Fruit tree canopy three-dimensional model reconstruction method and reconstruction system based on multimodal data

Abstract

The present invention provides a method and system for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data, and relates to the technical field of three-dimensional model reconstruction of fruit tree canopies based on multimodal data. The present invention synchronously acquires three-dimensional point cloud data and spectral images of the fruit tree canopy through laser radar and hyperspectral imaging, realizes sub-pixel spatial alignment with the help of ground control points, and performs voxel division based on the range of the point cloud bounding box, screens the number of point clouds to construct an initial seed set; adopts a structure-spectrum collaborative growth strategy, takes unvisited adjacent voxels as candidate voxels, calculates the mean spectral similarity between the adjacent voxels and the directly adjacent voxels in the seed set, terminates the growth when no new voxels are added after three consecutive iterations, outputs the final seed set, and converts the set into a three-dimensional canopy model that integrates spatial coordinates and full-band spectral reflectance, realizes voxel-level pathology annotation based on the chlorophyll sensitivity index and the blue-red slope, and completes the entire reconstruction and analysis process.

Description

Technical Field

[0001] The present invention relates to the technical field of fruit tree canopy three-dimensional model reconstruction based on multimodal data, and in particular to a fruit tree canopy three-dimensional model reconstruction method and reconstruction system based on multimodal data. Background Art

[0002] Information on the structure and physiological state of fruit tree canopies is crucial for precision agriculture management and the early identification of pests and diseases. Traditional three-dimensional canopy modeling methods often rely on single-modal data, such as LiDAR point clouds or visible light imagery. While these methods offer advantages in spatial structure modeling and species classification, they lack information from spectral dimensions like hyperspectral data and are unable to effectively represent the spectral reflectance characteristics of vegetation. This is particularly true for spatial identification of canopy growth status, physiological stress, or disease distribution. Furthermore, existing point cloud and image registration methods often only achieve rough spatial alignment and lack the ability to account for natural plant disturbances, such as wind-induced vibrations, and sub-pixel mapping accuracy. This results in significant registration errors, impacting the accuracy of modeling results.

[0003] Existing technologies attempt to project hyperspectral data onto point cloud models, but due to insufficient spatial registration accuracy and wind-induced displacement of branches and leaves, the spectral and structural data are spatially misaligned, making it impossible to establish a sub-voxel mapping relationship. Although pure spectral analysis methods can detect the physiological state of leaves, they lack three-dimensional spatial positioning capabilities and find it difficult to distinguish the distribution of diseases in the inner and outer layers of the canopy. More importantly, existing canopy segmentation algorithms, such as the region growing method, only rely on a single criterion of spatial continuity or spectral similarity, which can easily lead to overgrowth in complex fruit tree scenes, such as the mistaken inclusion of the background sky, or growth interruption, such as the omission of occluded areas. For example, the morphological method of Chen et al. ignores spectral consistency and misjudges dead branches as healthy canopies.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The object of the present invention is to provide a method and system for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data, comprising the following steps:

[0008] Step 1: Obtain the three-dimensional structural information and spectral information of the fruit tree canopy in the monitored area. The three-dimensional structural information is point cloud data obtained by laser radar scanning, and the spectral information is a hyperspectral image corresponding to the canopy. Use ground control points to unify the point cloud and image into a world coordinate system, and establish a precise mapping relationship between the point cloud coordinates and the hyperspectral pixels.

[0009] Step 2: Based on the bounding box range of the registered point cloud data, the range is divided into equal intervals at a preset voxel scale to generate voxel units covering the entire canopy space to construct the current voxel set. The number of point clouds contained in each voxel unit is counted, and whether it is occupied is determined according to the preset occupancy threshold. The voxel units that meet the conditions are formed into a seed voxel set;

[0010] Step 3: Using the seed voxel set as the starting point for growth, the structure-spectrum co-growth strategy is used for iterative expansion. In each round of iteration, a candidate voxel that is directly spatially adjacent to any voxel in the seed voxel set is selected from the current voxel set. For this candidate voxel, the spectral feature similarity between it and the spatially adjacent voxel in the seed voxel set is calculated. Only when the spectral feature similarity of the candidate voxel meets the preset consistency threshold is it included in the seed voxel set. If no new voxel is included in the current voxel set for three consecutive iterations, the growth process is considered to be terminated and the final seed voxel set is output.

[0011] Step 4: Convert the final seed voxel set into a three-dimensional canopy model that integrates spatial structure and spectral information. In this model, each voxel unit contains its three-dimensional spatial coordinates and its corresponding spectral reflectance data. Based on the spectral reflectance characteristics of the voxel unit, the physiological state of the vegetation it represents is automatically labeled to complete the three-dimensional reconstruction and physiological state analysis of the canopy.

[0012] Furthermore, the logic for obtaining the three-dimensional structural information and spectral information of the fruit tree canopy in the monitored area is as follows:

[0013] At least six high-reflectivity ground control points are deployed in the monitoring area, and lidar scanning is used to obtain canopy point cloud data and the millimeter-level precision three-dimensional coordinates of the control points. A hyperspectral image covering the canopy area is simultaneously acquired, and the pixel coordinates of each control point in the image are extracted. The spatial transformation matrix between the point cloud and the hyperspectral image is calculated based on a bidirectional least squares matching algorithm, and thin-plate spline deformation compensation is applied to the canopy edge area to eliminate geometric distortion caused by wind-induced branch and leaf displacement.

[0014] A sub-pixel precision mapping relationship is established between the point cloud and the hyperspectral pixels to ensure that the registration error does not exceed 0.2 pixels.

[0015] Furthermore, based on the bounding box range of the registered point cloud data, the logic of dividing the range into equal intervals at a preset voxel scale is as follows:

[0016] A 3D bounding box based on the registered point cloud data with a preset voxel size As the basic unit, it is divided into equal intervals in the directions of the X, Y, and Z coordinate axes to generate three-dimensional voxel units covering the entire canopy space;

[0017] The three-dimensional bounding box range is determined by the minimum value of the point cloud coordinates and maximum value Make sure that all point clouds are contained within the voxel unit;

[0018] Only when the number of point clouds in a voxel unit is greater than or equal to 3 and the standard deviation of the Euclidean distance between all point pairs in the voxel conform to When the voxel unit is determined to meet the occupancy condition and is included in the initial seed voxel set, after being included in the set, atmospheric scattering correction and radiation compensation related to the solar altitude angle are performed on the hyperspectral data mapped to the voxel unit to improve the consistency and comparability of the spectral data.

[0019] Furthermore, the logic of iterating by using the seed voxel set as the starting point of growth and adopting the structure-spectrum collaborative growth strategy is as follows:

[0020] The iterative process of the structure-spectrum collaborative voxel growth strategy specifically includes:

[0021] Iterate with the seed voxel set as the starting point of growth and mark the seed voxel set All voxels in the intersection with the current voxel set are in the visited state. For the seed voxel set of iteration t, traverse the 26 spatial adjacent positions of each visited voxel unit in the current voxel set. If there is an unvisited voxel unit in the adjacent position, mark it as a candidate voxel c, and confirm that the candidate voxel c is spatially adjacent to at least one voxel unit in the seed voxel set;

[0022] For each candidate voxel c, calculate its spectral angle similarity with all voxels in the seed voxel set The formula is:

[0023] ;

[0024] in, is a wavelength variable ranging from 400 nm to 2500 nm, For candidate voxels at wavelength The reflectivity, is the average reflectivity of the candidate voxels, is the i-th seed voxel in The reflectivity at is the average reflectivity of the i-th seed voxel, is the number of voxels in the seed voxel set, i is the index of the seed voxel in the seed voxel set;

[0025] If satisfied , then the candidate voxel is considered to be consistent with the seed voxel set There is consistency between them in the spectral space, allowing them to be joined and the seed voxel set to be updated.

[0026] Furthermore, the candidate voxels are added to the seed voxel set Another constraint is that the spatial connectivity satisfies the following conditions:

[0027] ;

[0028] in, is the sum of the volumes of candidate voxels and interstitial voxels in the tth iteration, where the interstitial voxel is the spatial position between the candidate voxel and the seed voxel set. The voxel units between them but not yet occupied by any voxel, is the volume of the seed voxel set, for Endosome prime number, is the voxel volume, are the candidate voxel center coordinates, for The coordinates of the voxel center with the closest Euclidean distance to the candidate voxel, represents the Euclidean distance.

[0029] Furthermore, the growth termination must simultaneously meet the following requirements: 1) the number of newly added voxels in three consecutive iterations does not exceed 0.1% of the total number of seed voxels in the previous iteration; and 2) the spectral information entropy converges, specifically:

[0030] ;

[0031] in, represents the Shannon entropy of the reflectivity of the seed voxel set in the 680-750 nm band at the t-th iteration, is the reflectivity histogram partition index, is the number of reflectance histogram partitions, is the probability that the reflectivity value falls into the kth interval at the tth iteration;

[0032] When there are no new voxels in three consecutive iterations and the number of new voxels in three consecutive iterations does not exceed 0.1% of the total number of seed voxel sets in the previous iteration, the growth is terminated and the final seed voxel set is output. .

[0033] Furthermore, the final seed voxel set Each voxel unit in is resolved into a three-dimensional space entity, whose properties include: geometric properties and spectral properties:

[0034] The geometric attributes are the voxel center coordinates and preset voxel sizes , the spectral attribute is the hyperspectral reflectance vector mapped by the voxel; a three-dimensional canopy model integrating structure and spectrum is constructed based on the geometric attributes and spectral attributes, and the model is stored in the form of voxel units, each unit containing coordinates and reflectance data;

[0035] The logic for automatically labeling the physiological status of vegetation based on its spectral characteristics is as follows:

[0036] Calculate the chlorophyll sensitivity index of voxel cells in the 3D canopy model at 680 nm and 750 nm: ,when , the voxel is judged to be in a stress state, otherwise it is marked as healthy; are the reflectance of the voxel at 750nm and 680nm, respectively;

[0037] For voxels judged to be stressed, the reflectance slope of the blue band is further calculated. ,like , it is identified as a bacterial disease; the reflectivity slope of the red band ,like , it is determined to be a fungal disease; if neither of the two conditions is met, it is marked as an unknown stress; are the reflectance of the voxel in the 490nm, 530nm, and 700nm bands, respectively;

[0038] The pathological label of each voxel and its spatial coordinates are used to output a three-dimensional canopy pathology spatial distribution map for vegetation disease location and health status assessment.

[0039] The present invention further provides a system for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data. The system is used to execute the above-mentioned method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data, comprising:

[0040] A data acquisition module is used to obtain the three-dimensional structural information and spectral information of the fruit tree canopy in the monitored area. The three-dimensional structural information is point cloud data obtained by laser radar scanning, and the spectral information is a hyperspectral image corresponding to the canopy. The point cloud and image are unified into a world coordinate system through ground control points to establish a precise mapping relationship between point cloud coordinates and hyperspectral pixels.

[0041] The set determination module is used to divide the bounding box range of the registered point cloud data into equal intervals at a preset voxel scale, generate voxel units covering the entire canopy space to construct the current voxel set, count the number of point clouds contained in each voxel unit, and determine whether it is occupied according to a preset occupancy threshold. The voxel units that meet the conditions are formed into a seed voxel set;

[0042] An iterative update module is used to use the seed voxel set as the starting point for growth and iteratively expand using a structural-spectral collaborative growth strategy. In each round of iteration, a candidate voxel that is directly spatially adjacent to any voxel in the seed voxel set is selected from the current voxel set. For each candidate voxel, the spectral feature similarity between it and the spatially adjacent voxels in the seed voxel set is calculated. Only when the spectral feature similarity of the candidate voxel meets a preset consistency threshold is it included in the seed voxel set. If no new voxel is included in the current voxel set for three consecutive iterations, the growth process is determined to be complete and the final seed voxel set is output.

[0043] The model analysis module is used to convert the final seed voxel set into a three-dimensional canopy model that integrates spatial structure and spectral information. In this model, each voxel unit contains its three-dimensional spatial coordinates and its corresponding spectral reflectance data. Based on the spectral reflectance characteristics of the voxel unit, the physiological state of the vegetation it represents is automatically labeled to complete the three-dimensional reconstruction and physiological state analysis of the canopy.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The present invention unifies the three-dimensional spatial structure information and the spectral reflectance information into the same coordinate system by constructing a precise mapping relationship between the point cloud based on ground control points and the hyperspectral image, thereby effectively solving the problem of insufficient multimodal data registration accuracy in the existing technology; using the precise registration result, the corresponding structural points in the hyperspectral image can be accurately located in the voxel division stage, thereby improving the basic accuracy of the subsequent joint modeling of canopy structure and physiological state; further, by generating a seed voxel set through spatial division and occupancy judgment of the point cloud data, redundant or non-canopy area data can be filtered out, thereby significantly improving the reliability and representativeness of the voxel growth starting point, and avoiding erroneous growth and model drift caused by improper initial area selection.

[0046] The present invention introduces a structural-spectral collaborative growth strategy in the voxel expansion process: in each round of iteration, the candidate voxels are strictly limited to be spatially adjacent to the current voxel set, and their spectral characteristics are required to meet a consistency threshold with the candidate voxels. Only voxels that meet both spatial connectivity and spectral similarity conditions can be included in the growth model; this strategy effectively avoids the pseudo-extension problem caused by structural noise or spectral outliers, ensuring the natural closure of the model boundary and the coherence of the spectral expression.

[0047] The present invention adopts a dynamic convergence mechanism of "terminating growth after three consecutive rounds of no new voxels", which can automatically determine whether the model has reached a stable state, avoid over-expansion or falling into local abnormal fluctuations, and improve the robustness and intelligence level of canopy boundary judgment. Through the synergistic effect of the above technical features, the present invention can achieve three-dimensional canopy modeling with high structural restoration and high spectral expression accuracy, which is significantly better than the existing modeling effect based on single modality or rule expansion methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Schematic diagram of the overall method flow of the present invention;

[0049] Figure 2 This is a flow chart of the overall system module of the present invention;

[0050] Figure 3 is a curve diagram of spectral angle similarity-average reflectivity difference of the present invention;

[0051] Figure 4 A line graph showing the average reflectivity of candidate voxels versus the average reflectivity of seed voxels in the present invention;

[0052] Figure 5 is a reflectance slope curve of the chlorophyll sensitivity index-red band of the present invention;

[0053] Figure 6 This is a line graph of the reflectance slope of the chlorophyll sensitivity index-blue band of the present invention. DETAILED DESCRIPTION

[0054] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0055] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0056] Example:

[0057] See also Figures 1-6 , the present invention provides a technical solution:

[0058] A method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data, comprising the following steps:

[0059] Step 1: Obtain the three-dimensional structural information and spectral information of the fruit tree canopy in the monitored area. The three-dimensional structural information is point cloud data obtained by laser radar scanning, and the spectral information is a hyperspectral image corresponding to the canopy. Use ground control points to unify the point cloud and image into a world coordinate system, and establish a precise mapping relationship between the point cloud coordinates and the hyperspectral pixels.

[0060] The logic for obtaining the three-dimensional structural information and spectral information of the fruit tree canopy in the monitored area is as follows:

[0061] At least six high-reflectivity ground control points are deployed in the monitoring area, and lidar scanning is used to obtain canopy point cloud data and the millimeter-level precision three-dimensional coordinates of the control points. A hyperspectral image covering the canopy area is simultaneously acquired, and the pixel coordinates of each control point in the image are extracted. The spatial transformation matrix between the point cloud and the hyperspectral image is calculated based on a bidirectional least squares matching algorithm, and thin-plate spline deformation compensation is applied to the canopy edge area to eliminate geometric distortion caused by wind-induced branch and leaf displacement.

[0062] Establish sub-pixel precision mapping between point cloud and hyperspectral pixels, ensuring that the registration error does not exceed 0.2 pixels;

[0063] To achieve spatial registration between the laser point cloud and the hyperspectral image, a set of ground control points with known three-dimensional spatial coordinates and observable image pixel coordinates must be deployed within the monitoring area. No fewer than six control points should be deployed for the following reasons: 3D affine / projection transformation fitting usually requires at least six control points to ensure the uniqueness and robustness of the solution. The control points should be evenly distributed across the entire canopy to avoid increased registration errors at the edges. Highly reflective materials, such as painted metal balls and white PVC, should be used for the control points to ensure they are clearly identifiable in both the lidar and hyperspectral images.

[0064] LiDAR uses laser beam scanning to obtain three-dimensional structural information, and can output high-density point clouds including canopy surface, trunk structure, and branch and leaf distribution. During the scanning process, the reflection intensity of control points is significantly higher than that of background vegetation, making it easier to extract the control point positions from the point cloud. To ensure registration accuracy, the coordinate accuracy of the control points must reach the millimeter level, such as within ±2mm. This is often obtained through a combination of high-precision RTK-GNSS and a total station.

[0065] Synchronous acquisition refers to using a hyperspectral imaging system, such as a visible-near infrared push-broom hyperspectral camera, to collect spectral images of the canopy area within a time window synchronized with the lidar scan as much as possible. The hyperspectral image is a three-dimensional data cube. ,in Representing the band dimension, control points can also appear as obvious feature points in hyperspectral images due to their special color or reflectivity characteristics, such as bright circular spots. Their two-dimensional pixel coordinates can be obtained through image processing methods such as brightness threshold plus template matching;

[0066] The bidirectional least squares matching algorithm is a geometric registration method that achieves optimal alignment between the coordinates of the control points of a point cloud and the coordinates of the image pixels. The core idea of ​​the algorithm is to simultaneously minimize the projection error of the point cloud onto the image plane and the reconstruction error of the image pixels back-projected into three-dimensional space, so that the transformation matrix achieves the minimum residual in both spaces. The output is a spatial transformation matrix, which includes rotation, translation, and scale change parameters, which is used to accurately map the three-dimensional coordinates to the image coordinate system.

[0067] In actual monitoring of fruit tree canopies, wind disturbances cause leaves and branches to jitter at different times, resulting in non-rigid offsets between point clouds and images. Thin plate splines are a smooth, non-rigid deformation model that can model local nonlinear differences between point sets. Based on control point registration, this method applies TPS deformation compensation to the canopy edge area (most affected by wind), ensuring that the registration relationship between the point cloud and the image remains locally continuous and physically consistent, significantly improving registration quality.

[0068] After completing the transformation matrix calculation and deformation compensation, each point in the 3D point cloud is projected into the image coordinate system to obtain its corresponding pixel coordinates. Since the pixel coordinates are floating-point data, rounding only the integer pixels may introduce significant errors. Therefore, bilinear interpolation or cubic interpolation methods are used to correspond the 3D points to the sub-pixel positions on the image. Finally, a one-to-one correspondence table between point cloud points and hyperspectral pixels is constructed to ensure that a hyperspectral vector can be obtained for each point cloud. To ensure accuracy, this scheme sets the maximum allowable projection residual to 0.2 pixels, and further reduces the error through residual elimination and iterative optimization.

[0069] Step 2: Based on the bounding box range of the registered point cloud data, the range is divided into equal intervals at a preset voxel scale to generate voxel units covering the entire canopy space to construct the current voxel set. The number of point clouds contained in each voxel unit is counted, and whether it is occupied is determined according to the preset occupancy threshold. The voxel units that meet the conditions are formed into a seed voxel set;

[0070] Based on the bounding box range of the registered point cloud data, the logic of dividing the range into equal intervals at the preset voxel scale is as follows:

[0071] A 3D bounding box based on the registered point cloud data with a preset voxel size As the basic unit, it is divided into equal intervals in the directions of the X, Y, and Z coordinate axes to generate three-dimensional voxel units covering the entire canopy space;

[0072] The registered point cloud data refers to the 3D point set that has been registered with the LiDAR point cloud and hyperspectral image in the previous stage. It has been unified into a common world coordinate system to ensure a one-to-one correspondence between the structure and the spectral information. The 3D bounding box is a minimum rectangular space bounding box that is used to completely contain all point clouds. The bounding box can be constructed by counting the minimum / maximum values ​​of all point cloud points along each axis.

[0073] Preset voxel sizes It is the side length of a single cube voxel, which determines the spatial resolution of the 3D grid. The smaller the resolution, the finer the model. The equal spacing represents the X, Y, and Z directions, from the minimum to the maximum value. The step size is used to divide the spatial coordinates, and the entire space is divided into three-dimensional voxel grids to form a regular spatial grid structure;

[0074] The three-dimensional bounding box range is determined by the minimum value of the point cloud coordinates and maximum value Make sure that all point clouds are contained within the voxel unit;

[0075] The number of voxels divided in each coordinate direction is:

[0076] ;

[0077] Only when the number of point clouds in a voxel unit is greater than or equal to 3 and the standard deviation of the Euclidean distance between all point pairs in the voxel conform to When the voxel unit is determined to meet the occupancy condition and is included in the seed voxel set, after being included in the set, atmospheric scattering correction and radiation compensation related to the solar altitude angle are performed on the hyperspectral data mapped to the voxel unit to improve the consistency and comparability of the spectral data;

[0078] At least three points are needed to form the minimum geometric structure to prevent the introduction of noise or isolated points. The Euclidean distance is calculated for any two points in the voxel, and the standard deviation of the distance is calculated for all points. , represents the maximum possible distance between any two points in the voxel, is the threshold of the allowed space compactness;

[0079] The established sub-pixel mapping relationship between point clouds and hyperspectral images can be used to reversely check the spectral pixel positions corresponding to all points in each voxel, extract its hyperspectral data, and eliminate spectral distortions caused by atmospheric molecules / aerosol scattering. For example, MODTRAN / 6S or empirical linear methods can be used for correction, considering the influence of the sun angle on the radiation intensity during shooting, and using the cosine law to adjust the spectral reflectance value, so that images collected at different times / angles are comparable.

[0080] Step 3: Using the seed voxel set as the starting point for growth, the structure-spectrum co-growth strategy is used for iterative expansion. In each round of iteration, a candidate voxel that is directly spatially adjacent to any voxel in the seed voxel set is selected from the current voxel set. For this candidate voxel, the spectral feature similarity between it and the spatially adjacent voxel in the seed voxel set is calculated. Only when the spectral feature similarity of the candidate voxel meets the preset consistency threshold is it included in the seed voxel set. If no new voxel is included in the current voxel set for three consecutive iterations, the growth process is considered to be terminated and the final seed voxel set is output.

[0081] The logic of using the seed voxel set as the starting point of growth and adopting the structure-spectrum collaborative growth strategy for iteration is as follows:

[0082] The iterative process of the structure-spectrum collaborative voxel growth strategy specifically includes:

[0083] Iterate with the seed voxel set as the starting point of growth and mark the seed voxel set All voxels in the intersection with the current voxel set are in the visited state. For the seed voxel set of iteration t, traverse the 26 spatial adjacent positions of each voxel unit in the current voxel set. If there is an unvisited voxel unit in the adjacent position, mark it as a candidate voxel c, and confirm that the candidate voxel c is spatially adjacent to at least one voxel unit in the seed voxel set;

[0084] Each voxel has 26 directly adjacent voxels (excluding itself), which are called the three-dimensional Moore neighborhood. Candidate voxels are adjacent voxels that have not participated in the iterative calculation before. Only candidate voxels that are adjacent to at least one voxel in the seed voxel set are retained to ensure structural connectivity;

[0085] Candidate voxels are added to the seed voxel set Another constraint is that the spatial connectivity satisfies the following conditions:

[0086] ;

[0087] in, is the sum of the volumes of candidate voxels and interstitial voxels in the tth iteration, is the volume of the seed voxel set, for Endosome prime number, is the voxel volume, are the candidate voxel center coordinates, for The coordinates of the voxel center with the closest Euclidean distance to the candidate voxel, represents the Euclidean distance;

[0088] is the set of candidate voxels and seed voxels in the tth iteration The total volume of the space between the void voxels, the void voxel refers to the spatial position between the candidate voxel and the seed voxel set The voxel units between them but not yet occupied by any voxel;

[0089] Setting strict thresholds , which means that there is almost no interrupted voxel jump space between the candidate voxels and the existing model, and they are very closely connected in space; The larger the value, the larger the spatial jump between the candidate voxel and the existing model, which may be unreasonable fracture growth, such as sudden generation through air. If it is smaller, it means that the structure is compact and the spatial connectivity is good, which allows growth.

[0090] It is used to restrict candidate voxels from being too far away. Even if their spectra are consistent, they are not allowed to be directly connected if they are far apart in structure. The maximum distance is set to 3 times the voxel length to accommodate edge voxels without sacrificing the compactness of the model. The smaller it is, the closer the candidate voxel is to the existing model structure and the compact spatial structure is. The larger it is, the more severe the structural break is and access is not allowed, even if the spectrum is consistent.

[0091] These two constraints work together to restrict the growth behavior of candidate voxels from the perspectives of macroscopic voxel density (ratio) and microscopic geometric adjacency (minimum distance), respectively, ensuring the continuity of the model's spatial structure and smooth edge transitions. This suppresses abnormal structural expansion, such as jumping branches and erroneous growth, and improves the physical rationality and geometric consistency of the 3D canopy model.

[0092] For each candidate voxel c, calculate its spectral angle similarity with all voxels in the seed voxel set The formula is:

[0093] ;

[0094] in, is a wavelength variable ranging from 400 nm to 2500 nm, For candidate voxels at wavelength The reflectivity, is the average reflectivity of the candidate voxels, is the i-th seed voxel in The reflectivity at is the average reflectivity of the i-th seed voxel, is the number of voxels in the seed voxel set, i is the index of the seed voxel in the seed voxel set;

[0095] is the average spectral angle similarity of the candidate voxel c, which measures the average similarity between the candidate voxel and all adjacent voxels in the spectral space. The smaller the similarity, the more similar it is. An angle of 0 means complete consistency. Bands representing hyperspectral reflectance, the integration process covers all bands of interest, Reflects the plant tissue status of the voxel, such as water content, chlorophyll, etc. , Used to remove the influence of spectral intensity differences and focus on spectral shape matching;

[0096] It represents the spectral co-deviation of two voxels, which is similar to covariance and measures the consistency of the two voxels at each wavelength after they deviate from their respective mean values. and are the energy or standard deviation of the two voxel spectra, respectively, which are used for normalization to offset the absolute amplitude difference and focus only on shape matching; For cosine similarity, the similarity is mapped to an angle. The smaller the angle, the better. A smaller angle means the spectral direction is more consistent.

[0097] The specific data of some sample numbers and spectral angle similarity are shown in Table 1.

[0098]

[0099] By analyzing the above 15 groups of voxel spectral similarity data, a positive correlation was observed between the average reflectance difference between voxels and the spectral angle similarity. For example, the average reflectances of the candidate voxels and seed voxels of sample number 1 were 0.31 and 0.29, respectively, with a corresponding difference of 0.02. The number of seed voxels was 5, and the calculated spectral angle similarity was only 0.153, indicating that under the conditions of small reflectance difference and moderate number of neighbors, the spectral shapes between voxels are highly consistent. Similarly, the difference of sample number 6 is smaller (0.01) and the number of seed voxels is larger (8), and its spectral angle similarity further decreases to 0.134, indicating that more spatial neighborhood metrics can improve the robustness of similarity calculation.

[0100] For most samples (e.g., samples 3, 7, 9, and 13), the spectral angle similarity remained low, ranging from 0.16 to 0.21, when the reflectance difference remained between 0.02 and 0.04. This indicates that as long as the reflectance difference between the candidate voxel and the seed voxel does not exceed 0.05 and the number of seed voxels is at least 4, the voxel's spectral shape exhibits little significant deviation and can be successfully incorporated into the canopy model. For example, sample 8, with a reflectance difference of 0.04 and 6 seed voxels, corresponds to a similarity of 0.210. While slightly higher than the 0.153 for sample 1, it remains below the commonly used growth threshold of approximately 0.55 (corresponding to a cosine similarity of 0.85).

[0101] When the difference approaches the upper threshold (e.g., voxels 4 and 10, both with a difference of 0.055), the similarity rises above 0.25, reflecting increased spectral shape deviation and closer to the rejection boundary for the growth criterion. The spectral angle similarity of 0.251 for voxels 4 and 0.256 for voxels 10 are both significantly higher than those for voxel differences of 0.02–0.04. This indicates that when the reflectance difference exceeds 0.05, even with a large number of seed voxels (4 for voxels 10), significant spectral angle deviation will occur, necessitating a comprehensive assessment based on spatial connectivity and other criteria. Overall, this data distribution validates the mathematical logic between "reflectance shape difference" and "spectral similarity angle" in the formula, providing empirical evidence for setting voxel growth thresholds.

[0102] If satisfied , then the candidate voxel is considered to be consistent with the seed voxel set There is consistency between them in the spectral space, allowing them to join and update the seed voxel set;

[0103] The growth termination must simultaneously meet the following requirements: 1) the number of newly added voxels in three consecutive iterations does not exceed 0.1% of the total number of seed voxels in the previous iteration; and 2) the spectral information entropy converges, specifically:

[0104] ;

[0105] in, represents the Shannon entropy of the reflectivity of the seed voxel set in the 680-750 nm band at the t-th iteration, is the reflectivity histogram partition index, is the number of reflectance histogram partitions, is the probability that the reflectivity value falls into the kth interval at the tth iteration;

[0106] When there are no new voxels in three consecutive iterations and the number of new voxels in three consecutive iterations does not exceed 0.1% of the total number of seed voxel sets in the previous iteration, the growth is terminated and the final seed voxel set is output. ;

[0107] The band reflectivity is divided into intervals, extract all seed voxel sets The reflectivity curve of each voxel in the range of 680–750 nm is obtained, and the distribution of these reflectivity data is counted to construct a normalized histogram: the total number of samples is N, and the kth interval has falls into the value, then the probability is , substitute into the entropy formula for calculation;

[0108] entropy A measure of the spectral uncertainty or diversity of the seed voxel set. If the reflectance distribution is more dispersed, that is, the reflectance differences between different bands are large, or the differences between voxels are large, the entropy value is high; if the reflectance is concentrated at certain values, that is, the voxel spectral consistency is enhanced, the entropy value is low;

[0109] like More evenly distributed, more dispersed, and higher entropy. More concentrated, the distribution is more concentrated, the entropy is lower, if the number of voxels N increases but the distribution remains unchanged, it does not affect the entropy, the entropy remains unchanged, if If there are more, the partition will be refined, the entropy may increase, and it will be more sensitive;

[0110] It means that the current information entropy change amplitude has dropped to less than 5% compared with the previous round, indicating that the spectral distribution of the seed voxel set has stabilized and the newly added voxels have not significantly changed the spectral characteristics of the model. If the entropy value changes for three consecutive rounds are less than the threshold (which can be expanded to a sliding window judgment), it can be judged that the spectral structure has converged.

[0111] Step 4: Convert the final seed voxel set into a 3D canopy model that integrates spatial structure and spectral information. In this model, each voxel unit contains its 3D spatial coordinates and its corresponding spectral reflectance data. Based on the spectral reflectance characteristics of the voxel unit, the physiological state of the vegetation it represents is automatically labeled to complete the 3D reconstruction and physiological state analysis of the canopy.

[0112] The final seed voxel set Each voxel unit in is resolved into a three-dimensional space entity, whose properties include: geometric properties and spectral properties:

[0113] This is a structured definition for modeling voxel units. Each voxel is a three-dimensional cube whose position and properties are described by the following two types of information: Geometric properties refer to the position of the voxel in three-dimensional space (i.e., its center coordinates X, Y, Z and the length of the cube side). , used to reconstruct its physical form and spatial relationship; the spectral attribute is the hyperspectral reflectance vector corresponding to the voxel, representing the spectral reflectance of each band in the range of 400–2500 nm in the voxel mapping area;

[0114] The geometric attributes are the voxel center coordinates and preset voxel sizes , the spectral attribute is the hyperspectral reflectance vector mapped by the voxel; a three-dimensional canopy model integrating structure and spectrum is constructed based on the geometric attributes and spectral attributes, and the model is stored in the form of voxel units, each unit containing coordinates and reflectance data;

[0115] This step completes the three-dimensional digital modeling of the canopy. The goal is to establish a unified representation of spatial structure and spectral characteristics: the voxel grid constitutes the spatial skeleton of the entire canopy. Each voxel contains its center position and a set of multidimensional spectral characteristics, which is a coupling of structure and function. The voxel grid structure is stored to facilitate spatial traversal, adjacency judgment, and subsequent voxel-level diagnosis.

[0116] The logic for automatically labeling the physiological status of vegetation based on its spectral characteristics is as follows:

[0117] Calculate the chlorophyll sensitivity index of voxel cells in the 3D canopy model at 680 nm and 750 nm: ,when , the voxel is judged to be in a stress state, otherwise it is marked as healthy;

[0118] is the reflectivity of the voxel at 750nm (near-infrared reflection band) and 680nm (chlorophyll absorption band), is the chlorophyll sensitivity index, which measures the difference between red light absorption and near-infrared reflectance; The larger it is, the healthier the vegetation is and the more active its photosynthesis is; the smaller it is, the more restricted its photosynthesis is and the more likely it is diseased or under nutrient stress.

[0119] It is a near-infrared band, which refers to the reflection peak of the internal structure of healthy leaves and is an indicator of the integrity of the cell structure. The chlorophyll absorption valley refers to the chlorophyll concentration sensitive area, that is, the stronger the absorption, the lower the reflection;

[0120] like ,but , vegetation is healthy, if ,but , the vegetation status is suspicious, if ,but A negative number indicates an atypical or incorrect match;

[0121] For voxels judged to be stressed, the reflectance slope of the blue band is further calculated. ,like , it is identified as a bacterial disease; the reflectivity slope of the red band ,like , it is determined to be a fungal disease; if neither of the two conditions is met, it is marked as an unknown stress;

[0122] It reflects the characteristics of the rapid change zone of chlorophyll and indicates the rising speed of the reflectivity in the blue-edge band (490–530nm). The rapid rise of blue-edge reflectivity is related to changes in leaf cell structure and bacterial osmotic pressure. Bacterial diseases usually lead to an accelerated rise in blue-edge reflectivity. The value increased significantly;

[0123] It is the ratio of the reflectance at 680nm to that at 700nm, reflecting the relative depth of the red valley. Healthy vegetation has the strongest absorption at 680nm (red valley), and the reflection at 700nm rises rapidly. Fungal diseases will destroy the cell structure, causing the red valley to become shallower (increased reflectance at 680nm). rise; are the reflectance of the voxel in the 490nm, 530nm, and 700nm bands, respectively; and They are located at the lower and upper ends of the typical blue edge band, respectively. The blue edge range (about 490–530nm) is particularly sensitive to chlorophyll and secondary pigments. It is the green light band, which refers to the carotenoid reflection peak. It is the blue-green transition band, which is the lutein absorption area. is the starting point of the red edge, reflecting the transition from the leaf absorption valley to the near-infrared reflection area;

[0124] Some sample numbers and specific data of chlorophyll sensitivity index are shown in Table 2.

[0125]

[0126] By analyzing the spectral indicators of the first 15 voxel samples, the chlorophyll sensitivity index (CSI) can effectively distinguish between healthy and stressed states. For example, the CSI of sample number 1 is 0.382 (lower than 0.45), so it is judged to be in a stressed state; its blue edge slope BES is 0.028 (lower than 0.03), and the red valley ratio RVI is 1.045 (lower than 1.1). Both do not meet the criteria for bacterial or fungal diseases, so it is marked as "unknown stress". In contrast, the CSI of sample number 2 is 0.512 (higher than 0.45), which is directly judged to be healthy without further disease classification. This shows that CSI has a high discriminant power in the initial screening of voxel health.

[0127] In stressed samples with a CSI < 0.45, BES and RVI can indicate different disease types. For example, sample 3 has a CSI of 0.421 and a BES of 0.042 (>0.03), meeting the criteria for a bacterial disease. Sample 7 has a CSI of only 0.215 and a high RVI of 1.150 (>1.1), resulting in a fungal disease classification. Samples 4 and 11 (CSI 0.305 and 0.278, respectively) both exhibit high BES (0.050 and 0.040) and excessive RVI (1.035 and 1.200), but both are labeled as bacterial diseases because the BES threshold takes precedence. This demonstrates that, among the stressed voxels identified by the CSI, BES is particularly sensitive for identifying bacterial diseases, while RVI provides an effective supplementary criterion for fungal diseases. This dual-indicator combination enables precise localization of pathological conditions in fruit tree canopy voxels.

[0128] The pathological label of each voxel and its spatial coordinates are used to output a three-dimensional canopy pathology spatial distribution map for vegetation disease location and health status assessment.

[0129] See also Figure 2 The present invention further provides a system for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data. The system is used to execute the above-mentioned method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data, comprising:

[0130] A data acquisition module is used to obtain the three-dimensional structural information and spectral information of the fruit tree canopy in the monitored area. The three-dimensional structural information is point cloud data obtained by laser radar scanning, and the spectral information is a hyperspectral image corresponding to the canopy. The point cloud and image are unified into a world coordinate system through ground control points to establish a precise mapping relationship between point cloud coordinates and hyperspectral pixels.

[0131] The set determination module is used to divide the bounding box range of the registered point cloud data into equal intervals at a preset voxel scale, generate voxel units covering the entire canopy space to construct the current voxel set, count the number of point clouds contained in each voxel unit, and determine whether it is occupied according to a preset occupancy threshold. The voxel units that meet the conditions are formed into a seed voxel set;

[0132] An iterative update module is used to use the seed voxel set as the starting point for growth and iteratively expand using a structural-spectral collaborative growth strategy. In each round of iteration, a candidate voxel that is directly spatially adjacent to any voxel in the seed voxel set is selected from the current voxel set. For each candidate voxel, the spectral feature similarity between it and the spatially adjacent voxels in the seed voxel set is calculated. Only when the spectral feature similarity of the candidate voxel meets a preset consistency threshold is it included in the seed voxel set. If no new voxel is included in the current voxel set for three consecutive iterations, the growth process is determined to be complete and the final seed voxel set is output.

[0133] The model analysis module is used to convert the final seed voxel set into a three-dimensional canopy model that integrates spatial structure and spectral information. In this model, each voxel unit contains its three-dimensional spatial coordinates and its corresponding spectral reflectance data. Based on the spectral reflectance characteristics of the voxel unit, the physiological state of the vegetation it represents is automatically labeled to complete the three-dimensional reconstruction and physiological state analysis of the canopy.

[0134] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0135] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0136] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0137] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data, characterized in that: The specific steps include: Step 1: Obtain the three-dimensional structural information and spectral information of the fruit tree canopy in the monitored area. The three-dimensional structural information is point cloud data obtained by laser radar scanning, and the spectral information is a hyperspectral image corresponding to the canopy. Use ground control points to unify the point cloud and image into a world coordinate system, and establish a precise mapping relationship between the point cloud coordinates and the hyperspectral pixels. Step 2: Based on the bounding box range of the registered point cloud data, the range is divided into equal intervals at a preset voxel scale to generate voxel units covering the entire canopy space to construct the current voxel set. The number of point clouds contained in each voxel unit is counted, and whether it is occupied is determined according to the preset occupancy threshold. The voxel units that meet the conditions are formed into a seed voxel set; Step 3: Using the seed voxel set as the starting point for growth, the structure-spectrum co-growth strategy is used for iterative expansion. In each round of iteration, a candidate voxel that is directly spatially adjacent to any voxel in the seed voxel set is selected from the current voxel set. For this candidate voxel, the spectral feature similarity between it and the spatially adjacent voxel in the seed voxel set is calculated. Only when the spectral feature similarity of the candidate voxel meets the preset consistency threshold is it included in the seed voxel set. If no new voxel is included in the current voxel set for three consecutive iterations, the growth process is considered to be terminated and the final seed voxel set is output. Step 4: Convert the final seed voxel set into a three-dimensional canopy model that integrates spatial structure and spectral information. In this model, each voxel unit contains its three-dimensional spatial coordinates and its corresponding spectral reflectance data. Based on the spectral reflectance characteristics of the voxel unit, the physiological state of the vegetation it represents is automatically labeled to complete the three-dimensional reconstruction and physiological state analysis of the canopy.

2. The method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data according to claim 1, characterized in that: The logic for obtaining the three-dimensional structural information and spectral information of the fruit tree canopy in the monitored area is as follows: At least six high-reflectivity ground control points are deployed in the monitoring area, and lidar scanning is used to obtain canopy point cloud data and the millimeter-level precision three-dimensional coordinates of the control points. A hyperspectral image covering the canopy area is simultaneously acquired, and the pixel coordinates of each control point in the image are extracted. The spatial transformation matrix between the point cloud and the hyperspectral image is calculated based on a bidirectional least squares matching algorithm, and thin-plate spline deformation compensation is applied to the canopy edge area to eliminate geometric distortion caused by wind-induced branch and leaf displacement. A sub-pixel precision mapping relationship is established between the point cloud and the hyperspectral pixels to ensure that the registration error does not exceed 0.2 pixels.

3. The method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data according to claim 2, characterized in that: Based on the bounding box range of the registered point cloud data, the logic of dividing the range into equal intervals at the preset voxel scale is as follows: A 3D bounding box based on the registered point cloud data with a preset voxel size As the basic unit, it is divided into equal intervals in the directions of the X, Y, and Z coordinate axes to generate three-dimensional voxel units covering the entire canopy space; The three-dimensional bounding box range is determined by the minimum value of the point cloud coordinates and maximum value Make sure that all point clouds are contained within the voxel unit; Only when the number of point clouds in a voxel unit is greater than or equal to 3 and the standard deviation of the Euclidean distance between all point pairs in the voxel conform to When the voxel unit is determined to meet the occupancy condition and is included in the initial seed voxel set, after being included in the set, atmospheric scattering correction and radiation compensation related to the solar altitude angle are performed on the hyperspectral data mapped to the voxel unit to improve the consistency and comparability of the spectral data.

4. The method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data according to claim 3, characterized in that: The logic of using the seed voxel set as the starting point for growth and adopting the structure-spectrum collaborative growth strategy for iteration is as follows: The iterative process of the structure-spectrum collaborative voxel growth strategy specifically includes: Iterate with the seed voxel set as the starting point of growth and mark the seed voxel set All voxels in the intersection with the current voxel set are in the visited state. For the seed voxel set of iteration t, traverse the 26 spatial adjacent positions of each visited voxel unit in the current voxel set. If there is an unvisited voxel unit in the adjacent position, mark it as a candidate voxel c, and confirm that the candidate voxel c is spatially adjacent to at least one voxel unit in the seed voxel set; For each candidate voxel c, calculate its spectral angle similarity with all voxels in the seed voxel set The formula is: ； in, is a wavelength variable ranging from 400 nm to 2500 nm, For candidate voxels at wavelength The reflectivity, is the average reflectivity of the candidate voxels, is the i-th seed voxel in The reflectivity at is the average reflectivity of the i-th seed voxel, is the number of voxels in the seed voxel set, i is the index of the seed voxel in the seed voxel set; If satisfied , then the candidate voxel is considered to be consistent with the seed voxel set There is consistency between them in the spectral space, allowing them to be joined and the seed voxel set to be updated.

5. The method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data according to claim 4, characterized in that: Candidate voxels are added to the seed voxel set Another constraint is that the spatial connectivity satisfies the following conditions: ； in, is the sum of the volumes of candidate voxels and interstitial voxels in the tth iteration, where the interstitial voxel refers to the spatial position between the candidate voxel and the seed voxel set The voxel units between them but not yet occupied by any voxel, is the volume of the seed voxel set, for Endosome prime number, is the voxel volume, are the coordinates of the candidate voxel center, for The coordinates of the voxel center with the closest Euclidean distance to the candidate voxel, represents the Euclidean distance.

6. The method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data according to claim 5, characterized in that: The growth termination must simultaneously meet the following requirements: 1) the number of newly added voxels in three consecutive iterations does not exceed 0.1% of the total number of seed voxels in the previous iteration; and 2) the spectral information entropy converges, specifically: ； in, represents the Shannon entropy of the reflectivity of the seed voxel set in the 680-750 nm band at the t-th iteration, is the reflectivity histogram partition index, is the number of reflectance histogram partitions, is the probability that the reflectivity value falls into the kth interval at the tth iteration; When there are no new voxels in three consecutive iterations and the number of new voxels in three consecutive iterations does not exceed 0.1% of the total number of seed voxel sets in the previous iteration, the growth is terminated and the final seed voxel set is output. .

7. The method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data according to claim 6, characterized in that: The final seed voxel set Each voxel unit in is resolved into a three-dimensional space entity, whose properties include: geometric properties and spectral properties: The geometric attributes are voxel center coordinates and preset voxel sizes , the spectral attribute is the hyperspectral reflectance vector mapped by the voxel; a three-dimensional canopy model integrating structure and spectrum is constructed based on the geometric attributes and spectral attributes, and the model is stored in the form of voxel units, each unit containing coordinates and reflectance data; The logic for automatically labeling the physiological status of vegetation based on its spectral characteristics is as follows: Calculate the chlorophyll sensitivity index of voxel cells in the 3D canopy model at 680 nm and 750 nm: ,when , the voxel is judged to be in a stress state, otherwise it is marked as healthy; are the reflectance of the voxel at 750nm and 680nm, respectively; For voxels judged to be stressed, the reflectance slope of the blue band is further calculated. ,like , it is identified as a bacterial disease; the reflectivity slope of the red band ,like , it is determined to be a fungal disease; if neither of the two conditions is met, it is marked as an unknown stress; are the reflectance of the voxel in the 490nm, 530nm, and 700nm bands, respectively; The pathological label of each voxel and its spatial coordinates are used to output a three-dimensional canopy pathology spatial distribution map for vegetation disease location and health status assessment.

8. A fruit tree canopy 3D model reconstruction system based on multimodal data, characterized by: The system is used to execute the method for reconstructing a three-dimensional model of a fruit tree canopy based on multimodal data according to any one of claims 1 to 7, comprising: A data acquisition module is used to obtain the three-dimensional structural information and spectral information of the fruit tree canopy in the monitored area. The three-dimensional structural information is point cloud data obtained by laser radar scanning, and the spectral information is a hyperspectral image corresponding to the canopy. The point cloud and image are unified into a world coordinate system through ground control points to establish a precise mapping relationship between point cloud coordinates and hyperspectral pixels. The set determination module is used to divide the bounding box range of the registered point cloud data into equal intervals at a preset voxel scale, generate voxel units covering the entire canopy space to construct the current voxel set, count the number of point clouds contained in each voxel unit, and determine whether it is occupied according to a preset occupancy threshold. The voxel units that meet the conditions are formed into a seed voxel set; An iterative update module is used to use the seed voxel set as the starting point for growth and iteratively expand using a structural-spectral collaborative growth strategy. In each round of iteration, a candidate voxel that is directly spatially adjacent to any voxel in the seed voxel set is selected from the current voxel set. For each candidate voxel, the spectral feature similarity between it and the spatially adjacent voxels in the seed voxel set is calculated. Only when the spectral feature similarity of the candidate voxel meets a preset consistency threshold is it included in the seed voxel set. If no new voxel is included in the current voxel set for three consecutive iterations, the growth process is determined to be complete and the final seed voxel set is output. The model analysis module is used to convert the final seed voxel set into a three-dimensional canopy model that integrates spatial structure and spectral information. In this model, each voxel unit contains its three-dimensional spatial coordinates and its corresponding spectral reflectance data. Based on the spectral reflectance characteristics of the voxel unit, the physiological state of the vegetation it represents is automatically labeled to complete the three-dimensional reconstruction and physiological state analysis of the canopy.

Images