A corn whole life cycle monitoring method and system based on air-ground cooperation

Abstract

The application is suitable for the field of agricultural monitoring technology, and particularly relates to a corn whole life cycle monitoring method and system based on air-ground cooperation, which comprises the following steps: outputting inferred stalk diameter and top plane coordinates through a UAV; obtaining measured stalk diameter and base plane coordinates through a ground robot; constructing a first plant distribution matrix and a second plant distribution matrix; matching the first plant distribution matrix with the second plant distribution matrix; and according to the matching result, associating the top monitoring data collected by the UAV with the side monitoring data collected by the ground robot to the same corn plant. Through the matching based on the relative position relationship between the plant stalk diameter and the group, the application overcomes the data misplacement problem caused by the positioning drift of the UAV and the ground robot, does not need to rely on expensive high-precision positioning equipment, provides a reliable data basis for planting decisions such as variable rate fertilization and precision pesticide spraying, and improves the accuracy of corn precision management.

Description

Technical Field

[0001] This invention belongs to the field of agricultural monitoring technology, and in particular relates to a method and system for monitoring the entire life cycle of maize based on air-ground collaboration. Background Technology

[0002] Air-ground collaboration is a three-dimensional operational paradigm that integrates drones, ground sensors, and intelligent equipment. It enables platforms in different spatial dimensions (space-based, air-based, and ground-based) to perform their respective functions, exchange data, and intelligently coordinate under unified scheduling, achieving a closed-loop management system for wide-area perception, refined diagnosis, and precise execution of target objects. The high-precision maneuverability of drones is deeply integrated with the real-time verification and execution capabilities of ground equipment, ultimately forming an intelligent system of "observation-analysis-decision-action."

[0003] In the current corn growth monitoring process, the monitoring data of the same corn plant is collected by drones and ground robots in collaboration. However, both drones and ground robots have limited positioning accuracy and are prone to position drift, which leads to mismatch between air and ground data. This results in inaccurate monitoring data for the corn plant, affecting the subsequent strategy formulation and implementation. Summary of the Invention

[0004] The purpose of this invention is to provide a corn full life cycle monitoring method based on air-ground collaboration, which aims to solve the problem that the positioning accuracy of both drones and ground robots is limited and prone to position drift, resulting in mismatch between air and ground data, inaccurate monitoring data of the corn plant, and affecting subsequent strategy formulation and execution.

[0005] This invention is implemented as follows: a method for monitoring the entire life cycle of maize based on air-ground collaboration, the method comprising:

[0006] Multispectral images of cornfields are acquired from a top view using drones, and the inferred stalk diameter and top plane coordinates of each corn plant are detected and output. Corn stalk images are acquired from a side view using ground robots, and the measured stalk diameter and base plane coordinates of each corn plant are detected.

[0007] Based on data measured by UAVs, a first plant distribution matrix is ​​constructed, which includes the inferred stem diameter and top plane coordinates of all plants; based on data measured by ground robots, a second plant distribution matrix is ​​constructed, which includes the measured stem diameter and base plane coordinates of all plants.

[0008] The first plant distribution matrix is ​​matched with the second plant distribution matrix. The matching criteria include the relative positional relationship between plants based on planar coordinates, and the numerical correspondence between the inferred stem diameter and the measured stem diameter.

[0009] Based on the matching results, the top monitoring data collected by the drone and the side monitoring data collected by the ground robot are associated with the same corn plant.

[0010] Preferably, the steps of constructing a first plant distribution matrix containing the inferred stem diameter and top plane coordinates of all plants based on data measured by UAVs, and constructing a second plant distribution matrix containing the measured stem diameter and base plane coordinates of all plants based on data measured by ground robots, specifically include:

[0011] For UAV data, the top plane coordinates of each corn plant are used as nodes. The Euclidean distance between each plant and the coordinates of all other plants is calculated, and its K nearest neighbor plants are found to form a local topological subgraph centered on that plant.

[0012] For ground robot data, perform the same operations as for UAV data, and construct a local topological subgraph with the same K value for each plant based on the base plane coordinates;

[0013] In each local topological subgraph, the diameter of the central plant, its distance vector to its K neighbors, and the diameters of the K neighbors themselves are combined into a feature vector. The set of feature vectors of all plants constitutes the first plant distribution matrix and the second plant distribution matrix.

[0014] Preferably, the step of matching the first plant distribution matrix with the second plant distribution matrix, based on the relative positional relationship between plants according to planar coordinates and the step of inferring the numerical correspondence between the stem diameter and the measured stem diameter, specifically includes:

[0015] Calculate the comprehensive similarity between the feature vectors of any two local topological subgraphs in the first plant distribution matrix and the second plant distribution matrix. This comprehensive similarity is a weighted sum of the difference in the diameter of the central plant and the cosine similarity of the two distance vectors.

[0016] Using a graph matching algorithm, with the overall similarity as the edge weight, we find the optimal matching pair that maximizes the total similarity of all plant correspondences in the two matrices;

[0017] Geometric verification is performed on the optimal matching pairs. Using the successfully matched plant pairs, an optimal rigid body transformation matrix is ​​calculated to transform the UAV coordinate system to the ground robot coordinate system, and an open-ground plant mapping table is generated.

[0018] Preferably, the step of associating the top monitoring data collected by the UAV and the side monitoring data collected by the ground robot with the same corn plant based on the matching results specifically includes:

[0019] Based on the open-ground plant mapping table, the multispectral vegetation index data obtained by the drone for each corn plant is bound to the ground surface data collected by the ground robot for the corresponding plant.

[0020] Using the base plane coordinates collected by the ground robot as a reference, the top plane coordinates of the corresponding plant collected by the UAV are corrected by applying the optimal rigid body transformation matrix to obtain the geographical location of the plant after fusion.

[0021] The multi-source data of each corn plant is merged and bound to its geographical location, and then updated to the same digital map to form a monitoring file for the entire growth period at the single plant scale.

[0022] Preferably, the ground robot uses lidar to measure the plant diameter.

[0023] Another objective of this invention is a corn full life cycle monitoring system based on air-ground collaboration, the system comprising:

[0024] The data acquisition module is used to acquire multispectral images of cornfields from a top view using a drone, detect and output the inferred stalk diameter and top plane coordinates of each corn plant; and to acquire corn stalk images from a side view using a ground robot, and detect the measured stalk diameter and base plane coordinates of each corn plant.

[0025] The matrix construction module is used to construct a first plant distribution matrix containing the inferred stem diameter and top plane coordinates of all plants based on data measured by UAVs; and to construct a second plant distribution matrix containing the measured stem diameter and base plane coordinates of all plants based on data measured by ground robots.

[0026] The location matching module is used to match the first plant distribution matrix with the second plant distribution matrix. The matching criteria include the relative positional relationship between plants based on planar coordinates, and the numerical correspondence between the inferred stem diameter and the measured stem diameter.

[0027] The data matching module is used to associate the top monitoring data collected by the drone with the side monitoring data collected by the ground robot to the same corn plant based on the matching results.

[0028] Preferably, the matrix construction module includes:

[0029] The aerial data topology unit is used to calculate the Euclidean distance between the top plane coordinates of each corn plant and the coordinates of all other plants for UAV data, and to find its K nearest neighbor plants, forming a local topological subgraph centered on that plant.

[0030] Ground data topology unit is used to perform the same operations on ground robot data as on UAV data, and to construct a local topology subgraph with the same K value for each plant based on the base plane coordinates;

[0031] The matrix generation unit is used to combine the diameter of the central plant, its distance vector to K neighbors, and the diameters of the K neighbors themselves into a feature vector in each local topological subgraph. The set of feature vectors of all plants constitutes the first plant distribution matrix and the second plant distribution matrix.

[0032] Preferably, the location matching module includes:

[0033] The similarity calculation unit is used to calculate the comprehensive similarity between the feature vectors of any two local topological subgraphs in the first plant distribution matrix and the second plant distribution matrix. The comprehensive similarity is a weighted sum of the difference in the diameter of the central plant and the cosine similarity of the two distance vectors.

[0034] The plant matching unit is used to find the optimal matching pair that maximizes the total similarity of all plant correspondences in two matrices by using a graph matching algorithm with the comprehensive similarity as the edge weight.

[0035] The mapping table construction unit is used to perform geometric verification on the optimal matching pair. Using the successfully matched plant pairs, it calculates an optimal rigid body transformation matrix that transforms the UAV coordinate system to the ground robot coordinate system, and generates an open-ground plant mapping table.

[0036] Preferably, the data matching module includes:

[0037] The data binding unit is used to bind the multispectral vegetation index data obtained by the drone for each corn plant with the ground data collected by the ground robot for the corresponding plant, according to the open-ground plant mapping table.

[0038] The geolocation transformation unit is used to correct the top plane coordinates of the corresponding plant collected by the UAV based on the base plane coordinates collected by the ground robot and apply the best rigid body transformation matrix to obtain the fused geolocation of the plant.

[0039] The archive construction unit is used to bind the fused multi-source data of each corn plant with its geographical location and update it to the same digital map to form a monitoring archive of the entire growth period at the single plant scale.

[0040] Preferably, the ground robot uses lidar to measure the plant diameter.

[0041] This invention overcomes the data misalignment problem caused by positioning drift between UAVs and ground robots by matching the relationship between plant stem diameter and relative position of the population, without relying on expensive high-precision positioning equipment. The method realizes fully automatic and high-precision single-plant-level fusion of multi-source heterogeneous monitoring data, providing a reliable data foundation for planting decisions such as variable fertilization and precision spraying, and improving the accuracy of precision management of maize. Attached Figure Description

[0042] Figure 1 A flowchart illustrating a corn life-cycle monitoring method based on air-ground collaboration, provided as an embodiment of the present invention;

[0043] Figure 2 A flowchart for constructing the first plant distribution matrix and the second plant distribution matrix provided in an embodiment of the present invention;

[0044] Figure 3 A flowchart illustrating the steps for inferring the numerical correspondence between the stem diameter and the measured stem diameter provided in this embodiment of the invention;

[0045] Figure 4 A flowchart illustrating how, based on matching results, top monitoring data collected by a drone and side monitoring data collected by a ground robot are associated with the same corn plant, as provided in this embodiment of the invention.

[0046] Figure 5 This is an architecture diagram of a corn life cycle monitoring system based on air-ground collaboration, provided as an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0048] like Figure 1 As shown in the figure, an embodiment of the present invention provides a method for monitoring the entire life cycle of maize based on air-ground collaboration. The method includes:

[0049] The S100 uses a drone to collect multispectral images of cornfields from a top-down perspective, detects and outputs the inferred stalk diameter and top plane coordinates of each corn plant; and uses a ground robot to collect corn stalk images from a side-down perspective, and detects the measured stalk diameter and base plane coordinates of each corn plant.

[0050] In this step, a drone equipped with a multispectral camera takes aerial photos from directly above the cornfield from a vertical perspective, acquiring high-resolution digital images covering the entire field. Image recognition is then performed on these high-resolution digital images to identify each individual corn plant. For each plant instance, the pixel features of its stem base in the image are analyzed to infer the stem diameter and determine the position of the plant's crown center point in the image coordinate system. Combined with the drone's own GNSS position and inertial measurement unit data, the top planar coordinates of the plant are calculated. Simultaneously, a ground robot moves through the field, its onboard laser sensor acquiring detailed images or point clouds of the corn stems at close range. A point cloud fitting algorithm is used to obtain the measured stem diameter of each corn plant, and the robot's own positioning system is used to determine the base planar coordinates of the stem base.

[0051] S200: Based on data measured by UAVs, a first plant distribution matrix is ​​constructed, which includes the inferred stem diameter and top plane coordinates of all plants; based on data measured by ground robots, a second plant distribution matrix is ​​constructed, which includes the measured stem diameter and base plane coordinates of all plants.

[0052] In this step, for the data collected by the UAV, the inferred stem diameter and top plane coordinates of all plants are read to form a first plant distribution matrix. The position of each plant and the inferred stem diameter output by the UAV are recorded through the first plant distribution matrix. Similarly, for the data collected by the ground robot, the measured stem diameter of all plants and the coordinates of the corresponding plants are read to construct a second plant distribution matrix. Through the two independent first and second plant distribution matrices, the plant positions and corresponding plant sizes measured by the UAV and the ground robot can be determined respectively.

[0053] S300, Match the first plant distribution matrix with the second plant distribution matrix. The matching criteria include the relative positional relationship between plants based on planar coordinates, and the numerical correspondence between the inferred stem diameter and the measured stem diameter.

[0054] In this step, the first and second plant distribution matrices are matched, and a dual constraint condition is introduced for joint solution. The first constraint is the relative position relationship: based on the coordinates in the matrix, the distance between any two plants in the plant group is calculated to form a spatial topological fingerprint. This relative position between plants in the same field is stable and unchanging whether observed from the air or the ground. The second constraint is the numerical correspondence relationship: using the stem diameter column in the two matrices, since the same feature - stem diameter - is measured, the diameter values ​​of corresponding plants should be very similar. The solution is performed under the above two constraints to output the plant ID pairs that satisfy the above constraints.

[0055] The S400, based on the matching results, associates the top monitoring data collected by the drone with the side monitoring data collected by the ground robot with the same corn plant.

[0056] In this step, based on the obtained plant ID mapping table, a data association operation is performed. All top monitoring data collected by the UAV for the plant with ID i, including leaf area index calculated from reflectance spectrum, normalized difference vegetation index (NDVI), and chlorophyll content inversion value, are bound with all side monitoring data collected by the ground robot for the corresponding plant, such as leaf thermal infrared temperature, stem height laser measurement value, and local high-definition images of pests and diseases. This data is then merged into a complete multi-dimensional data archive for the same corn plant. All the associated individual plant data are integrated and visualized on a unified digital map.

[0057] like Figure 2 As shown, in a preferred embodiment of the present invention, the steps of constructing a first plant distribution matrix containing the inferred stem diameter and top plane coordinates of all plants based on data measured by a UAV, and constructing a second plant distribution matrix containing the measured stem diameter and base plane coordinates of all plants based on data measured by a ground robot, specifically include:

[0058] S201. For UAV data, take the top plane coordinates of each corn plant as a node, calculate its Euclidean distance to the coordinates of all other plants, and find its K nearest neighbor plants to form a local topological subgraph centered on that plant.

[0059] In this step, the top plane coordinates of each corn plant in the first plant distribution matrix are read ( , For the i-th plant currently serving as the center, calculate the Euclidean distance between this plant and the coordinates of all other plants in the matrix. Select the K smallest values ​​from these distances, such as K=4 or 6, to determine the K nearest neighbors of the center plant in the field. The distances of these K nearest neighbors and their distances to the center plant together form a local topological subgraph with the center plant as the origin and its neighbors as nodes. This local topological subgraph is used to describe the relative distances and orientations between plants.

[0060] S202, for ground robot data, perform the same operation as for UAV data, and construct a local topological subgraph with the same K value for each plant based on the base plane coordinates.

[0061] In this step, using the exact same logic and parameters, a corresponding local topological subgraph is constructed for the ground robot data, and the basal plane coordinates of each corn plant in the second plant distribution matrix are read. , For each plant j that is the center, repeat the same operation as S201, calculate the Euclidean distance to all other plants, and select its K nearest neighbor plants according to the same K value, thereby forming a local topological subgraph based on the ground coordinate system.

[0062] S203, in each local topological subgraph, the diameter of the central plant, its distance vector to K neighbors, and the diameters of the K neighbors themselves are combined into a feature vector. The set of feature vectors of all plants constitutes the first plant distribution matrix and the second plant distribution matrix.

[0063] In this step, for each constructed local topological subgraph, whether from a drone or a ground robot, it is encoded into a high-dimensional feature vector. This high-dimensional feature vector contains three parts of information in a fixed order: the diameter value of the central plant itself, etc. A distance vector containing K elements, which records the precise distances from the central plant to its K neighbors, and a neighbor diameter vector containing K elements, which records the diameter values ​​of the K neighbor plants. The feature vectors corresponding to all plants in the dataset are arranged in order as rows to generate a structured first plant distribution matrix or a second plant distribution matrix.

[0064] like Figure 3 As shown, in a preferred embodiment of the present invention, the step of matching the first plant distribution matrix with the second plant distribution matrix, based on the relative positional relationship between plants according to planar coordinates and the step of inferring the numerical correspondence between the stem diameter and the measured stem diameter, specifically includes:

[0065] S301, calculate the comprehensive similarity between the feature vectors of any two local topological subgraphs in the first plant distribution matrix and the second plant distribution matrix. This comprehensive similarity is the weighted sum of the difference in the diameter of the central plant and the cosine similarity of the two distance vectors.

[0066] In this step, a local topological subgraph feature vector centered on plant A from UAV data and a local topological subgraph feature vector centered on plant B from ground data are extracted from the structured first and second plant distribution matrices. The comprehensive similarity between the two is calculated. The comprehensive similarity includes stem difference and spatial overlap. Regarding stem difference, the absolute difference between the inferred stem diameter of A and the measured stem diameter of B is calculated. The smaller this value, the more consistent the two are in key biological characteristics. Regarding spatial overlap, the cosine similarity between the distance vectors of A and B is calculated. The higher the cosine similarity, the more similar the spatial distribution patterns of the neighbors around the two plants are. The normalized comprehensive similarity is output through preset weight coefficients, where the diameter difference weight is 0.4 and the spatial overlap weight is 0.6.

[0067] S302 uses a graph matching algorithm, with the overall similarity as the edge weight, to find the optimal matching pair that maximizes the total similarity of all plant correspondences in the two matrices.

[0068] In this step, the plant sets of the UAV and the plant sets of the ground robot are abstracted into two graphs, where each plant is a node, and the potential matching relationship between any two plants forms an edge. The weight of the edge is the comprehensive similarity. A graph matching algorithm is used to solve this bipartite graph. Under the constraint that a UAV plant can match at most one ground plant, the algorithm is used to maximize the sum of the edge weights of all successfully paired plant pairs, and finally outputs a plant pairing mapping list.

[0069] S303, perform geometric verification on the optimal matching pair, and use the successfully matched plant pairs to calculate an optimal rigid body transformation matrix that transforms the UAV coordinate system to the ground robot coordinate system, and generate an open-ground plant mapping table.

[0070] In this step, the most similar matching pairs are selected from the plant pairing mapping list as reliable anchor points. Using the precise coordinates of the ground plants in the reliable anchor points as the reference and the coordinates of the UAV plants as the source, the least squares method is used to calculate the optimal rigid body transformation matrix. The optimal rigid body transformation matrix is ​​used to describe the mapping rules for transforming the entire coordinate system of the UAV to the coordinate system of the ground robot. The coordinates of all initially matched UAV plants are transformed using the optimal rigid body transformation matrix, and the Euclidean distance between the transformed position and the position of its paired ground plant is calculated. If the position deviation of a certain pair of plants is greater than a preset threshold, such as 10 cm, it is considered a mismatched pair and is removed. The empty ground plant mapping table is then output.

[0071] like Figure 4 As shown, in a preferred embodiment of the present invention, the step of associating the top monitoring data collected by the UAV and the side monitoring data collected by the ground robot with the same corn plant based on the matching results specifically includes:

[0072] S401, based on the open-ground plant mapping table, binds the multispectral vegetation index data obtained by the drone for each corn plant with the ground data collected by the ground robot for the corresponding plant.

[0073] In this step, based on the open-field plant mapping table as the authoritative index for data association, for each pair of matching IDs in the mapping table, the drone is assigned to that plant. The top monitoring data collected (including parameters such as Normalized Difference Vegetation Index (NDVI), Leaf Area Index (LAI), and Chlorophyll Content (Chl)) were extracted and analyzed based on the corresponding ground robot plant ( The system extracts the collected monitoring data (including leaf nitrogen, phosphorus and potassium content, canopy temperature obtained by thermal imager and local feature images of pests and diseases captured by high-definition camera), records the two sets of data from different sources through data structure binding, and associates them with a unique fused plant ID.

[0074] S402 uses the base plane coordinates collected by the ground robot as a reference and applies the optimal rigid body transformation matrix to correct the top plane coordinates of the corresponding plant collected by the UAV, thereby obtaining the geographical location of the plant after fusion.

[0075] In this step, the base plane coordinates collected by the ground robot are read through the optimal rigid body transformation matrix. For each pair of matched plants, the ground coordinates are determined as the reference geographical location of the plant. The corresponding top plane coordinates of the UAV are read and transformed by the rigid body transformation matrix. The transformed UAV coordinates are corrected to the ground robot coordinate system. The weighted average algorithm is used to fuse the ground reference coordinates and the corrected UAV coordinates to generate the fused geographical location of the plant.

[0076] S403 integrates multi-source data from each corn plant and binds it to its geographical location, updating it to the same digital map to form a monitoring archive of the entire growth period at the single-plant scale.

[0077] In this step, a digital map is created. For each corn plant, its geographical location is used as a spatial anchor point, and all the multi-source monitoring data associated with it are written into the database record of the map as attribute information, thereby forming a digital archive of the corn plant's entire growth period.

[0078] like Figure 5 As shown in the figure, an air-ground collaborative corn life cycle monitoring system is provided by an embodiment of the present invention. The system comprises:

[0079] The data acquisition module 100 is used to acquire multispectral images of the cornfield from a top view using a drone, detect and output the inferred stalk diameter and top plane coordinates of each corn plant; and to acquire corn stalk images from a side view using a ground robot, and detect the measured stalk diameter and base plane coordinates of each corn plant.

[0080] The matrix construction module 200 is used to construct a first plant distribution matrix containing the inferred stem diameter and top plane coordinates of all plants based on data measured by UAVs; and to construct a second plant distribution matrix containing the measured stem diameter and base plane coordinates of all plants based on data measured by ground robots.

[0081] The matrix construction module includes:

[0082] The aerial data topology unit is used to calculate the Euclidean distance between the top plane coordinates of each corn plant and the coordinates of all other plants for UAV data, and to find its K nearest neighbor plants, forming a local topological subgraph centered on that plant.

[0083] Ground data topology unit is used to perform the same operations on ground robot data as on UAV data, and to construct a local topology subgraph with the same K value for each plant based on the base plane coordinates;

[0084] The matrix generation unit is used to combine the diameter of the central plant, its distance vector to K neighbors, and the diameters of the K neighbors themselves into a feature vector in each local topological subgraph. The set of feature vectors of all plants constitutes the first plant distribution matrix and the second plant distribution matrix.

[0085] The position matching module 300 is used to match the first plant distribution matrix with the second plant distribution matrix. The matching basis includes the relative positional relationship between plants based on planar coordinates, and the numerical correspondence between the inferred stem diameter and the measured stem diameter.

[0086] The location matching module includes:

[0087] The similarity calculation unit is used to calculate the comprehensive similarity between the feature vectors of any two local topological subgraphs in the first plant distribution matrix and the second plant distribution matrix. The comprehensive similarity is a weighted sum of the difference in the diameter of the central plant and the cosine similarity of the two distance vectors.

[0088] The plant matching unit is used to find the optimal matching pair that maximizes the total similarity of all plant correspondences in two matrices by using a graph matching algorithm with the comprehensive similarity as the edge weight.

[0089] The mapping table construction unit is used to perform geometric verification on the optimal matching pair. Using the successfully matched plant pairs, it calculates an optimal rigid body transformation matrix that transforms the UAV coordinate system to the ground robot coordinate system, and generates an open-ground plant mapping table.

[0090] The data matching module 400 is used to associate the top monitoring data collected by the UAV and the side monitoring data collected by the ground robot with the same corn plant based on the matching results.

[0091] The data matching module includes:

[0092] The data binding unit is used to bind the multispectral vegetation index data obtained by the drone for each corn plant with the ground data collected by the ground robot for the corresponding plant, according to the open-ground plant mapping table.

[0093] The geolocation transformation unit is used to correct the top plane coordinates of the corresponding plant collected by the UAV based on the base plane coordinates collected by the ground robot and apply the best rigid body transformation matrix to obtain the fused geolocation of the plant.

[0094] The archive construction unit is used to bind the fused multi-source data of each corn plant with its geographical location and update it to the same digital map to form a monitoring archive of the entire growth period at the single plant scale.

[0095] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring the entire life cycle of maize based on air-ground collaboration, characterized in that, The method includes: Multispectral images of cornfields are acquired from a top view using drones, and the inferred stalk diameter and top plane coordinates of each corn plant are detected and output. Corn stalk images are acquired from a side view using ground robots, and the measured stalk diameter and base plane coordinates of each corn plant are detected. Based on data measured by UAVs, a first plant distribution matrix is ​​constructed, which includes the inferred stem diameter and top plane coordinates of all plants; based on data measured by ground robots, a second plant distribution matrix is ​​constructed, which includes the measured stem diameter and base plane coordinates of all plants. The first plant distribution matrix is ​​matched with the second plant distribution matrix. The matching criteria include the relative positional relationship between plants based on planar coordinates, and the numerical correspondence between the inferred stem diameter and the measured stem diameter. Based on the matching results, the top monitoring data collected by the drone and the side monitoring data collected by the ground robot are associated with the same corn plant.

2. The method for monitoring the entire life cycle of maize based on air-ground collaboration according to claim 1, characterized in that, The steps of constructing a first plant distribution matrix containing the inferred stem diameter and top plane coordinates of all plants based on data measured by UAVs, and constructing a second plant distribution matrix containing the measured stem diameter and base plane coordinates of all plants based on data measured by ground robots, specifically include: For UAV data, the top plane coordinates of each corn plant are used as nodes. The Euclidean distance between each plant and the coordinates of all other plants is calculated, and its K nearest neighbor plants are found to form a local topological subgraph centered on that plant. For ground robot data, perform the same operations as for UAV data, and construct a local topological subgraph with the same K value for each plant based on the base plane coordinates; In each local topological subgraph, the diameter of the central plant, its distance vector to its K neighbors, and the diameters of the K neighbors themselves are combined into a feature vector. The set of feature vectors of all plants constitutes the first plant distribution matrix and the second plant distribution matrix.

3. The method for monitoring the entire life cycle of maize based on air-ground collaboration according to claim 1, characterized in that, The step of matching the first plant distribution matrix with the second plant distribution matrix, based on the relative positional relationship between plants using planar coordinates and the step of inferring the numerical correspondence between the stem diameter and the measured stem diameter, specifically includes: Calculate the comprehensive similarity between the feature vectors of any two local topological subgraphs in the first plant distribution matrix and the second plant distribution matrix. This comprehensive similarity is a weighted sum of the difference in the diameter of the central plant and the cosine similarity of the two distance vectors. Using a graph matching algorithm, with the overall similarity as the edge weight, we find the optimal matching pair that maximizes the total similarity of all plant correspondences in the two matrices; Geometric verification is performed on the optimal matching pairs. Using the successfully matched plant pairs, an optimal rigid body transformation matrix is ​​calculated to transform the UAV coordinate system to the ground robot coordinate system, and an open-ground plant mapping table is generated.

4. The method for monitoring the entire life cycle of maize based on air-ground collaboration according to claim 1, characterized in that, The step of associating the top monitoring data collected by the drone and the side monitoring data collected by the ground robot with the same corn plant based on the matching results specifically includes: Based on the open-ground plant mapping table, the multispectral vegetation index data obtained by the drone for each corn plant is bound to the ground surface data collected by the ground robot for the corresponding plant. Using the base plane coordinates collected by the ground robot as a reference, the top plane coordinates of the corresponding plant collected by the UAV are corrected by applying the optimal rigid body transformation matrix to obtain the geographical location of the plant after fusion. The multi-source data of each corn plant is merged and bound to its geographical location, and then updated to the same digital map to form a monitoring file for the entire growth period at the single plant scale.

5. The method for monitoring the entire life cycle of maize based on air-ground collaboration according to claim 1, characterized in that, The ground robot uses lidar to measure the diameter of the plant.

6. A corn life-cycle monitoring system based on air-ground collaboration, characterized in that, The system includes: The data acquisition module is used to acquire multispectral images of cornfields from a top view using a drone, detect and output the inferred stalk diameter and top plane coordinates of each corn plant; and to acquire corn stalk images from a side view using a ground robot, and detect the measured stalk diameter and base plane coordinates of each corn plant. The matrix construction module is used to construct a first plant distribution matrix containing the inferred stem diameter and top plane coordinates of all plants based on data measured by UAVs; and to construct a second plant distribution matrix containing the measured stem diameter and base plane coordinates of all plants based on data measured by ground robots. The location matching module is used to match the first plant distribution matrix with the second plant distribution matrix. The matching criteria include the relative positional relationship between plants based on planar coordinates, and the numerical correspondence between the inferred stem diameter and the measured stem diameter. The data matching module is used to associate the top monitoring data collected by the drone with the side monitoring data collected by the ground robot to the same corn plant based on the matching results.

7. The corn life-cycle monitoring system based on air-ground collaboration according to claim 6, characterized in that, The matrix construction module includes: The aerial data topology unit is used to calculate the Euclidean distance between the top plane coordinates of each corn plant and the coordinates of all other plants for UAV data, and to find its K nearest neighbor plants, forming a local topological subgraph centered on that plant. Ground data topology unit is used to perform the same operations on ground robot data as on UAV data, and to construct a local topology subgraph with the same K value for each plant based on the base plane coordinates; The matrix generation unit is used to combine the diameter of the central plant, its distance vector to K neighbors, and the diameters of the K neighbors themselves into a feature vector in each local topological subgraph. The set of feature vectors of all plants constitutes the first plant distribution matrix and the second plant distribution matrix.

8. The corn full life cycle monitoring system based on air-ground collaboration according to claim 6, characterized in that, The location matching module includes: The similarity calculation unit is used to calculate the comprehensive similarity between the feature vectors of any two local topological subgraphs in the first plant distribution matrix and the second plant distribution matrix. The comprehensive similarity is a weighted sum of the difference in the diameter of the central plant and the cosine similarity of the two distance vectors. The plant matching unit is used to find the optimal matching pair that maximizes the total similarity of all plant correspondences in two matrices by using a graph matching algorithm with the comprehensive similarity as the edge weight. The mapping table construction unit is used to perform geometric verification on the optimal matching pair. Using the successfully matched plant pairs, it calculates an optimal rigid body transformation matrix that transforms the UAV coordinate system to the ground robot coordinate system, and generates an open-ground plant mapping table.

9. The corn life-cycle monitoring system based on air-ground collaboration according to claim 6, characterized in that, The data matching module includes: The data binding unit is used to bind the multispectral vegetation index data obtained by the drone for each corn plant with the ground data collected by the ground robot for the corresponding plant, according to the open-ground plant mapping table. The geolocation transformation unit is used to correct the top plane coordinates of the corresponding plant collected by the UAV based on the base plane coordinates collected by the ground robot and apply the best rigid body transformation matrix to obtain the fused geolocation of the plant. The archive construction unit is used to bind the fused multi-source data of each corn plant with its geographical location and update it to the same digital map to form a monitoring archive of the entire growth period at the single plant scale.

10. The corn life-cycle monitoring system based on air-ground collaboration according to claim 6, characterized in that, The ground robot uses lidar to measure the diameter of the plant.

Images