A method for calculating canopy volume

By combining the improved α-shape algorithm and the KNN algorithm, the slice thickness and α value are dynamically adjusted, which solves the problem of inaccurate calculation of citrus canopy volume in the existing technology, realizes more accurate canopy volume measurement, and supports dynamic monitoring and management of fruit tree growth.

CN118429412BActive Publication Date: 2024-11-15CITRUS RES INST SOUTHWEST UNIV
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410583001.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-11
Publication Date
2024-11-15
Estimated Expiration
2044-05-11

AI Technical Summary

Technical Problem

Existing methods for calculating canopy volume cannot accurately reflect the actual distribution shape of citrus trees, resulting in a large deviation between the measurement results and the actual volume, which affects the monitoring and optimization of fruit tree growth dynamics and management.

Method used

An improved α-shape algorithm combined with the KNN algorithm is used to iteratively find the threshold αb for each point cloud slice, calculate the area and density difference of the outer contour of the slice, dynamically merge the slices, adjust the α value using the average point spacing of the point cloud, and reconstruct the canopy volume.

Benefits of technology

It can more accurately calculate the actual distribution shape of citrus tree canopies, eliminate cavities and gaps, provide a more realistic measurement of canopy volume, and support the dynamic monitoring and management of fruit tree growth.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118429412B_ABST
    Figure CN118429412B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of point cloud processing, and specifically discloses a canopy volume calculation method, comprising the following steps: S1, inputting point cloud data; S2, slicing the crown point cloud data indirectly; S3, obtaining the area of ​​the outer contour polygon of the crown slice; S4, calculating the area change percentage of each adjacent slice; S5, calculating the overall density of each slice, and calculating the difference between the slice density and the average density; S6, if the area change percentage R of two adjacent slices is greater than the average area percentage of the adjacent slices, or the density difference between the previous slice and the current slice is less than the overall average difference, the slice is marked as 1, otherwise it is marked as 0; S7, if the marks of two adjacent slices are the same, the two slices are merged; S8, reconstructing the slice point cloud, and calculating the volume of the reconstructed area; S9, adding the volumes of all slices to obtain the total volume of the crown. The technical solution of the present invention can accurately calculate the actual distribution shape of the crown of citrus trees.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of point cloud processing, and in particular to a canopy volume calculation method. Background Art

[0002] The crown of a fruit tree is the main part that absorbs light for respiration and photosynthesis. It contains most branches, all leaves and fruits, and provides the nutrients and energy needed for growth. The crown of a fruit tree is directly related to the growth and productivity of the tree, and can be used as an important indicator for tree biomass monitoring, growth estimation, yield prediction, water consumption estimation, health monitoring, and long-term productivity monitoring. In addition, it can quantify the impact of pruning and evaluate tree parameters. The detection of the canopy characteristics of citrus trees is a key task in the production, breeding and management of citrus trees. It is inseparable from the fine management projects of orchards such as water and fertilizer irrigation, soil moisture detection, fruit tree breeding and output value estimation of citrus trees. At the same time, accurate measurement of canopy volume information can also provide a basis for the identification and decision-making of crop pesticide dosage, provide technical support for pruning robots, and guide managers to make correct agricultural production decisions.

[0003] At present, most agricultural researchers still use manual measurement (MM) to measure the volume of tree crowns. The volume obtained in this way is not only time-consuming and labor-intensive, but also has a large deviation from the actual volume, which is not conducive to further analysis by researchers.

[0004] LiDAR has the advantages of high accuracy and long distance, and is widely used in canopy feature extraction. At present, the reconstruction algorithms based on point cloud mainly include convex hull algorithm (CH), convex hull slicing algorithm (CHBS), voxel method (VB), platform method (PCH), alpha-shape algorithm (AS) and alpha-shape slicing algorithm (ASBS). The reconstruction effect of each algorithm is shown in the figure below. Figure 1-Figure 7 As shown, in the figure, x, y and z represent the east-west canopy distribution, north-south canopy distribution and canopy height respectively.

[0005] After reconstructing these algorithms and manual measurement methods on citrus trees, it was found that the manual measurement method regarded the tree crown as an ellipsoid, and the shape of citrus trees is often complex and irregular. This simplified model calculated more of the external space between the ellipsoid model and the canopy, and did not pay attention to the gaps and holes inside the canopy, so the volume value obtained was much larger than the actual canopy volume;

[0006] The convex hull algorithm regards the fruit tree as a convex polyhedron. Although it removes some external space of the ellipsoid model, it does not notice the gaps and holes. The convex hull slicing algorithm uses the same principle as the convex hull method. Although it removes some holes and gaps, the effect is very poor and it still overestimates the volume. The table method uses external geometric features to solve the three-dimensional canopy information, but it simplifies the crown morphology to a certain extent and ignores the details of the internal holes. The voxel method considers the cubes on the crown surface, but it is difficult to select cubes of appropriate sizes for each fruit tree with different crown widths, and the shape and layout of the cubes cannot fully adapt to the crown morphology. Although the alpha-shape algorithm can obtain relatively accurate results, the structure of citrus trees is complex, and the distribution of fruits and branches and leaves of each tree is different, so it is difficult to determine the appropriate α value for reconstruction. Although the alpha-shape slicing algorithm performs well in calculating canopy volume, the canopy shape and size of each citrus tree are different, so slicing with uniform thickness in the vertical direction has a great impact on the calculation of the canopy volume. Moreover, the alpha-shape slicing algorithm uses the same alpha value when reconstructing each slice, without taking into account the fact that the density of each part of the fruit tree is different.

[0007] In summary, the current algorithms have some problems that make it impossible to accurately calculate the canopy characteristics of citrus trees. Therefore, it is necessary to develop a calculation method that is both accurate and close to the actual distribution shape of citrus trees, which is crucial for monitoring the growth dynamics of fruit trees and optimizing orchard management. Summary of the invention

[0008] The invention provides a canopy volume calculation method, which can accurately calculate the actual distribution shape of the canopy of citrus trees.

[0009] In order to solve the above technical problems, this application provides the following technical solutions:

[0010] A canopy volume calculation method comprises the following steps:

[0011] S1, input the point cloud data of a citrus tree crown;

[0012] S2, slice the tree crown point cloud data longitudinally and indirectly;

[0013] S3, use the preset α-shape algorithm to iteratively find the threshold α of each point cloud slice b , and obtain the area of ​​the outer contour polygon of the crown slice under this threshold;

[0014] S4, calculating the percentage change R of the area of ​​each adjacent slice;

[0015] S5, using the KNN algorithm to calculate the overall density of each slice, and calculating the difference D between the slice density and the average density;

[0016] S6. If the area change percentage R of two adjacent slices is greater than the average area percentage Mean-R of the adjacent slices, or the density difference D between the previous slice and the current slice is less than the overall average difference Mean-D, then the slice is marked as 1, otherwise it is marked as 0;

[0017] S7. If the marks of two adjacent slices are the same, merge the two slices, otherwise continue to traverse the next slice until all slices are traversed;

[0018] S8, using the alpha-shape algorithm to reconstruct the slice point cloud, using the average point spacing of the current slice point cloud as the alpha value, iterating the alpha value until a complete area is generated, and then calculating the volume of the reconstructed area;

[0019] S9. Add the volumes of all slices to obtain the total crown volume.

[0020] Further, in step S3, the iteration formula is:

[0021] α i+1 =α i +Δα(i=1,2,…,n)

[0022] Among them, Δα represents the iteration step size, and i represents the number of iterations.

[0023] Furthermore, in step S3, the discretized Green's formula is used to calculate the area, and the formula is as follows:

[0024]

[0025] Among them, S is the area of ​​the current crown slice, k is the total number of boundary points of the current crown slice, (x i ,y i ) is the two-dimensional projection coordinate of point i; (x i +1,y i +1) is the two-dimensional projection coordinate of point i+1.

[0026] Furthermore, in step S4, the area change percentage R of each adjacent slice is calculated as follows:

[0027]

[0028] in, Represents the area ratio of adjacent slices.

[0029] Furthermore, in step S5, the KNN algorithm is used to calculate the overall density of each slice, and the calculation formula is:

[0030]

[0031] density i =mean(localDensity)

[0032] Among them, k is the number of neighbors, Data i is the data point of the ith segment; idx i is the neighbor index of the i-th slice; represents the distance between the i-th slice and its k nearest neighbors.

[0033] Further, in step S5, the difference D between the slice density and the average density is calculated using the following formula:

[0034] D i =|density i -mean(density)|

[0035] Among them, mean() is the average function.

[0036] Furthermore, in step S2, the slices are divided into 50 equally spaced slices.

[0037] This scheme dynamically merges slices by considering the area change ratio and density difference of adjacent slices: the change in slice area often reflects the actual shape and growth characteristics of the crown. If the area change ratio of a slice is greater than the average area change ratio, it means that the crown growth in this part is relatively vigorous. These areas may be newly grown branches or leaves, or rapidly growing fruits. If the density difference of a slice is less than the average density difference, it means that the health of the crown in this part is relatively stable and the growth is relatively balanced. According to the density and shape of the point cloud data. When processing point cloud data, this scheme will dynamically modify the slice thickness and α value to make it closer to the actual crown, thereby effectively eliminating holes and gaps and better describing the shape and density distribution of the crown. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is the rendering of the tree crown reconstruction using the manual measurement method;

[0039] Figure 2 This is the tree crown reconstruction effect using the convex hull algorithm;

[0040] Figure 3 This is the tree crown reconstruction effect using the convex hull slicing algorithm;

[0041] Figure 4 This is the rendering of the tree crown reconstruction using the voxel method;

[0042] Figure 5 This is the tree crown reconstruction effect using the alpha-shape algorithm;

[0043] Figure 6 This is the tree crown reconstruction effect using the alpha-shape slicing algorithm;

[0044] Figure 7 This is the tree crown reconstruction effect using the platform algorithm;

[0045] Figure 8 This is a flow chart of a canopy volume calculation method according to Embodiment 1;

[0046] Fig. 9 This is a tree crown reconstruction effect diagram of a canopy volume calculation method in Example 1;

[0047] Fig.10 This is a flow chart of orchard 3D map segmentation in Example 2;

[0048] Fig.11 This is a side view of the tree crown reconstruction in a canopy volume calculation method in Example 2;

[0049] Fig.12 Schematic diagram of volume comparison of different calculation methods in Example 3. DETAILED DESCRIPTION

[0050] The following is further described in detail through specific implementation methods:

[0051] Embodiment 1

[0052] like Figure 8 As shown, a canopy volume calculation method of this embodiment includes the following steps:

[0053] S1, input the point cloud data of a citrus tree crown;

[0054] S2. Slice the tree crown point cloud data longitudinally and evenly. In this embodiment, the data is divided into 50 equally spaced slices.

[0055] S3. When using the α-shape algorithm to construct the contour, there is a threshold α b , when α≥α b When , the boundary polygon formed by it can contain all the points in the point set P.

[0056] Use the preset α-shape algorithm to iteratively find the threshold α for each point cloud slice b , and the area of ​​the outer contour polygon of the crown slice is obtained under this threshold. The iterative formula is as follows:

[0057] α i+1 =α i +Aα(i=1,2,…,n)

[0058] Wherein, Δα represents the iteration step, i represents the number of iterations, and in this algorithm, the initial α is set to 0.05, and i is set to 0.1. In this embodiment, the preset α-shape algorithm is the "Improved α-shape Algorithm", which is an algorithm for constructing a two-dimensional plane published by Cheng Gang in the Journal of Agricultural Machinery.

[0059] The discretized Green's formula is used to calculate the area, the formula is as follows:

[0060]

[0061] Among them, S is the area of ​​the current crown slice, k is the total number of boundary points of the current crown slice, (x i ,y i ) is the two-dimensional projection coordinate of point i; (x i +1,y i +1) is the two-dimensional projection coordinate of point i+1.

[0062] S4. Calculate the percentage change R of the area of ​​each adjacent slice using the following formula:

[0063]

[0064] in, Represents the area ratio of adjacent slices.

[0065] S5. Use the KNN algorithm (K-nearest neighbor algorithm) to calculate the overall density of each slice. The calculation formula is:

[0066]

[0067] density i =mean(localDensity)

[0068] Where k is the number of nearest neighbors, and this embodiment uses 10; Data i is the data point of the ith segment; idx i is the neighbor index of the i-th slice; represents the distance between the i-th slice and its k nearest neighbors (excluding itself);

[0069] Then calculate the difference D between the slice density and the average density, and the calculation formula is:

[0070] D i =|density i -mean(density)|

[0071] Among them, mean() is the average function.

[0072] S6. If the area change percentage R of two adjacent slices is greater than the average area percentage Mean-R of adjacent slices, or the difference D between the density of two adjacent slices and the average density is less than the overall average difference Mean-D, then the slice is marked as 1, otherwise it is marked as 0. The area change percentage R of all two adjacent slices is added together, and then the average is calculated to obtain the average area percentage Mean-R of the adjacent slices; the area change percentage R of all two adjacent slices is added together, and then the average is calculated to obtain the overall average difference Mean-D.

[0073] S7. If the marks of two adjacent slices are the same, merge the two slices, otherwise continue to traverse the next slice until all slices are traversed.

[0074] S8, such as Fig. 9 As shown, the alpha-shape algorithm is used to reconstruct the slice point cloud, the average point spacing of the current slice point cloud is used as the alpha value, the alpha value is iterated until a complete area is generated, and then the volume of the reconstructed area is calculated.

[0075] This example uses a custom function PointsAverageSpacing(Data) to calculate the average distance between points in the point cloud. In this function, a kd-tree structure is used to quickly find the nearest neighbor of each point and calculate the sum of the distances between all points. The average distance value is obtained by dividing the total distance by the number of points in the point cloud. After obtaining the average distance, the value is used to construct the alpha-shape.

[0076] Alpha-shape is a geometric shape description method that defines the maximum allowable distance to form a continuous surface in a point cloud by controlling a parameter. This embodiment uses the alpha-shape function in MATLAB to construct alpha-shape, where the parameter distance is initialized to the average point spacing of the current slice point cloud. Then, the distance is iteratively adjusted until the number of regions in alpha-shape is 1, and the iterative formula is distance = distance + the average point spacing of the current slice point cloud. This ensures the integrity and connectivity of the shape. Finally, the volume of the reconstructed area is calculated.

[0077] S9. Add the volumes of all slices to obtain the total volume of the crown and output the result.

[0078] The branches of naturally round-headed citrus trees will droop naturally when they are in fruit-bearing state. The fruits are concentrated in the middle and lower parts of the crown. There are usually more fruits in the lower part, and the branches and leaves are concentrated in the middle and upper parts. For example, a 3-year-old citrus tree has most of its top branches that are newly grown and have no fruit, and they are relatively sparse.

[0079] The method of this embodiment dynamically merges slices by considering the area change ratio and density difference of adjacent slices: the change in slice area often reflects the actual shape and growth characteristics of the crown. If the area change ratio of a slice is greater than the average area change ratio, it means that the crown growth of this part is relatively vigorous. These areas may be newly grown branches or leaves, or rapidly growing fruits; if the density difference of a slice is less than the average density difference, it means that the health of the crown of this part is relatively stable and the growth is relatively balanced; according to the density and shape of the point cloud data, the method of this embodiment will dynamically modify the slice thickness and α value when processing the point cloud data to make it closer to the actual crown situation, thereby effectively eliminating holes and gaps and better describing the shape and density distribution of the crown.

[0080] Embodiment 2

[0081] The difference between this embodiment and the first embodiment is that the steps of the method of this embodiment are implemented by writing in MATLAB R2023a software.

[0082] The user inputs a single citrus tree point cloud file in the format of .txt or .pcd file;

[0083] For a single scan of a citrus tree, the point cloud can be directly used as input data after cropping; for the scan of the entire orchard, Fig.10 The method shown segments the orchard 3D map into individual citrus trees.

[0084] When the user has a point cloud of a single citrus tree, or the user uses a folder containing multiple single citrus tree point cloud files as input for batch processing, the algorithm will return to the user an Excel spreadsheet containing the file name, run time, and volume value to display the calculated data. The specific development process is as follows:

[0085] Specify the input citrus tree point cloud file path, create an Excel spreadsheet and specify the output save path; start looping through each point cloud file;

[0086] The point cloud is initially divided into 50 slices with equal spacing;

[0087] The improved α-shape algorithm is used to calculate the contour of each slice, and then the discretized Green's formula is used to calculate the area of ​​each slice. The KNN algorithm is used to calculate the density of each slice.

[0088] The percentage change R of the area of ​​each adjacent slice and the difference densityDifference(D) between the slice density and the average density are calculated.

[0089] Loop through each slice. If either (R > averageR) or (densityDifference(1:end - 1) < meanDensityDifference) is satisfied, mark the slice as 1; otherwise, mark it as 0. Here, averageR refers to the average percentage change in area between adjacent slices, and meanDensityDifference refers to the difference between the average slice density and the average density.

[0090] Perform classification and merging: When the marks of two adjacent slices are the same, merge these two slices, and then continue to loop through the marks of the next adjacent two slices; if the marks are different, directly continue to loop through the marks of the next adjacent two slices. Repeat this step until all slices have been looped through.

[0091] Loop through all the merged slices, construct a kdtree to query the average point spacing of the current slice's point cloud as the α value for three-dimensional reconstruction of the slice using alpha-shape until a complete shape is reconstructed; otherwise, continuously iterate α. The iteration formula is: α = α + the average point spacing of the current slice's point cloud. That is, α = α + the average point spacing of the current slice's point cloud will be executed once per loop until a complete shape is reconstructed.

[0092] After the reconstruction is completed, calculate the volume of the reconstructed model of this slice.

[0093] After reconstructing all slices, accumulate the volume values sumV of all slices as the volume of the entire citrus tree, and then append the file name, running time, and volume value of this tree to an Excel table. After the operation ends, continue to loop through the next citrus tree until all citrus trees in the folder have been calculated.

[0094] As shown in the side reconstruction effect Fig.11 As shown, where α refers to the α value used for alpha-shape reconstruction of the current slice, V refers to the volume value of the current slice, and Sum(V) refers to the volume value of this citrus tree. The dynamic slicing in this embodiment more comprehensively considers the area and density characteristics of the slices, making it easier to identify the dense areas of leaves and branches, the concentrated areas of fruits, and the sparse density areas of the roots and tree tops. The optimal α values for these areas are automatically determined by continuously iterating the point spacing α value: for the dense areas of branches and fruits, use a smaller α value to improve the reconstruction (0.0244 - 0.0775), which can better capture the detailed shapes and structures of the branches, leaves, and fruits, making the model more realistic in structure; for the root and tree top slices with few branches and fruits, use a slightly larger α value for loose reconstruction (0.1161 - 0.3306), which can capture the overall shape and structure, making the overall contour of the tree crown smoother and more stable.

[0095] Embodiment 3

[0096] In this embodiment, the existing typical canopy reconstruction algorithm and manual measurement method are compared and analyzed with this solution.

[0097] Fig.12 The comparison chart of the volume values ​​of 160 citrus trees measured by the MM method and 7 point cloud-based algorithms is shown, where Number refers to the number of point cloud trees, in units of trees; Crown Volumes refers to the volume value of the tree reconstructed under the specified algorithm, in units of cubic meters. Table 1 shows the R of each algorithm. 2 Compare the analysis results with the regression coefficients.

[0098] Table 1

[0099]

[0100] Based on these comparative experiments, we further analyzed: Fig.12 It can be seen that the relationship between the crown volumes measured by these eight methods is:

[0101] V MM >V CH >V CHBS >V VB >V PCH >V AS >V ASBS >V DR

[0102] The canopy volume calculated by MM is larger than the results calculated by all point cloud-based algorithms. The canopy volume calculated by MM is larger than the canopy volume calculated by all point cloud-based algorithms. This is because the MM method treats the tree crown as an ellipsoid. The shape of blood orange trees is often complex and irregular. This simplified model will not only take into account the holes and gaps inside the canopy, but also the external space between the canopy and the ellipsoid model. Therefore, the volume value calculated using the MM method is much larger than the actual volume of the tree crown.

[0103] The CH algorithm has the largest calculation result among all point cloud-based algorithms. This is because the CH algorithm treats fruit trees as convex polyhedra, taking into account the holes, cracks, and depressions inside the crown, and the convex hull it generates is larger than the crown outline. Although the CHBS algorithm removes some of the holes and gaps caused by the CH algorithm, it still has the same problems as the CH algorithm, so the results still overestimate the canopy volume. Although the VB algorithm takes into account the cube of the crown surface, the shape and layout of the cube cannot fully adapt to the crown shape, and it is difficult to choose the appropriate cube size for each fruit tree with different crown widths, so its final volume The PCH algorithm can solve the three-dimensional crown information based on the full use of the external geometric features of the crown in principle. It is a simple, fast and effective method for measuring the crown volume, but it also simplifies the crown morphology to a certain extent and ignores details such as the holes inside the crown, which leads to an increase in the error of volume estimation. The AS algorithm can accurately extract the boundaries of the tree crown point cloud when α is reasonably set. ASBS can successfully remove holes and gaps in the tree crown to a certain extent, so the volume calculated by the AS algorithm is smaller. The volume calculated by the ASBS algorithm is the smallest

[0104] According to the analysis in Table 1, V MM The increase in the difference in regression coefficients with other volume measurements suggests that V MM The difference between the volume measurements and those of other methods increases; V MM Linear fit R of volume measurements with other methods 2 As the volume difference increases, it decreases. This is because the ellipsoid model takes into account both the holes and gaps inside the crown and the external space between the crown and the ellipsoid model, resulting in a significant overestimation of the crown volume. The CH algorithm removes part of the external space between the crown and the ellipsoid model, and the CHBS algorithm subsequently removes some of the holes and gaps. Subsequent algorithms have gradually removed more holes and gaps to a certain extent to optimize the extraction of crown boundary details. Therefore, the regression coefficient and R discussed previously 2 The change is the result of the increasing difference in the calculated volumes between the MM and point cloud-based algorithms. The CH algorithm has a lower R2 compared to the other algorithms, which indicates that its predictive ability is relatively weak;

[0105] Compared with other algorithms, the regression equation intercepts of CH and CHBS algorithms are larger, which means that CH and CHBS algorithms overestimate the volume of tree crowns than other algorithms. VB algorithm is a voxel-based algorithm and does not calculate the volume of cubes without point clouds. CHBS V VBThe reason for the large volume is that the PCH algorithm fits the data well, the slope of the fitting equation is moderate, the intercept is small, and it has a significant correlation with the AS method, indicating that its estimated volume is relatively accurate. There are great differences between the CH algorithm and the AS algorithm in terms of reconstruction effect and accuracy. The CH algorithm simplifies the crown shape and ignores internal details and holes. The AS algorithm can generate crown shapes of different fineness according to the parameter α, is more sensitive to external details, and can accurately reconstruct the crown shape. The intercepts of the AS and ASBS algorithms are smaller than those of other algorithms (between 0.013-0.91), and only the VB algorithm has the largest intercept (0.161 and 0.183), which indicates that the VB algorithm cannot well represent the internal structure and contour details of the crown using small cubes, and overestimates the volume of the crown. The R2 values ​​of AS and ASBS are both 0.999, and the intercepts and slopes are very small, indicating that the two are strongly correlated, and because the volume value obtained by ASBS is the smallest, its slope (1.068-2.690) is greater than that of all other algorithms.

[0106] Although ASBS performs well in determining the crown volume, due to the different crown shapes and sizes of each citrus tree, vertically slicing the crown with a consistent thickness has the following disadvantages: it cannot adapt to the irregular shape of the crown, especially the edge or top of the crown, which may result in too many or too few slices; slicing the center of the crown will produce too many data points; it cannot accurately reflect the complex structure and cavity inside the crown; it may ignore some details and features of the crown, such as branching and leaf density; therefore, it may relatively overestimate the crown volume. In addition, the height of the experimental fruit trees ranged from 0.83 meters to 2.616 meters, and the number of slices per tree was also different, generally between 10 and 30 slices after dynamic slicing, while citrus fruit trees have dense branches and leaves, complex morphology, and different branch structures and distribution densities. If the fixed value of α = 0.05 is still used, the volume will be underestimated due to the smooth transition of the surface at some locations, the reconstruction results will lack details, or information will be lost due to excessive denoising. Using the average point spacing of the iterative point cloud as α can better reflect the distribution of point cloud data, adapt to point clouds of different densities, retain the characteristics of the original point cloud data, and make the reconstructed volume more adaptable, accurate, and reliable.

[0107] Since the true volume of the tree crown cannot be obtained, the ASBS algorithm with better performance among the six algorithms, the PCH algorithm with a wider range of applications, and the CHBS algorithm that also uses the slicing method are used to further compare with the algorithm proposed in this solution. This embodiment uses these four methods to calculate the average number of slices per tree, the average α value of each slice, the average volume consistency value, and the average volume per slice of 160 blood orange trees. The volume consistency of the fruit tree is calculated by dividing the standard deviation of the slice volume by the average slice volume, and the calculation formula is:

[0108]

[0109] n represents the number of slices, V m Indicates the current slice volume, V ave represents the average slice volume.

[0110] Table 2

[0111]

[0112] As shown in Table 2, CHBS and ASBS used a fixed number of slices 10 on each tree and reconstructed each slice in the same way, resulting in extremely consistent volume consistency (0.556 and 0.665). However, the growth pattern of citrus trees is inherently uneven, with large differences in density and volume between parts (such as the root, middle, and top of the crown), which CHBS and ASBS did not take into account when stratifying, resulting in the reconstructed models being too consistent and stable to accurately represent the growth of citrus trees. Since PCH divides the circular platform at a height of 0.4 meters, it dynamically stratifies the citrus trees to a certain extent. Specifically, it divides the top of each tree into cones and platform-segments the remaining area according to the crown height. Therefore, its volume consistency (0.710) is closer to the citrus distribution pattern than ASBS and CHBS, but because it still relies on the basic geometric model to calculate the volume, the resulting volume will be too large and cannot fully eliminate holes and gaps. This scheme dynamically stratifies each fruit tree and uses different a values ​​for different parts of the fruit tree. According to the results (1.532), this scheme better represents the growth characteristics and morphological features of various parts of the citrus tree (such as crown density and branch / fruit distribution), the reconstruction results are also more continuous and smoother, and the alpha shape is more consistent with the original data, thus obtaining a more accurate crown volume.

[0113] Based on the density and shape of the point cloud data, this solution will dynamically modify the slice thickness and α value when processing the point cloud data to make it closer to the actual crown situation, thereby effectively eliminating loopholes and gaps. Fig.12As shown in the figure, the volume values ​​measured by this scheme on 160 experimental trees tend to be stable, and the average volume value is lower than that of all six algorithms (VDR = 0.84). Table 1 shows that the slope and intercept of this scheme are more significant in the regression equation with other algorithms (the intercept is more significant than the influence of PCH, AS and ASBS studied in the previous section). In addition, the previous analysis shows that for algorithms with higher accuracy, the fitting R2 value with MM is lower. This is because the ellipsoid model is very different from the actual volume of the crown. It ignores that each tree is unique, and the distribution of fruits, density of branches and leaves, and gaps and holes between crowns are all different (the same is true for the CH algorithm). The R2 of this scheme and other algorithms is about 0.96, indicating that the volume obtained by the algorithm is stable and the fit is satisfactory. The reason why R2 is lower than ASBS and other algorithms is that this scheme starts from the internal structure of the crown, uses the correlation between density and area to dynamically slice, and iterates to find the optimal a value for reconstruction based on the distribution of branches, leaves and fruits in each slice. Traditional algorithms all use a simple and uniform slice and rebuild approach, rather than dynamically correcting for changes in the underlying structure of the canopy as this scheme does, so the R2 is higher in many other algorithms, while this scheme is lower. This is critical for analyzing the growth patterns of citrus trees and predicting future fruit supply.

[0114] In general, this solution not only solves the problem that the thickness and number of crown point cloud slices are difficult to adapt to the shape and size of the crown itself when using ASBS and CHBS algorithms, but also solves the problem that the PCH algorithm cannot capture detailed features, resulting in inaccurate and incomplete reconstructed canopy models. It has better adaptive ability, more precise volume calculation, better ability to remove noise and outliers, and adaptability that is closer to the actual shape of trees. This is of great significance for the later growth pattern analysis of citrus trees and fruit yield estimation.

[0115] The above are only embodiments of the present invention. The invention is not limited to the fields involved in this implementation case. The common sense such as the known specific structures and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field to which the invention belongs before the application date or the priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The protection scope required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to explain the content of the claims.

Claims

1. A canopy volume calculation method, characterized in that: The steps include: S1, input the point cloud data of a citrus tree crown; S2, slice the tree crown point cloud data longitudinally and indirectly; S3. Use the preset α-shape algorithm to iteratively find the threshold αb of each point cloud slice, and obtain the area of ​​the outer contour polygon of the crown slice under this threshold; the iterative formula is: a i+1 =a i +Δα (i=1,2,…,n) Among them, Δα represents the iteration step size, and i represents the number of iterations; The discretized Green's formula is used to calculate the area, the formula is as follows: Among them, S is the area of ​​the current crown slice, k is the total number of boundary points of the current crown slice, (x i , yi ) is the two-dimensional projection coordinate of point i; (xi+1, yi+1) is the two-dimensional projection coordinate of point i+1; S4. Calculate the percentage change R of the area of ​​each adjacent slice using the following formula: in, represents the area ratio of adjacent slices; S5. Use the KNN algorithm to calculate the overall density of each slice. The calculation formula is: density i =mean(localDensity) Among them, k is the number of neighbors, Data i is the data point of the i-th segment; idx i is the neighbor index of the i-th slice; represents the distance between the i-th slice and its k nearest neighbors; Calculate the difference D between the slice density and the average density using the formula: D i =|density i -mean(density)| Among them, mean() is the average function; S6. If the area change percentage R of two adjacent slices is greater than the average area percentage Mean-R of the adjacent slices, or the density difference D between the previous slice and the current slice is less than the overall average difference Mean-D, then the slice is marked as 1, otherwise it is marked as 0; S7. If the marks of two adjacent slices are the same, merge the two slices, otherwise continue to traverse the next slice until all slices are traversed; S8, using the alpha-shape algorithm to reconstruct the slice point cloud, using the average point spacing of the current slice point cloud as the alpha value, iterating the alpha value until a complete area is generated, and then calculating the volume of the reconstructed area; S9. Add the volumes of all slices to obtain the total crown volume.

2. The canopy volume calculation method according to claim 1, characterized in that: In step S2, the image is divided into 50 slices with equal spacing.

Citation Information

Cited By

  • Method for calculating effective volume of fruit tree canopy

    CN121414818A