that the tree trunks are clustered together; The tree trunk point cloud set is obtained by extracting the feature point cloud at breast height (DBH) of the fruit trees in the orchard's 3D point cloud map using a Euclidean clustering algorithm. Then, the feature point clouds belonging to the same fruit tree at DBH are segmented using a judgment formula, which is as follows: in, It is a characteristic dot cloud and The distance between the centers of the circles; It is a characteristic dot cloud and The maximum distance between the centers of the circles; It is the radius of the characteristic circle; It is a characteristic circle and The observation time difference; and It is a constant.

6. A computer device comprising a processor and a memory, the memory storing a computer program, characterized in that, The processor is used to execute the computer program to implement the orchard variable application prescription map generation method according to any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the orchard variable application prescription map generation method according to any one of claims 1-4.

Images