A method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn

Abstract

The invention discloses a method for extracting the distribution of high-throughput leaf angles and leaf inclination angles of field corn, comprising the following steps: scanning and acquiring point cloud data of field corn; performing stem-leaf separation on the point cloud data of field corn; clustering the obtained stalk points of field corn using a density-based DBSCAN spatial clustering algorithm, and extracting the stalk points of each corn plant respectively; fitting the extracted stalk points of each corn plant using a least squares algorithm to obtain a stalk point fitting line of each corn plant; and calculating a multi-layer leaf angle distribution curve and a multi-layer leaf inclination angle distribution curve of field corn. The extraction method proposed by the invention can extract the distribution of leaf angles and leaf inclination angles of field corns with different growth periods, different varieties and different planting densities, realizes the dynamic monitoring of leaf angles and leaf inclination angles of corns of different varieties at different growth periods, and provides an effective tool for obtaining phenotypic information of field corn.

Description

Technical Field

[0001] The invention relates to the technical field of high-throughput leaf angle and leaf inclination angle distribution extraction of field corn, and specifically to a high-throughput leaf angle and leaf inclination angle distribution extraction method of field corn. Background Art

[0002] The rapid population growth, the reduction of available arable land and global climate change around the world have made food security issues increasingly serious. Corn is the largest food crop in my country, accounting for about 35% of the total grain planting area and close to 40% of the total annual grain output. It plays a pivotal role in ensuring national food security. However, my country's corn yield is currently only about 60% of the world's advanced level, facing a strong practical need to use excellent varieties to select and improve corn yield. Accurate and high-throughput extraction of crop phenotypic parameters is the key to breeding. Leaf angle and leaf inclination, as important structural parameters, are of great significance to promoting corn breeding technology, cultivating high-yield and high-quality varieties, and alleviating food crises.

[0003] Leaf angle refers to the angle between the root of the leaf and the direction of the stem, and is the main trait that determines the compactness of corn. A smaller leaf angle can reduce the shading between leaves, ensuring that corn can maintain good photosynthesis even when the planting density is high, thereby achieving the goal of increasing yield.

[0004] Leaf inclination angle refers to the angle between the normal of the leaf fitting plane and the vertical direction. As an important crop structural parameter, leaf inclination angle is closely related to the spectral reflectance and radiation transfer process of the crop canopy. It is also a basic parameter of the crop three-dimensional radiation transfer model.

[0005] The existing methods for extracting corn leaf angles and leaf inclination angles mainly include two methods: skeleton extraction based on the calculation of normal vectors based on the plane fitting. The skeleton extraction method is mostly used to calculate the leaf angle, while the method based on the plane fitting to calculate the normal vector is mostly used to calculate the leaf inclination angle of vegetation with flat leaves and no curvature. At present, the extraction of corn leaf angles and leaf inclination angles mainly faces the following problems: due to the overlapping of corn leaves in the field, cross-occlusion is relatively serious. At present, the research on corn leaf angles and leaf inclination angles is mostly concentrated on the scale of a single plant, and almost no research focuses on the extraction of corn leaf angles and leaf inclination angles in the field; corn leaves are slender and curved, which makes it difficult to fit the plane to calculate the normal vector, resulting in greater difficulty in extracting the leaf inclination angle of corn. Therefore, few studies focus on the extraction of corn leaf inclination angles. Summary of the invention

[0006] The purpose of the present invention is to provide a method for extracting the leaf angle and leaf inclination angle distribution of field corn with high throughput and high precision, so as to provide a scientific method and technical support for the screening of ideal plant traits of corn, promote corn breeding technology, and cultivate high-yield and high-quality varieties.

[0007] The present invention provides a method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn, comprising the following steps:

[0008] Scan and obtain point cloud data of corn in the field;

[0009] Separate the stem and leaf of the point cloud data of field corn to obtain the stem points and leaf points of field corn;

[0010] The stalk points of field corn were thinned out, and the density-based DBSCAN spatial clustering algorithm was used to cluster the stalk points of field corn, and the stalk points of each corn plant were extracted respectively;

[0011] The least squares algorithm is used to fit the extracted stalk points of each corn plant to obtain the fitting line of the stalk points of each corn plant;

[0012] Points within a certain range from the fitting lines of each stalk point are selected from the leaf points of field corn, and the points within the range are divided into multiple layers according to their heights. The density-based DBSCAN spatial clustering algorithm is used to cluster the leaf points in each layer into multiple leaf point clusters. Planes are fitted to each leaf point cluster in each layer and the normal vectors of each leaf point cluster are calculated.

[0013] The multi-layer leaf angle distribution curve and multi-layer leaf inclination angle distribution curve of field corn are calculated according to the normal vector of the fitting plane of each leaf point cluster in each layer.

[0014] Furthermore, the method for scanning and acquiring point cloud data of field corn includes:

[0015] Arrange two or more ground-based lidar scanning stations in the corn field;

[0016] Set the scanning mode and scanning range of the ground-based laser radar;

[0017] The scanning parameters of the ground-based laser radar are set, wherein the scanning parameters of the ground-based laser radar include scanning resolution, quality, horizontal and vertical scanning angles.

[0018] Furthermore, before the point cloud data of the field corn is separated into stems and leaves, the point cloud data of the field corn is preprocessed;

[0019] The preprocessing of the point cloud data of field corn includes solving and splicing of multi-site scanning data, point cloud denoising, ground point filtering and plot segmentation;

[0020] After preprocessing, corn point cloud data of multiple plots are obtained.

[0021] Furthermore, the stem-leaf separation of the point cloud data of field corn includes the following steps:

[0022] All data points in the corn point cloud data of each plot are traversed by a cylinder to obtain the neighborhood points of each data point;

[0023] Construct a covariance matrix for the neighborhood points of each data point, use singular value decomposition to obtain the multidimensional eigenvalues ​​and corresponding eigenvectors of the neighborhood points of each data point, and judge the distribution of the neighborhood points of each data point based on the distribution of the multidimensional eigenvalues ​​of the neighborhood points of each data point;

[0024] The eigenvector corresponding to the maximum eigenvalue in the multidimensional eigenvalues ​​is the direction vector of the neighborhood point of the data point, and the angle between the direction vector of the neighborhood point of the data point and the vertical direction is calculated;

[0025] The stalk points and leaf points in the corn point cloud data of the plot are separated according to the distribution of the neighborhood points of each data point and the size of the angle between the direction vector of the neighborhood points of each data point and the vertical direction.

[0026] Furthermore, a distance-based method was used to thin out the stalk points in the corn point cloud data of each plot, so that the minimum distance between points was 0.01 m.

[0027] Furthermore, the clustering of the stalk points of field corn using the density-based DBSCAN spatial clustering algorithm includes:

[0028] Select a point p from the stalk points in the corn point cloud data of each plot;

[0029] Find nearby points N whose distance from point p is within Ep and includes Ep Eps (q), then N Eps (q) is defined as the set of points q including point p itself, expressed as:

[0030] N Eps (q)={q∈D|distance(p, q)≤Eps} (1)

[0031] Where Eps represents the neighborhood radius value of clustering time point p;

[0032] N Eps (q) indicates the nearby points within and including Eps from point p;

[0033] When N Eps If the number of (q) is ≥ MinPts, then point p and its nearby points form a stalk point cluster;

[0034] When N Eps If the number of (q) is less than MinPts, the points near point p are marked as noise points;

[0035] Where MinPts represents the threshold of the number of points in the neighborhood whose distance to point p is Eps.

[0036] Furthermore, the least squares algorithm is used to fit the extracted stalk points of each corn plant to obtain a fitting line of the stalk points of each corn plant, including:

[0037] Select any point (x0, y0, z0) on the stalk of each corn plant;

[0038] Obtain the direction vector (m, n, p) of the fitting line of each corn stalk point, and the linear equation of the fitting line of each corn stalk point is:

[0039]

[0040] Further, fitting a plane to the blade point cluster and obtaining a normal vector of the blade point cluster includes:

[0041] Using the coordinates of several three-dimensional points (x i ,y i , z i ) Establish a fitting plane equation, where i = 1, 2, 3, ...;

[0042] The constraint condition of the fitting plane equation is a 2 +b 2 +c 2 =1, and the sum of the distances from the fitting plane to all points in the blade point cluster is the smallest, where the expression of the fitting plane equation is:

[0043] ax+by+cz=d (3)

[0044] Let the average coordinates of all points in each leaf point cluster be but:

[0045]

[0046] Subtracting equation (3) from equation (4), we get:

[0047]

[0048] Assumption Matrix Column Matrix Then equation (5) is equivalent to AX = 0

[0049] Perform singular value decomposition on the matrix A, and the eigenvector corresponding to the minimum singular value of the matrix A is the normal vector of the fitting plane of the leaf point cluster.

[0050] Further, among the leaf points in the corn point cloud data of each of the plots, points within the ranges of [0.05m, 0.10m], [0.05m, 0.15m] and [0.10m, 0.15m] from the stalk fitting line are selected respectively, and the points within each range are divided into 5 layers according to the height;

[0051] The density-based DBSCAN algorithm is used to cluster the leaf points in each layer into multiple leaf point clusters, and planes are fitted to each leaf point cluster in each layer and the normal vectors of each leaf point cluster are calculated.

[0052] According to the definition of leaf angle, the leaf angle of each leaf point cluster in each layer is obtained by using the normal vector of the fitting plane of each leaf point cluster in each layer. The average value of multiple leaf angles in each layer is calculated to obtain the five-layer leaf angle distribution curve of field corn.

[0053] Further, among the leaf points in the corn point cloud data of each of the plots, points at a distance of [0.15m, 0.20m], [0.20m, 0.25m], [0.15m, 0.25m] and [0.20m, 0.30m] from the stalk fitting line are selected respectively, and the points in each range are divided into 5 layers according to the height;

[0054] The density-based DBSCAN algorithm is used to cluster the leaf points in each layer into multiple leaf point clusters, and planes are fitted to each leaf point cluster in each layer and the normal vectors of each leaf point cluster are calculated.

[0055] According to the definition of leaf inclination angle, the leaf inclination angle of each leaf point cluster in each layer is obtained by using the normal vector of the fitting plane of each leaf point cluster in each layer. The average value of multiple leaf inclination angles in each layer is calculated to obtain the five-layer leaf inclination angle distribution curve of field corn.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The high-throughput leaf angle and leaf inclination angle distribution extraction method for field corn based on ground-based lidar proposed in the present invention can extract the leaf angle and leaf inclination angle distribution of field corn in different growth stages, different varieties and different planting densities, and realize the dynamic monitoring of leaf angles and leaf inclination angles of various varieties of corn in different growth stages, providing an effective tool for breeding experts to obtain field corn phenotypic information at low cost and high efficiency, further achieving the purpose of accelerating the breeding process, increasing corn yield and alleviating the food crisis. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0059] Figure 1 It is a flow chart of a method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn proposed by the present invention;

[0060] Figure 2 This is the result of separation of stems and leaves of corn in five plots in the field;

[0061] Figure 3 It is the result of DBSCAN clustering based on the stem points obtained after separation of stems and leaves;

[0062] Figure 4 is the extracted leaf points within a certain range from the stem fitting line;

[0063] Figure 5 It is the process and result of extracting the leaf angle and leaf inclination angle distribution curve. DETAILED DESCRIPTION

[0064] The technical solutions in the embodiments of the present invention are clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention, but it should be understood that the protection scope of the present invention is not limited by the specific implementation methods.

[0065] Example

[0066] like Figure 1-5 As shown, a method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn comprises the following steps:

[0067] Step 1: Scan and obtain point cloud data of corn in the field.

[0068] Since the ground-based laser radar uses a fixed scanning method and the field corn is severely blocked, a single-site scan can only obtain corn point cloud data for a local area. Therefore, at least two sites need to be deployed in a community to obtain a relatively complete point cloud. In addition, when using ground-based laser radar to obtain corn point clouds, a large scanning range will result in a large amount of data, thereby reducing the data processing speed. Due to the high similarity of field corn, a small scanning distance will make it difficult to stitch multiple sites.

[0069] Therefore, when scanning and acquiring point cloud data of corn in the field:

[0070] Set a reasonable scanning mode and scanning range, and obtain field corn point cloud data by setting the scanning parameters of the ground-based laser radar. The scanning parameters of the ground-based laser radar include block scanning resolution, quality, horizontal and vertical scanning angles, etc.

[0071] Step 2: Preprocess the point cloud data of field corn.

[0072] The point cloud data of field corn were preprocessed, including multi-site data stitching, point cloud denoising, ground point filtering and plot segmentation.

[0073] Step 3: Separate the pre-processed field corn point cloud data into stems and leaves to obtain stem points and leaf points, including the following steps:

[0074] A cylinder is used to traverse all the data points in the pre-processed corn point cloud data of each plot to obtain the neighborhood points of each data point;

[0075] Construct a covariance matrix for the neighborhood points of each data point, use singular value decomposition to obtain the multidimensional eigenvalues ​​and corresponding eigenvectors of the neighborhood points of each data point, and judge the distribution of the neighborhood points of each data point based on the distribution of the multidimensional eigenvalues ​​of the neighborhood points of each data point;

[0076] The eigenvector corresponding to the maximum eigenvalue in the multidimensional eigenvalue is the direction vector of the neighborhood point of the data point, and the angle between the direction vector of the neighborhood point of the data point and the vertical direction is calculated;

[0077] According to the distribution of the neighborhood points of each data point and the angle between the direction vector of the neighborhood points of each data point and the vertical direction, the stem points and leaf points in the cell point cloud data are separated. The specific judgment method is as follows:

[0078] When the neighborhood points of a data point are linearly distributed and the angle between the direction vector of the neighborhood points of the data point and the vertical direction is small or close to zero, the data point is a stem point;

[0079] When the neighborhood points of a data point are distributed in a planar manner and the angle between the direction vector of the neighborhood points of the data point and the vertical direction is large, the data point is a leaf point.

[0080] Step 4: Extraction of corn leaf angle and leaf inclination distribution.

[0081] Step 4.1: Thin out the stem points.

[0082] Since the point cloud density obtained by the ground-based lidar is high, resulting in a slow data processing speed, the distance-based method is first used to thin the stalk points in the corn point cloud data of each plot so that the minimum distance between points is 0.01m.

[0083] Step 4.2: Cluster the corn stalk points in the field using the density-based DBSCAN spatial clustering algorithm with noise, and extract the stalk points of each corn plant, including:

[0084] Select any point p from the corn stalk point cloud data of each plot;

[0085] Find nearby points N whose distance from point p is within and including Eps Eps (q), then NEps (q) is defined as the set of points q including point p itself, expressed as:

[0086] N Eps (q)={q∈D|distance(p, q)≤Eps} (1)

[0087] Where Eps represents the neighborhood radius value of clustering time point p;

[0088] N Eps (q) indicates the nearby points within and including Eps from point p;

[0089] When N Eps If the number of (q) is ≥ MinPts, then point p and its nearby points form a stalk point cluster;

[0090] When N Eps If the number of (q) is less than MinPts, the points near point p are marked as noise points;

[0091] Where MinPts represents the threshold of the number of points in the neighborhood whose distance to point p is Eps.

[0092] Step 4.3: Use the least squares algorithm to fit the extracted stalk points of each corn plant to obtain the stalk point fitting line of each corn plant, including:

[0093] Select any point (x0, y0, z0) on the stalk of each corn plant;

[0094] Obtain the direction vector (m, n, p) of the fitting line of each corn stalk point, and the linear equation of the fitting line of each corn stalk point is:

[0095]

[0096] Step 4.4: Select points within a certain range from the fitting lines of each stalk point from the leaf points of the field corn, and divide the points within the range into multiple layers according to their heights. Use the density-based DBSCAN spatial clustering algorithm to cluster the leaf points in each layer into multiple leaf point clusters. Fit a plane to each leaf point cluster in each layer and calculate the normal vector of each leaf point cluster.

[0097] The multi-layer leaf angle distribution curve and multi-layer leaf inclination angle distribution curve of field corn are calculated according to the normal vector of the fitting plane of each leaf point cluster in each layer.

[0098] Step 4.4.1: Fit a plane to the blade point cluster and find the normal vector of the blade point cluster, including:

[0099] Using the coordinates of several three-dimensional points (x i ,y i, z i ) Establish a fitting plane equation, where i = 1, 2, 3, ...;

[0100] The constraint condition for fitting the plane equation is a 2 +b 2 +c 2 =1, and the sum of the distances from the fitting plane to all points in the blade point cluster is the smallest, where the expression of the fitting plane equation is:

[0101] ax+by+cz=d (3)

[0102] Let the average coordinates of all points in each leaf point cluster be but:

[0103]

[0104] Subtracting equation (3) from equation (4), we get:

[0105]

[0106] Assumption Matrix Column Matrix Then equation (5) is equivalent to AX = 0

[0107] Perform singular value decomposition on the matrix A, and the eigenvector corresponding to the minimum singular value of the matrix A is the normal vector of the fitting plane of the leaf point cluster.

[0108] Step 4.4.2: From the leaf points in the corn point cloud data of each plot, select points within the ranges of [0.05m, 0.10m], [0.05m, 0.15m] and [0.10m, 0.15m] from the stalk fitting line, and divide the points in each range into 5 layers according to their height;

[0109] The density-based DBSCAN algorithm is used to cluster the leaf points in each layer into multiple leaf point clusters, and planes are fitted to each leaf point cluster in each layer and the normal vectors of each leaf point cluster are calculated.

[0110] According to the definition of leaf angle, the leaf angle of each leaf point cluster in each layer is obtained by using the normal vector of the fitting plane of each leaf point cluster in each layer. The average value of multiple leaf angles in each layer is calculated to obtain the five-layer leaf angle distribution curve of field corn.

[0111] Step 4.4.3: Due to the cross-occlusion of corn leaves in the field, according to statistics, the distance from the middle of each leaf to the stem is about 0.1m-0.3m. Therefore, among the separated leaf points, in the leaf points in the corn point cloud data of each plot, the points at the distance from the stem fitting line [0.15m, 0.20m], [0.20m, 0.25m], [0.15m, 0.25m] and [0.20m, 0.30m] are selected to calculate the leaf inclination angle, and the points in each range are divided into 5 layers according to the height. The density-based DBSCAN algorithm is used to cluster the leaf points in each layer into multiple leaf point clusters, and the planes are fitted to each leaf point cluster in each layer and the normal vectors of each leaf point cluster are calculated;

[0112] According to the definition of leaf inclination angle, the leaf inclination angle of each leaf point cluster in each layer is obtained by using the normal vector of the fitting plane of each leaf point cluster in each layer. The average value of multiple leaf inclination angles in each layer is calculated to obtain the five-layer leaf inclination angle distribution curve of field corn.

[0113] According to the above method, the leaf angle and leaf inclination angle distribution curves of the five layers of corn were obtained respectively, and compared with the measured values, the optimal leaf point for calculating the leaf angle and leaf inclination angle distribution curve was obtained.

[0114] The specific implementation modes of the present invention are further described in detail below with reference to the accompanying drawings and examples.

[0115] The corn in the example is located in the Xiaotangshan National Precision Agriculture Demonstration Base in Changping District, Beijing (40.18°N, 116.44°E). The experimental area has a total of 60 plots, including 5 varieties A1-Zhengdan 958, A2-Xianyu 335, A3-Jingnongke 728, A4-Chengdan 30 and A5-Jingpin 6, 3 replicates and 4 planting densities, namely 3000 plants / mu, 4500 plants / mu, 6000 plants / mu and 7000 plants / mu. The data can be used to analyze the changes in the distribution of leaf angles and leaf inclination angles of corn of different varieties and different planting densities.

[0116] 1. Ground-based LiDAR data acquisition

[0117] FARO Focuss 350 ground-based laser radar was used to obtain high-precision corn point cloud data. Before scanning, image control points were evenly arranged in the test area for later stitching of multi-site data. To ensure more complete and efficient scanning of the data of each plot, the "5-point method" was adopted. A total of 5 sites were arranged in four plots, one station in each of the four corners and one station in the middle, and a total of 39 stations were arranged in 60 plots.

[0118] The scanning range of FARO Focus 350 is 0.6m-350m, which is set to 50m in this example. The ranging error is ±1mm, the point error is 2mm, the beam divergence angle is 0.3mrad, the output light diameter is 2.12mm, the horizontal scanning angle is 0°~360°, the vertical scanning angle is -60°~90°, and the scanning time per station is 2′45″.

[0119] 2. Ground-based LiDAR data preprocessing

[0120] Data preprocessing mainly includes four parts: multi-site scanning data solution and stitching, denoising, cell segmentation and ground point filtering.

[0121] FARO SCENE software is used to solve and stitch the acquired ground-based LiDAR data. FARO SCENE can efficiently and easily process and manage scan data through automatic object recognition, scan stitching and positioning;

[0122] The Statistical Outlier Removal (SOR) algorithm integrated in the CloudCompare software is used to perform denoising by determining the standard deviation multiplier threshold and the number of points used to estimate the average distance.

[0123] Use cloth filter algorithm CSF to separate corn point cloud and ground point cloud;

[0124] Draw the vector boundary lines of each cell based on the point cloud data, adjust the vector boundary lines of each cell and the point cloud data to the same height, and complete the cell segmentation in batches in the CloudCompare software.

[0125] 3. Separation of stems and leaves

[0126] Manual stem-leaf separation was performed on 10 plots. A cylinder with a height of 0.11 m and a diameter of 0.07 m was used to traverse the stem points and leaf points respectively to obtain their neighborhood points. The multidimensional feature values ​​of the neighborhood points and the angle between the direction vector and the vertical direction were used as training values. The Random Forest (RF) classifier was used to complete the stem-leaf separation of corn in the remaining plots. Figure 2 .

[0127] 4. Extraction of corn leaf angle and leaf inclination distribution

[0128] 4.1. Use DBSCAN clustering algorithm to cluster the thinned stalk points, where MinPts is 10 and Eps is 0.1m. Figure 3 .

[0129] 4.2. Fitting a straight line to the clustering results of the stem points:

[0130] The leaf points within the range of [0.05m, 0.10m], [0.05m, 0.15m], [0.10m, 0.15m], [0.15m, 0.20m], [0.20m, 0.25m] and [0.20m, 0.30m] from the stem fitting line are extracted to calculate the leaf angle and leaf inclination angle distribution curves, such as Figure 4 .

[0131] The calculated leaf angle and leaf inclination angle diagrams are compared with the measured leaf angle and leaf inclination angle values ​​to obtain the optimal leaf angle and leaf inclination angle distribution curves, such as Figure 5 , where a1 and b1 are leaf points within the range of [0.05m, 0.10m] and [0.15m, 0.25m] from the stem fitting line, respectively; a2 and b2 are leaf point clusters; a3 and b3 are clustering results divided into 5 layers according to height; a4 and b4 are the extracted leaf angle and measured leaf angle distribution curves and the extracted leaf inclination angle and measured leaf inclination angle distribution curves, respectively.

[0132] By comparison, it is found that the optimal leaf points for extracting leaf angle and leaf inclination angle distribution are the points within the range of [0.05m, 0.10m] and [0.15m, 0.25m] from the stem fitting line respectively.

[0133] Finally, it should be noted that what is disclosed above is only a specific embodiment of the present invention, but the embodiment of the present invention is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn, characterized in that: The following steps are involved: Scan and obtain point cloud data of corn in the field; Separate the stem and leaf of the point cloud data of field corn to obtain the stem points and leaf points of field corn; The stalk points of field corn were thinned out, and the density-based DBSCAN spatial clustering algorithm was used to cluster the stalk points of field corn, and the stalk points of each corn plant were extracted respectively; The least squares algorithm is used to fit the extracted stalk points of each corn plant to obtain the fitting line of the stalk points of each corn plant; Points within a certain range from the fitting lines of each stalk point are selected from the leaf points of field corn, and the points within the range are divided into multiple layers according to their heights. The density-based DBSCAN spatial clustering algorithm is used to cluster the leaf points in each layer into multiple leaf point clusters. Planes are fitted to each leaf point cluster in each layer and the normal vectors of each leaf point cluster are calculated. The multi-layer leaf angle distribution curve and multi-layer leaf inclination angle distribution curve of field corn are calculated according to the normal vector of the plane fitted by each leaf point cluster in each layer; Before the point cloud data of the field corn is separated into stems and leaves, the point cloud data of the field corn is also preprocessed; The preprocessing of the point cloud data of field corn includes solving and splicing of multi-site scanning data, point cloud denoising, ground point filtering and plot segmentation; After preprocessing, corn point cloud data of multiple plots are obtained; The method of separating the stems and leaves of the point cloud data of field corn comprises the following steps: All data points in the corn point cloud data of each plot are traversed by a cylinder to obtain the neighborhood points of each data point; Construct a covariance matrix for the neighborhood points of each data point, use singular value decomposition to obtain the multidimensional eigenvalues ​​and corresponding eigenvectors of the neighborhood points of each data point, and judge the distribution of the neighborhood points of each data point based on the distribution of the multidimensional eigenvalues ​​of the neighborhood points of each data point; The eigenvector corresponding to the maximum eigenvalue in the multidimensional eigenvalues ​​is the direction vector of the neighborhood point of the data point, and the angle between the direction vector of the neighborhood point of the data point and the vertical direction is calculated; Separating the stalk points and leaf points in the corn point cloud data of the plot according to the distribution of the neighborhood points of each data point and the size of the angle between the direction vector of the neighborhood point of each data point and the vertical direction; The density-based DBSCAN spatial clustering algorithm is used to cluster the stalk points of field corn, including: Select a point p from the stalk points in the corn point cloud data of each plot; Find nearby points N whose distance from point p is within and including Eps Eps (q), then N Eps (q) is defined as the set of points q including q itself, expressed as: N Eps (q)={q∈D|distance(p,q)≤Eps} (1) Where Eps represents the neighborhood radius value of clustering time point p; N Eps (q) indicates the nearby points within and including Eps from point p; When N Eps If the number of (q) is ≥ MinPts, then point p and its nearby points form a stalk point cluster; When N Eps If the number of (q) is less than MinPts, the points near point p are marked as noise points; Where MinPts represents the threshold of the number of points in the neighborhood whose distance to point p is Eps; Fitting a plane to the blade point cluster and obtaining a normal vector of the blade point cluster includes: Using the coordinates of several three-dimensional points (x i ,y i , z i ) Establish a fitting plane equation, where i = 1, 2, 3, ...; The constraint condition of the fitting plane equation is a 2 +b 2 +c 2 =1, and the sum of the distances from the fitting plane to all points in the blade point cluster is the smallest, where the expression of the fitting plane equation is: ax+by+cz=d (3) Let the average coordinates of all points in each leaf point cluster be but: Subtracting equation (3) from equation (4), we get: Assumption Matrix Column Matrix Then equation (5) is equivalent to AX=0 Perform singular value decomposition on the matrix A, and the eigenvector corresponding to the minimum singular value of the matrix A is the normal vector of the fitting plane of the leaf point cluster.

2. The method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn according to claim 1, characterized in that: The method for scanning and acquiring point cloud data of field corn comprises: Arrange two or more ground-based lidar scanning stations in the corn field; Set the scanning mode and scanning range of the ground-based laser radar; The scanning parameters of the ground-based laser radar are set, wherein the scanning parameters of the ground-based laser radar include scanning resolution, quality, horizontal and vertical scanning angles.

3. The method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn according to claim 1, characterized in that: The distance-based method was used to thin out the stalk points in the corn point cloud data of each plot, so that the minimum distance between points was 0.01m.

4. The method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn according to claim 1, characterized in that: The least squares algorithm is used to fit the extracted stalk points of each corn plant to obtain a stalk point fitting line of each corn plant, including: Select any point (x0, y0, z0) on the stalk of each corn plant; Get the direction vector (m,n,p) of the fitting line of each corn stalk point, then the linear equation of the fitting line of each corn stalk point is:

5. The method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn according to claim 1, characterized in that: Among the leaf points in the corn point cloud data of each plot, points within the range of [0.05m, 0.10m], [0.05m, 0.15m] and [0.10m, 0.15m] from the stalk fitting line were selected, and the points in each range were divided into 5 layers according to the height; The density-based DBSCAN algorithm is used to cluster the leaf points in each layer into multiple leaf point clusters, and planes are fitted to each leaf point cluster in each layer and the normal vectors of each leaf point cluster are calculated. According to the definition of leaf angle, the leaf angle of each leaf point cluster in each layer is obtained by using the normal vector of the fitting plane of each leaf point cluster in each layer. The average value of multiple leaf angles in each layer is calculated to obtain the five-layer leaf angle distribution curve of field corn.

6. The method for extracting high-throughput leaf angle and leaf inclination angle distribution of field corn according to claim 1, characterized in that: Among the leaf points in the corn point cloud data of each plot, points at the distances from the stalk fitting line [0.15m, 0.20m], [0.20m, 0.25m], [0.15m, 0.25m], and [0.20m, 0.30m] were selected, and the points in each range were divided into 5 layers according to the height; The density-based DBSCAN algorithm is used to cluster the leaf points in each layer into multiple leaf point clusters, and planes are fitted to each leaf point cluster in each layer and the normal vectors of each leaf point cluster are calculated. According to the definition of leaf inclination angle, the leaf inclination angle of each leaf point cluster in each layer is obtained by using the normal vector of the fitting plane of each leaf point cluster in each layer. The average value of multiple leaf inclination angles in each layer is calculated to obtain the five-layer leaf inclination angle distribution curve of field corn.

Images