Rice phenotypic method and system based on unmanned aerial vehicle laser radar

Abstract

The invention discloses a rice phenotypic method and system based on unmanned aerial vehicle laser radar, and the method comprises the following steps: S1, carrying out the data collection of rice in the key phenological period of rice growth, and carrying out the in-situ three-dimensional reconstruction based on the collected data, constructing a point cloud data set of the whole growth cycle of the rice; s2, preprocessing the point cloud data set, and calculating phenotypic parameters of rice based on the preprocessed point cloud data set; and S3, establishing a mapping relation between the phenotypic parameters and the rice growth state for assisting rice breeding work. According to the method, comprehensive rice phenotype information can be provided, and the growth vigor of the rice in the whole growth cycle can be detected. In addition, the external morphology and character changes in the rice growth process are quantified, and powerful data support is provided for a rice breeding expert group.

Description

Technical Field

[0001] The present invention relates to agricultural monitoring technology, and in particular to a rice phenotyping method and system based on unmanned aerial vehicle laser radar. Background Art

[0002] As one of the three major food crops, rice feeds more than half of the world's population. However, with the rapid increase in population and the reduction in effective arable land, the high yield of rice will directly affect global food security. In order to improve global food security and promote sustainable agricultural development, rice breeding has received more and more attention. The goal of breeding is to improve the productivity, adaptability, resistance and quality of rice by selecting and improving the phenotypic traits of rice. Because the phenotype directly reflects the key characteristics of rice growth, development, yield, quality, etc., the rapid acquisition of rice phenotype, that is, the external morphology and traits of rice, has a vital impact on breeding.

[0003] Traditional rice phenotyping Traditional crop phenotyping data acquisition mainly relies on manual measurement, which is inefficient, inaccurate and highly destructive, and cannot meet the needs of modern agricultural scientific research. Although various two-dimensional imaging technologies have been used to non-destructively extract rice canopy parameters, rice ears and leaves are identified through image processing technology. However, during the growth of rice, leaves will block each other and the canopy closure will increase. Two-dimensional images have great limitations in extracting certain features (such as plant height, canopy volume, etc.). In addition, optical imaging technology requires certain lighting conditions, and is greatly affected by the environment when used in outdoor fields, and the effect of each imaging cannot be guaranteed. In recent years, laser radar technology has developed rapidly. Light detection and ranging (LiDAR) is an active remote sensing technology that overcomes many shortcomings of passive remote sensing technologies. LiDAR directly uses a pulsed laser beam to calculate the distance from the laser scanner to the target, through which the three-dimensional geometric features of the target crop can be recorded in a point cloud data set. Therefore, using this technology for crop phenotyping can avoid the influence of light and obtain rich geometric information. The laser beam can penetrate the vegetation to obtain the internal structure of the vegetation canopy, and more accurately obtain phenotypic information such as plant height, canopy coverage index, canopy volume, etc.

[0004] In addition, the rapid development of drones in recent years has brought new possibilities for large-scale field phenotyping. Drones have excellent flexibility and good load capacity, and can be quickly deployed as large-scale phenotyping platforms in the field. Drones equipped with lidar can quickly and accurately obtain canopy information of rice fields. However, drone detection still has defects such as detection accuracy and efficiency. Summary of the invention

[0005] The first purpose of the present invention is to overcome the shortcomings of the above existing technologies and provide a rice phenotyping method based on drone laser radar. This rice phenotyping method based on drone laser radar can quickly and accurately calculate the phenotypic parameters of different varieties and their changes, provide comprehensive data support for rice breeding, and improve breeding efficiency.

[0006] The second object of the present invention is to provide a rice phenotyping system based on UAV lidar.

[0007] The first object of the present invention is achieved by the following technical solution: The rice phenotyping method based on drone laser radar comprises the following steps:

[0008] S1. Collect data on rice during the key phenological period of rice growth, and perform in-situ 3D reconstruction based on the collected data to construct a point cloud dataset of the entire rice growth cycle;

[0009] S2, preprocessing the point cloud data set, and calculating the phenotypic parameters of rice based on the preprocessed point cloud data set;

[0010] S3. Establish a mapping relationship between phenotypic parameters and rice growth status to assist rice breeding.

[0011] Preferably, step S1 includes the following specific steps:

[0012] S11. Setting the take-off point and landing point of the drone according to the actual planting range of the rice field, and generating a full coverage route of the drone based on the take-off point and landing point;

[0013] S12. Use ROS robot operating system to manage the publishing and subscription nodes of lidar, IMU and GNSS data in drones;

[0014] S13, tightly couple and fuse the lidar and IMU data through the SLAM algorithm Fast-lio2, and manage the map through the optimized ikd-tree structure;

[0015] S14. Obtain the three-dimensional point cloud of the rice canopy during the rice transplanting period, tillering period, jointing period, fruiting period and maturity period, and use it to construct a point cloud dataset for the entire growth cycle of rice.

[0016] Preferably, the preprocessing of the point cloud data set in step S2 includes the following steps:

[0017] S21, using the statistical outlier removal method to reduce the noise of the point cloud, where the standard deviation multiple threshold is uniformly set to 1.5, and then different neighborhood sizes are set according to different growth stages of rice. The neighborhood is set to 25 during the rice transplanting period, and then 5 is superimposed in each key growth stage;

[0018] S22, using the cloth simulation algorithm to separate ground points, the slope threshold is set to 10°, the vertical threshold is set to 0.1m, the grid size is set to 0.1m, the maximum number of iterations varies with the rice growth stage, the number of iterations is set to 800 during the rice transplanting period, and then 200 are superimposed in each key growth stage;

[0019] S23. Perform point cloud segmentation according to the planting area of ​​different rice varieties, and accurately separate the point cloud of each variety of rice.

[0020] Preferably, the process of calculating the phenotypic parameters of rice based on the preprocessed point cloud data set in step S2 is as follows:

[0021] S24, integrating point cloud denoising, ground point separation, plot division and phenotypic parameter calculation into one software;

[0022] S25, during the processing, the point cloud of each step is visualized;

[0023] The calculation of rice phenotypic parameters calculated by S26 are all embedded in the software.

[0024] The calculation of rice phenotypic parameters includes plant height calculation, volume ratio calculation, canopy cover index calculation and leaf area index calculation.

[0025] Step S3 includes the following specific steps:

[0026] Provide the expert panel with specific values ​​of phenotypic parameters of each rice variety at all key phenological stages;

[0027] Provide the expert group with information on the variation of phenotypic parameters of different rice varieties;

[0028] Rice varieties in the breeding garden are classified based on breeding evaluation criteria.

[0029] The second object of the present invention is achieved by the following technical solution: a rice phenotyping system based on UAV laser radar, comprising:

[0030] The drone module is used to initially sense the environmental status of the rice fields and generate a full-coverage flight route based on the environmental status, so as to achieve full-coverage flight over the rice fields and collect rice data;

[0031] The 3D reconstruction module is based on the rice data collected by the drone module. It combines the data input of the lidar and inertial measurement unit with the real-time positioning and mapping algorithm to realize the in-situ 3D reconstruction of rice in the breeding garden and construct a point cloud data set of the entire growth cycle of rice.

[0032] The phenotypic parameter calculation module integrates the point cloud preprocessing algorithm, the mapping relationship between the point cloud data structure and the phenotype, and the point cloud visualization during the calculation process into one software, assisting the expert group to calculate the phenotypic parameters of rice accurately and quickly.

[0033] The drone module includes a drone, a laser radar and an onboard computer. The laser radar and the onboard computer are both arranged on the drone, and the three-dimensional reconstruction module is deployed in the onboard computer.

[0034] The present invention has the following advantages over the prior art:

[0035] 1. The present invention can provide high-precision geometric information of rice fields by combining drones with laser radar technology, including field topography, three-dimensional structure information inside the canopy, etc. It can provide comprehensive rice phenotypic information and detect the growth of rice throughout its growth cycle. In addition, the external morphology and trait changes of rice during its growth process are quantified, providing strong data support for the rice breeding expert group.

[0036] 2. The present invention combines drone and lidar technology to achieve fast and accurate automatic acquisition of rice phenotypic parameters. Compared with traditional manual measurement and ground equipment measurement, it reduces damage to crops, improves detection efficiency, lowers the operating threshold, and saves resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The present invention is a flow chart of the rice phenotyping method based on UAV laser radar.

[0038] Figure 2 The framework diagram of the rice phenotyping system based on UAV laser radar of the present invention.

[0039] Figure 3 Schematic diagram of the rice phenotyping system based on UAV lidar of the present invention.

[0040] Figure 4 This is a diagram for segmenting rice when calculating plant height according to the present invention.

[0041] Figure 5 It is a process diagram for calculating volume ratio of the present invention.

[0042] Figure 6 It is a process diagram for calculating the canopy coverage index of the present invention.

[0043] Figure 7 This is a schematic diagram of the interface of the software integrating point cloud denoising, ground point separation, plot division and phenotypic parameter calculation of the present invention.

[0044] Figure 8 Schematic diagram of the point cloud denoising process in the present invention.

[0045] Fig. 9 It is a schematic diagram of ground point separation of the present invention.

[0046] Fig.10 This is the first schematic diagram of the land parcel division according to the present invention.

[0047] Fig.11 This is a second schematic diagram of the land parcel division according to the present invention.

[0048] Fig.12 This is a schematic diagram of the output results of the phenotypic parameter calculation of the present invention.

[0049] Among them, 1 is the drone, 2 is the controller, 3 is the onboard computer, 4 is the GNSS (Global Satellite Navigation System) module, 5 is the lidar, and 6 is the ground computer. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0051] like Figure 1 As shown, the rice phenotyping method based on UAV laser radar includes the following steps:

[0052] S1. Collect data on rice during the key phenological period of rice growth, and perform in-situ 3D reconstruction based on the collected data to construct a point cloud dataset of the entire rice growth cycle;

[0053] Step S1 includes the following specific steps:

[0054] S11. Set the take-off point and landing point of the drone according to the actual planting range of the rice field, and generate a full coverage route of the drone based on the take-off point and landing point; This embodiment uses a drone module composed of a laser radar carried by the drone to collect data. First, generate an operation path and realize autonomous flight based on the preliminary perceived environmental information and biological correlation issues, that is, obtain the position information and morphological information of all objects in the rice field (rice plants, ridges, obstacles, etc.) through the scanning results of the laser radar. Then, the paddy field environmental information is preliminarily analyzed by the onboard computer and the ground station, and then a flight path that can fully cover the paddy field is generated. Finally, the path is uploaded to the flight control to realize the autonomous flight mission. In this embodiment, the take-off point and landing point of the drone are set according to the actual planting range of the rice field. The straight-line distance between the take-off point and the landing point and the boundary of the planting area is more than 5m. A full-coverage route of the drone is generated based on the take-off point and the landing point. The route spacing is 5m. The flight speed of the drone is 3.5m / s during the rice transplanting period. As the rice grows, the flight speed is reduced by 0.3m / s at the next critical growth period, and the flight altitude is set to 13-15m.

[0055] S12. Use the ROS robot operating system to manage the data publishing and subscription nodes of the lidar, IMU (inertial measurement unit) and GNSS in the drone; during the flight of the drone, use the ROS robot operating system to manage the data publishing and subscription nodes of the lidar, IMU and GNSS in the drone to realize data collection for rice.

[0056] S13, tightly couple and fuse the lidar and IMU data through the SLAM algorithm Fast-lio2, and manage the map through the optimized ikd-tree structure;

[0057] S14. Obtain the three-dimensional point cloud of the rice canopy during the rice transplanting, tillering, jointing, fruiting and maturity stages, and use this to construct a point cloud data set for the entire growth cycle of rice. The onboard computer of the drone can calculate the current posture state of the drone by pre-integrating the data of the inertial measurement unit and performing inter-frame estimation of the lidar point cloud data. Combining the two to form a laser inertial navigation odometer can better make up for their respective shortcomings when used alone. Efficient and accurate posture estimation allows the drone to accurately locate its position and posture in the environment, thereby accurately and completely matching the continuously updated point cloud frames to the global map to generate a dense three-dimensional point cloud of the rice canopy.

[0058] S2, preprocessing the point cloud data set, and calculating the phenotypic parameters of rice based on the preprocessed point cloud data set;

[0059] The preprocessing of the point cloud data set in step S2 includes the following steps:

[0060] S21, using the statistical outlier removal method to reduce the noise of the point cloud, where the standard deviation multiple threshold is uniformly set to 1.5, and then different neighborhood sizes are set according to different growth stages of rice. The neighborhood is set to 25 during the rice transplanting period, and then 5 is superimposed in each key growth stage;

[0061] S22, using the cloth simulation algorithm to separate ground points, the slope threshold is set to 10°, the vertical threshold is set to 0.1m, the grid size is set to 0.1m, the maximum number of iterations varies with the rice growth stage, the number of iterations is set to 800 during the rice transplanting period, and then 200 are superimposed in each key growth stage;

[0062] S23. Perform point cloud segmentation according to the planting area of ​​different rice varieties, and accurately separate the point cloud of each variety of rice.

[0063] Point cloud denoising uses a statistical outlier removal method to remove noise points that are far away from the large point cloud; ground points are separated using a cloth simulation algorithm, which calculates the plane of the ground through iterative fitting and then removes the ground points; point cloud segmentation is performed by calling the open3D library to select the point cloud by creating a three-dimensional box that conforms to the planting range of different varieties of rice.

[0064] The process of calculating the phenotypic parameters of rice based on the preprocessed point cloud dataset in step S2 is as follows:

[0065] S24, integrate point cloud denoising, ground point separation, plot division and phenotypic parameter calculation into one software. The software interface is as follows Figure 7 As shown;

[0066] Noise reduction and horizontalization of point clouds, such as Figure 8 and Fig. 9 As shown. Due to the inevitable obstruction in the environment and the influence of wind that causes the drone fuselage and crop leaves to shake, the generated 3D point cloud will have unnecessary noise, and the slope and terrain features of the rice field will also affect the accuracy of parameter extraction. Therefore, this embodiment develops a process for reducing the error caused by the environment, which includes point cloud denoising, plane fitting and rotation alignment. First, the 3D point cloud is denoised and the Statistical Outlier Removal (SOR) algorithm is used to remove outliers. This method calculates the average distance between a given 3D point and a neighboring point. If the average distance is outside the standard range, the point is marked as an outlier. The points classified as outliers are marked and deleted from subsequent analysis. Then, the least squares plane fitting method is used to find the best fitting plane of the point cloud, and the rotation angle between the normal vector of the plane and the Z axis is calculated. Finally, the point cloud is rotated by the transformation matrix so that its normal vector is aligned with the Z axis.

[0067] Separate ground points (such as Fig. 9 and Fig.10 As shown,). In this embodiment, a ground filter is developed based on the cloth simulation (CSF) algorithm to separate ground points from the denoised point cloud, and the entire point cloud map will be divided into ground points and non-ground points. In order to make the CSF algorithm originally designed for ultra-large land monitoring more suitable for processing field rice point clouds, the filter is optimized by reducing its grids and nodes.

[0068] In order to obtain enough samples to support statistical analysis, the point cloud was divided into 3×3m plots according to the experimental design (e.g. Fig.10 and Fig.11As shown in Figure 2.1 ...

[0069] Output results (such as Fig.12 According to the above definitions, PH, LAI, and f c The volume and volume ratio of the ear layer are calculated, and the final output (denoising of point cloud, separation of ground points, division of plots, and calculation of phenotypic parameters) is an Excel table, and the data are arranged according to the plot number.

[0070] S25. During the processing, the point cloud of each step is visualized; when performing point cloud preprocessing, ground point separation and land parcel division operations, different parameters can be input on the interface for adjustment;

[0071] The calculation of rice phenotypic parameters calculated by S26 are all embedded in the software.

[0072] The calculation of rice phenotypic parameters includes plant height calculation, volume ratio calculation, canopy coverage index calculation and leaf area index calculation. The details are as follows:

[0073] For plant height (PH) calculation:

[0074] Since rice is grown mechanically, after segmenting each plot, we can divide it into smaller, uniformly sized areas (such as Figure 4 In this way, the height of each rice plant in the plot can be calculated, and then the statistical values ​​can be analyzed based on all the plant height data in the plot, such as the maximum, minimum and average heights of all rice plants in the plot. The specific calculation method is as follows: First, use the o3d.io.read_point_cloud function in the open3D library to load the point cloud (PCD format). Then, use the np.asarray function in the NumPy library to obtain the point cloud data index of each plot, including the elevation values ​​of all points in the point cloud and the values ​​of different height percentiles (such as Figure 4 Finally, the elevation value of the ground point is subtracted from the elevation value of the corresponding height percentile of the point cloud in the region (the 90th, 95th, 99th, and 100th percentiles are selected in this embodiment), as shown in the following formula:

[0075] PH=Z X％ -Z ground

[0076] Where Z X％ Z is the X% elevation value of the point cloud in the area. grounis the elevation of the ground point in the area.

[0077] For volume ratio calculation:

[0078] In order to more accurately describe the growth of rice, panicle volume and volume ratio were introduced to measure the resource allocation of rice plants during their growth, especially the balance between reproductive growth and vegetative growth.

[0079] The panicle layer is defined as the area above the natural height of the panicle of rice, and the volume ratio is defined as the ratio of the panicle volume to the canopy volume. This embodiment uses the alpha shape algorithm to calculate the canopy volume and panicle volume of each plot. Unlike the traditional convex hull algorithm, the Alpha shape algorithm can not only generate the outer boundary of the point set, but also capture the internal structure of the point set. It can generate complex boundaries and reveal the topological features of the data, such as holes and internal structures. Therefore, the algorithm will show better accuracy and take less time when dealing with short and densely planted crops such as rice. Figure 5 As shown, the specific calculation process is:

[0080] Use the o3d.geometry.AxisAlignedBoundingBox function in the open3d library to define a three-dimensional rectangular area aligned along the coordinate axis. The X and Y planes of the rectangular area are consistent with the size of the plot, and the height is the elevation of the ear layer. Use this three-dimensional rectangle to intercept the point cloud of the ear layer range of each plot, and create a new mask to save the ear layer point cloud.

[0081] The alpha shape algorithm is used to calculate the panicle volume and canopy volume respectively. For short and densely planted crops such as rice, selecting α = 0.1 can quickly and accurately calculate the canopy volume and panicle volume of rice based on the plot level while taking into account the calculation speed;

[0082] Calculate the rice canopy volume V C , the corresponding panicle volume V S Then calculate the volume ratio V Ratio , as shown below:

[0083]

[0084] For canopy cover index calculation:

[0085] The canopy coverage index reflects the coverage of vegetation on the horizontal projection plane and is an important vegetation structure indicator. It can provide a more comprehensive understanding of the growth and functional status of rice. The canopy coverage index is calculated for each plot, such as Figure 6 As shown, the calculation process is as follows:

[0086] Filter the point cloud with a maximum height of 60%. Use the np.max function in the NumPy library to calculate the maximum height of the point cloud, create a new mask to indicate whether each point meets the condition that the height is equal to 60% of the maximum height, and then use the pcd.select_by_index function in the open3d library to retain the points that meet the conditions;

[0087] Project the point cloud to the X, Y plane. Copy the filtered point cloud, project it to the X, Y plane (set the Z coordinate to 0), and then create a new point cloud object to store the projected points;

[0088] The grid method is used to calculate the point cloud area ratio. The X, Y plane is divided into a series of small grids, each with a side length of 0.03. For each grid, the point cloud data is traversed to calculate the number of points falling within the grid. If the number of points within the grid is greater than 5, it is recorded as a valid grid. The total area covered by the point cloud is calculated (the number of valid grids multiplied by the grid area). Finally, the point cloud area ratio is calculated, that is, the total area covered by the point cloud divided by the plot area.

[0089] Calculation for Leaf Area Index (LAI):

[0090] Monitoring and estimation of LAI can help agricultural producers optimize planting density, fertilization and irrigation strategies, and improve crop yield and quality. This embodiment will calculate LAI based on the Beer-Lambert law, that is, calculate LAI through the gap ratio and extinction coefficient, and the gap ratio can be directly calculated from the point cloud. The specific formula is as follows:

[0091]

[0092] Where C is the interstitial ratio, κ is the extinction coefficient, and ln is the natural logarithm.

[0093] The gap ratio is calculated using the following formula:

[0094]

[0095] Among them, N ground is the number of pre-separated ground points, N total The total number of points.

[0096] S3. Establish a mapping relationship between phenotypic parameters and rice growth status to assist rice breeding.

[0097] Step S3 includes the following specific steps:

[0098] Provide the expert panel with specific values ​​of phenotypic parameters of each rice variety at all key phenological stages;

[0099] Provide the expert group with information on the variation of phenotypic parameters of different rice varieties;

[0100] Rice varieties in the breeding garden are classified based on breeding evaluation criteria.

[0101] like Figure 2 As shown, the UAV-based LiDAR rice phenotyping system includes:

[0102] The drone module is used to initially sense the environmental status of the rice fields and generate a full-coverage flight route based on the environmental status, so as to achieve full-coverage flight over the rice fields and collect rice data;

[0103] The 3D reconstruction module is based on the rice data collected by the drone module. It combines the data input of the lidar and inertial measurement unit with the real-time positioning and mapping algorithm to realize the in-situ 3D reconstruction of rice in the breeding garden and construct a point cloud data set of the entire growth cycle of rice.

[0104] The phenotypic parameter calculation module integrates the point cloud preprocessing algorithm, the mapping relationship between the point cloud data structure and the phenotype, and the point cloud visualization during the calculation process into one software, assisting the expert group to calculate the phenotypic parameters of rice accurately and quickly.

[0105] like Figure 3 As shown, the drone module includes a drone, a laser radar and an onboard computer, the laser radar and the onboard computer are both arranged on the drone, and the three-dimensional reconstruction module is deployed in the onboard computer. Specifically, the drone of this embodiment is a six-rotor drone with a battery capacity of 16000mAh, a rotor size of 38.1cm, a maximum carrying weight of 8kg, and a Livox Mid-360 laser radar carried. It is a hybrid solid-state laser radar with a wavelength of 905nm. The model of the onboard computer is Intel NUC 13Pro, and the program in the three-dimensional reconstruction module is deployed in the onboard computer for real-time generation of three-dimensional point clouds of rice. In order to ensure better flight of the drone, the drone is also provided with a controller and a GNSS module.

[0106] The above specific implementation modes are preferred embodiments of the present invention and cannot be used to limit the present invention. Any other changes or other equivalent replacement methods that do not deviate from the technical solution of the present invention are included in the protection scope of the present invention.

Claims

1. A rice phenotyping method based on UAV laser radar, characterized in that: The following steps are involved: S1. Collect data on rice during the key phenological period of rice growth, and perform in-situ 3D reconstruction based on the collected data to construct a point cloud dataset of the entire rice growth cycle; S2, preprocessing the point cloud data set, and calculating the phenotypic parameters of rice based on the preprocessed point cloud data set; S3. Establish a mapping relationship between phenotypic parameters and rice growth status to assist rice breeding.

2. The rice phenotyping method based on UAV laser radar according to claim 1, characterized in that: Step S1 includes the following specific steps: S11. Setting the take-off point and landing point of the drone according to the actual planting range of the rice field, and generating a full coverage route of the drone based on the take-off point and landing point; S12. Use ROS robot operating system to manage the publishing and subscription nodes of lidar, IMU and GNSS data in drones; S13, tightly couple and fuse the lidar and IMU data through the SLAM algorithm Fast-lio2, and manage the map through the optimized ikd-tree structure; S14. Obtain the three-dimensional point cloud of the rice canopy during the rice transplanting period, tillering period, jointing period, fruiting period and maturity period, and use it to construct a point cloud dataset for the entire growth cycle of rice.

3. The rice phenotyping method based on UAV laser radar according to claim 1, characterized in that: The preprocessing of the point cloud data set in step S2 includes the following steps: S21, using the statistical outlier removal method to reduce the noise of the point cloud, where the standard deviation multiple threshold is uniformly set to 1.5, and then different neighborhood sizes are set according to different growth stages of rice. The neighborhood is set to 25 during the rice transplanting period, and then 5 is superimposed in each key growth stage; S22, using the cloth simulation algorithm to separate ground points, the slope threshold is set to 10°, the vertical threshold is set to 0.1m, the grid size is set to 0.1m, the maximum number of iterations varies with the rice growth stage, the number of iterations is set to 800 during the rice transplanting period, and then 200 are superimposed in each key growth stage; S23. Perform point cloud segmentation according to the planting area of ​​different rice varieties, and accurately separate the point cloud of each variety of rice.

4. The rice phenotyping method based on UAV laser radar according to claim 1, characterized in that: The process of calculating the phenotypic parameters of rice based on the preprocessed point cloud dataset in step S2 is as follows: S24, integrating point cloud denoising, ground point separation, plot division and phenotypic parameter calculation into one software; S25, during the processing, the point cloud of each step is visualized; The calculation of rice phenotypic parameters calculated by S26 are all embedded in the software.

5. The rice phenotyping method based on UAV laser radar according to claim 1, characterized in that: The calculation of rice phenotypic parameters includes plant height calculation, volume ratio calculation, canopy cover index calculation and leaf area index calculation.

6. The rice phenotyping method based on UAV laser radar according to claim 1, characterized in that: Step S3 includes the following specific steps: Provide the expert panel with specific values ​​of phenotypic parameters of each rice variety at all key phenological stages; Provide the expert group with information on the variation of phenotypic parameters of different rice varieties; Rice varieties in the breeding garden are classified based on breeding evaluation criteria.

7. The rice phenotyping system based on UAV laser radar is characterized by: include: The drone module is used to initially sense the environmental status of the rice fields and generate a full-coverage flight route based on the environmental status, so as to achieve full-coverage flight over the rice fields and collect rice data; The 3D reconstruction module is based on the rice data collected by the drone module. It combines the data input of the lidar and inertial measurement unit with the real-time positioning and mapping algorithm to realize the in-situ 3D reconstruction of rice in the breeding garden and construct a point cloud data set of the entire growth cycle of rice. The phenotypic parameter calculation module integrates the point cloud preprocessing algorithm, the mapping relationship between the point cloud data structure and the phenotype, and the point cloud visualization during the calculation process into one software, assisting the expert group to calculate the phenotypic parameters of rice accurately and quickly.

8. The rice phenotyping system based on UAV laser radar according to claim 7, characterized in that: The drone module includes a drone, a laser radar and an onboard computer. The laser radar and the onboard computer are both arranged on the drone, and the three-dimensional reconstruction module is deployed in the onboard computer.

Images