Method for estimating leaf area index of each leaf layer based on wheat canopy light distribution

Abstract

The invention discloses a method for estimating the leaf area index of each leaf layer based on wheat canopy light distribution. The method comprises the following steps: S1, collecting multispectral data and wheat canopy three-dimensional point cloud data through an unmanned aerial vehicle; s2, carrying out multispectral image splicing and canopy three-dimensional point cloud reconstruction; s3, generating a shape file of each cell by utilizing ArcGIS software, cutting the multispectral image by utilizing ENVI software, cutting the canopy point cloud by utilizing PCM software, and obtaining the multispectral image and three-dimensional point cloud of each cell; s4, extracting a vegetation index of each cell by utilizing MATLAB software; s5, counting the number of point clouds of different layers of the canopy point clouds; s6, constructing a canopy light distribution estimation model by using a machine learning algorithm based on the number characteristics of the point clouds of different layers; and S7, based on the estimated canopy light distribution characteristics and vegetation indexes, using a machine learning algorithm to construct different leaf layer LAI estimation models. According to the invention, large-scale effective monitoring of the leaf area index can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of crop monitoring, and in particular to a method for estimating leaf area index of each leaf layer based on wheat canopy light distribution. Background Art

[0002] Winter wheat leaves are the main organs for photosynthesis in crops, and the green leaf area index (LAI) is the core indicator for measuring its photosynthetic capacity. The photosynthetic efficiency of the canopy is directly related to the carbon fixation capacity and final yield of the crop, and the size of the LAI reflects the interception and utilization of light energy by the plant community. The canopy LAI not only represents the biomass and growth status of the entire canopy, but also reveals the photosynthetic contribution of different leaf layers. Especially in the filling stage, there are significant differences in the effects of LAI of different leaf layers on the absorption of light energy and photosynthetic efficiency. The upper leaves (such as flag leaves) often undertake most of the photosynthesis, while the lower leaves may be restricted by light when the canopy is closed. Accurately measuring the LAI of each leaf layer, especially in the filling stage, can help study the canopy aging process and the dynamic changes of photosynthetic capacity. In addition, LAI can also be used to guide management measures such as fertilization and irrigation to optimize crop resource utilization. Therefore, understanding and accurately estimating the LAI of each leaf layer is extremely important for improving crop management efficiency, predicting yield, and improving resource utilization.

[0003] There are two main traditional methods for determining LAI. One is to measure the actual size of leaves through destructive sampling in the field. This method requires selecting representative samples in the field, measuring the length and width of the leaves one by one, and then calculating the area of ​​a single leaf through a mathematical formula. Finally, the LAI of the entire field is calculated based on the plant density. Although this method is accurate, it is time-consuming and labor-intensive, especially in large-scale crop fields. The time and labor costs required are extremely high, and it is difficult to achieve efficient large-scale measurements. The second is based on optical equipment, such as handheld devices such as LAI-2000 and LAI-2200, which measure the light transmittance of the canopy and combine it with radiation transfer theory to estimate LAI. This method can obtain results more quickly and is suitable for measurements in smaller areas, but its application is also limited by the sampling method of handheld devices, making it difficult to estimate LAI over a large area. In addition, the measurement results of handheld devices depend on changes in canopy structure and lighting conditions, which may lead to a decrease in data accuracy. Therefore, the two traditional methods have their own advantages and disadvantages in terms of accuracy and efficiency. Especially in modern large-scale agricultural production, they still cannot meet the needs of efficient and accurate large-scale crop monitoring.

[0004] With the development of remote sensing technology and UAV platforms, UAV-based LAI estimation methods have gradually become mainstream. UAVs can collect data on large areas of farmland in a short time due to their flexibility, low cost and easy operation. They can also carry different types of sensors, such as RGB cameras, multispectral, hyperspectral, LiDAR, etc., providing a rich data source for LAI estimation. Common estimation methods are divided into physical regression models and empirical regression models. The physical regression model is based on the reflection and absorption characteristics of crops to light, and combines radiation transmission theory and complex mathematical formulas to estimate LAI. However, the model requires many parameters and is complex to calculate, making it difficult to apply to the optimal estimation under different conditions. Relatively speaking, the empirical regression model relies on the relationship between the spectral reflectance in UAV images and the crop phenotypic parameters. It is simple and efficient and has been widely used in crop LAI estimation. Nevertheless, due to the spectral saturation problem of densely planted crops such as wheat, the vegetation index obtained by UAVs has a decrease in accuracy at high LAI. To solve this problem, researchers introduced parameters such as texture features, canopy height, and canopy coverage, and built models through machine learning or deep learning algorithms to improve the accuracy of estimation. However, existing studies mostly focus on the estimation of the total LAI of the canopy, while there are still few studies on the estimation of LAI at different leaf layers, which limits the in-depth understanding and optimization of canopy photosynthetic efficiency. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a method for estimating the leaf area index of each leaf layer based on wheat canopy light distribution, which can reduce the labor and time costs invested in the original leaf area index measurement process and realize large-scale effective monitoring.

[0006] Technical solution: A method for estimating leaf area index of each leaf layer based on light distribution in the wheat canopy, comprising the following steps:

[0007] S1, collecting multispectral data and wheat canopy 3D point cloud data by UAV;

[0008] S2, multispectral image stitching and canopy 3D point cloud reconstruction using DJI Terra software;

[0009] S3, using ArcGIS software to generate shape files for each plot, using ENVI software to crop multispectral images, using PCM software to crop canopy point clouds, and obtaining multispectral images and three-dimensional point clouds for each plot;

[0010] S4, import the multispectral image of each plot into MATLAB software to extract the vegetation index of each plot;

[0011] S5, counting the number of point clouds at different layers of the canopy point cloud;

[0012] S6, based on the quantitative characteristics of point clouds at different layers, a canopy light distribution estimation model was constructed using a machine learning algorithm;

[0013] S7, based on the estimated canopy light distribution characteristics and vegetation index, a machine learning algorithm was used to construct an LAI estimation model for different leaf layers.

[0014] Further, in step S5, the steps for counting the number of point clouds at different layers of the canopy point cloud are as follows:

[0015] S51, importing the obtained point clouds of different plots into PCM software to obtain canopy height information;

[0016] S51, normalizing the number of point clouds in different cells according to the bottom area of ​​the minimum bounding box of the point cloud;

[0017] S53, according to the measurement position of the canopy vertical light distribution, the canopy point cloud is divided into 4 layers, namely: flag leaf and above layer, 2nd leaf layer-flag leaf layer, 3rd leaf layer-2nd leaf layer, 3rd leaf and below layer;

[0018] S54, counting the number of point clouds at different layers in step S53.

[0019] Furthermore, in the canopy light distribution estimation model, the number of canopy point clouds is used as an input parameter, and the canopy light distribution characteristics are used as output parameters, wherein the canopy light distribution characteristics include photosynthetically active radiation interception rate and relative light intensity;

[0020] The calculation formulas for the photosynthetically active radiation interception rate of each leaf layer and the relative light intensity of each leaf layer are as follows:

[0021]

[0022] Among them, LIR represents the photosynthetically active radiation interception rate, RLI represents the relative light intensity of each leaf layer; PAR n is the photosynthetically active radiation of each leaf layer, in μmol·m -2 ·s -1 ; TPAR is the photosynthetically active radiation at the top of the canopy, in μmol·m -2 ·s -1 ; n represents different leaf layers.

[0023] Furthermore, in the LAI estimation model for different leaf layers, the estimated canopy light distribution characteristics and vegetation index are used as input parameters, and the estimated leaf area index value of each leaf layer is used as output parameter;

[0024] The LAI calculation formula for each leaf layer is as follows:

[0025] LAI = leaf area of ​​each leaf layer per single stem × total number of single stems

[0026] The leaf area of ​​each leaf is calculated as follows:

[0027] Leaf area = leaf length × leaf width × 0.83

[0028] Among them, the blade length refers to the distance from the base to the tip of the blade, and the blade width refers to the widest part of the blade.

[0029] Compared with the prior art, the present invention has the following significant effects:

[0030] The present invention uses an unmanned aerial vehicle equipped with a laser radar to obtain three-dimensional point cloud data of wheat canopy, and crops the point cloud in layers to obtain the number of point clouds in different layers. Based on this, a canopy light distribution estimation model is constructed, and a leaf area index estimation model for different leaf layers is constructed by combining the canopy light distribution with the vegetation index. This can reduce the labor and time costs invested in the original leaf area index measurement process, and can realize large-scale effective monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a work flow chart of the present invention;

[0032] Figure 2 Flowchart for obtaining quantitative features of point clouds at different vertical layers of the canopy;

[0033] Figure 3 Get a schematic diagram for the canopy light distribution (CVLD) characteristics;

[0034] Figure 4 (a) is the measured value of the relative light intensity (RLI) of each leaf layer in the canopy light distribution characteristics, and (b) is the measured value of the light interception rate (LIR) of each leaf layer in the canopy light distribution characteristics;

[0035] Figure 5 This is a schematic diagram of the correlation between the measured values ​​of canopy light distribution and the leaf area index of each leaf layer;

[0036] Figure 6 (a) is a schematic diagram of the number of point clouds between each layer of the canopy in 2022-2023, (b) is a schematic diagram of the number of point clouds at each leaf layer and above in the canopy in 2022-2023, (c) is a schematic diagram of the number of point clouds between each layer of the canopy in 2023-2024, and (d) is a schematic diagram of the number of point clouds at each leaf layer and above in the canopy in 2023-2024;

[0037] Figure 7 Schematic diagram of the correlation between canopy light distribution parameters and each layer’s point cloud characteristics (CVPCD);

[0038] Figure 8 It is the accuracy map of the canopy light distribution parameter model constructed based on the point cloud features of each layer (CVPCD);

[0039] Fig. 9 These are the accuracy graphs of the leaf area index estimation model for each leaf layer at 0 days after anthesis (0DAA), 20 days after anthesis (20DAA), and 30 days after anthesis (30DAA). DETAILED DESCRIPTION

[0040] The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0041] The present invention provides a method for estimating the leaf area index of each leaf layer based on the light distribution of the wheat canopy. The specific process is as follows: Figure 1 As shown, the following steps are included:

[0042] Step 1: Use the DJI Mavic 3 multispectral drone to collect multispectral data.

[0043] The multispectral UAV used can obtain spectral information in four bands, namely green: 560±16nm, red: 650±16nm, red edge: 730±16nm, near infrared: 840±26nm. The flight plan and parameters were established using DJI Pilot 2 software, and the forward overlap and side overlap rates of the routes were maintained at 80% and 70%, respectively. The multispectral UAV flew at an altitude of 12m, with a lens angle of 90° vertically downward. Before the flight mission, three reflectors with predetermined reflectivity values ​​(25%, 50%, and 75%) were manually captured for subsequent radiation correction. The UAV equipped with a lidar flew at an altitude of 15 meters, with an oblique photography mode and a lens angle of 45°. Each flight required five routes around the experimental site. The two UAVs flew from 11 a.m. to 1 p.m. in clear sky and low wind conditions.

[0044] Step 2: Use DJI Terra software to stitch multispectral images and reconstruct the canopy 3D point cloud.

[0045] Before stitching the multispectral images, the calibration plate data was imported for radiation correction to improve the reliability of the output. After stitching, four single-band images were generated. The "Band Combination Tool" in ENVI (The Environment for Visualizing Images, a complete remote sensing image processing platform) software was used to create a multispectral image of the entire test area from the four single-band images. When reconstructing the canopy 3D point cloud, the point cloud density was configured to the maximum value to ensure high-quality results of the point cloud reconstruction and activate point cloud accuracy optimization. Cloudcompare software was used for point cloud denoising and downsampling.

[0046] Step 3: Use ArcGIS software to generate shape files for each plot, use ENVI software to crop multispectral images, and use PCM (Point Cloud Magic) software to crop canopy point clouds to obtain multispectral images and three-dimensional point clouds for each plot.

[0047] Step 4: Import the multispectral image of each plot into MATLAB software to extract the vegetation index of each plot. The specific vegetation index extracted is shown in Table 1 below.

[0048] Table 1 Vegetation indices extracted from multispectral images

[0049]

[0050]

[0051] Note: Nir, RedE, R, and G are the reflectances of near-infrared band, red edge band, red band, and green band, respectively.

[0052] Step 5: Count the number of point clouds in different layers.

[0053] Import the acquired point clouds of different plots into the PCM software to obtain the canopy height information. Normalize the number of point clouds of different plots according to the bottom area of ​​the minimum bounding box of the point cloud. Divide the canopy point cloud into 4 layers according to the measurement position of the vertical light distribution of the canopy, namely, the flag leaf and above layer (L1U), the second leaf layer-flag leaf layer (L2-L1), the third leaf layer-the second leaf layer (L3-L2), the third leaf layer and below (L3B), and count the number of point clouds in different layers. The specific flow chart for obtaining the number of point clouds of different vertical layers of the canopy is shown in the figure. Figure 2 As shown in Figure 6 As shown in (a) to (d) in the figure.

[0054] Step 6: Based on the quantitative characteristics of point clouds at different layers, a canopy light distribution estimation model is constructed using a machine learning algorithm ( Figure 8 ) and the coefficient of determination (R 2 ), root mean square error (RMSE) and mean absolute error (MAE) to evaluate the performance of the model.

[0055] The photosynthetically active radiation (PAR) of the top of the canopy (20 cm above the top of the ear), the flag leaf (L1), the second leaf (L2), the third leaf (L3), the fourth leaf (L4) and the bottom of the canopy was measured using the SUNSCAN canopy analyzer produced by the British Delta Company. Figure 3 The photosynthetically active radiation interception rate (LIR) of each leaf layer and the relative light intensity (RLI) of each leaf layer were calculated according to the following formula:

[0056]

[0057] Among them, PAR n is the photosynthetically active radiation of each leaf layer (μmol·m -2 ·s -1 ), TPAR is the photosynthetically active radiation at the top of the canopy (μmol·m -2 ·s -1 ); n represents different leaf layers. When n is the flag leaf layer, n-1 is the second leaf layer from the bottom, when n is the second leaf layer from the bottom, n-1 is the third leaf layer from the bottom, and when n is the third leaf layer from the bottom, n-1 is the bottom of the canopy.

[0058] With the number of canopy point clouds as input parameters and the canopy light distribution characteristics as output parameters, a canopy light distribution estimation model was constructed through a machine learning algorithm; among which, the canopy light distribution characteristics included the photosynthetic active radiation interception rate and the relative light intensity.

[0059] The measured value of wheat canopy light distribution (CVLD) is as follows Figure 4 As shown in (a) and (b) in R 2 , RMSE, and MAE values ​​are calculated as follows:

[0060]

[0061] Among them, N is the number of samples, the measured value and the predicted value are y i and is the average measurement value.

[0062] Step 7: Based on the estimated canopy light distribution characteristics and vegetation index, a machine learning algorithm is used to build an LAI estimation model for different leaf layers, and R 2 , RMSE and MAE were used to evaluate the performance of the model.

[0063] Wheat leaves were divided into flag leaf (L1), leaf 2 (L2), leaf 3 (L3), leaf 4 (L4), and leaf 5 (L5). The length and width of each leaf were measured with a ruler, and the leaf area of ​​each leaf was calculated using the following formula:

[0064] Leaf area (cm 2 ) = blade length (cm) × blade width (cm) × 0.83 (6)

[0065] Among them, the blade length refers to the distance from the base to the tip of the blade, and the blade width refers to the widest part of the blade.

[0066] The LAI of each leaf layer is calculated by the following formula:

[0067] LAI = leaf area of ​​each leaf layer per single stem (m 2 )×total number of single stems (7)

[0068] With the estimated canopy light distribution characteristics and vegetation index as input parameters and the estimated leaf area index values ​​of each leaf layer as output parameters, a machine learning algorithm was used to construct the LAI estimation model for different leaf layers.

[0069] The leaves of wheat at 0 days after anthesis (0DAA), 20 days after anthesis (20DAA), and 30 days after anthesis (30DAA) were selected to obtain the measured values ​​of leaf area index of each leaf layer and the estimated values ​​of the model. Fig. 9 As shown, the horizontal axis is the measured value and the vertical axis is the estimated value.

[0070] In order to evaluate the application effect of the method of the present invention in actual production and between years, the invention team conducted two seasons of winter wheat experiments in Zhenjiang, Jiangsu from 2022 to 2024, designed different cultivars, planting densities and nitrogen fertilizer gradients, and constructed different population structures. By investigating the leaf area index of each leaf layer during the filling period of winter wheat, canopy light distribution, multispectral data and three-dimensional point cloud data, the strong correlation between canopy light distribution and leaf area index of each leaf layer was found ( Figure 5 ), through the strong correlation between different layer point clouds and canopy light distribution ( Figure 7 ) to estimate the canopy light distribution, and combined the estimated canopy light distribution parameters and vegetation index to estimate the leaf area index of each leaf layer of winter wheat using a machine learning algorithm. It was found that R 2 The highest value can reach 0.92, and the lowest RMSE and MAE are 0.06 and 0.02 respectively. It can be seen that the leaf area index estimation model for each leaf layer proposed in the present invention can estimate the leaf area index of each leaf layer in a large area of ​​the field at low cost and high throughput.

[0071] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention or some parts of the embodiments.

[0072] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A method for estimating leaf area index of each leaf layer based on wheat canopy light distribution, characterized in that: The steps include: S1, collecting multispectral data and wheat canopy 3D point cloud data by UAV; S2, multispectral image stitching and canopy 3D point cloud reconstruction using DJI Terra software; S3, using ArcGIS software to generate shape files for each plot, using ENVI software to crop multispectral images, using PCM software to crop canopy point clouds, and obtaining multispectral images and three-dimensional point clouds for each plot; S4, import the multispectral image of each plot into MATLAB software to extract the vegetation index of each plot; S5, counting the number of point clouds at different layers of the canopy point cloud; S6, based on the quantitative characteristics of point clouds at different layers, a canopy light distribution estimation model was constructed using a machine learning algorithm; S7, based on the estimated canopy light distribution characteristics and vegetation index, a machine learning algorithm was used to construct an LAI estimation model for different leaf layers.

2. The method for estimating leaf area index of each leaf layer based on wheat canopy light distribution according to claim 1, characterized in that: In step S5, the steps for counting the number of point clouds at different layers of the canopy point cloud are as follows: S51, importing the obtained point clouds of different plots into PCM software to obtain canopy height information; S51, normalizing the number of point clouds in different cells according to the bottom area of ​​the minimum bounding box of the point cloud; S53, according to the measurement position of the canopy vertical light distribution, the canopy point cloud is divided into 4 layers, namely: flag leaf and above layer, 2nd leaf layer-flag leaf layer, 3rd leaf layer-2nd leaf layer, 3rd leaf and below layer; S54, counting the number of point clouds at different layers in step S53.

3. The method for estimating leaf area index of each leaf layer based on wheat canopy light distribution according to claim 1, characterized in that: In the canopy light distribution estimation model, the number of canopy point clouds is used as an input parameter, and the canopy light distribution characteristics are used as output parameters, wherein the canopy light distribution characteristics include photosynthetically active radiation interception rate and relative light intensity; The calculation formulas for the photosynthetically active radiation interception rate of each leaf layer and the relative light intensity of each leaf layer are as follows: Among them, LIR represents the photosynthetically active radiation interception rate, RLI represents the relative light intensity of each leaf layer; PAR n is the photosynthetically active radiation of each leaf layer, in μmol·m -2 ·s -1 ; TPAR is the photosynthetically active radiation at the top of the canopy, in μmol·m -2 ·s -1 ; n represents different leaf layers.

4. The method for estimating leaf area index of each leaf layer based on wheat canopy light distribution according to claim 3, characterized in that: In the LAI estimation model for different leaf layers, the estimated canopy light distribution characteristics and vegetation index are used as input parameters, and the estimated leaf area index value of each leaf layer is used as output parameter; The LAI calculation formula for each leaf layer is as follows: LAI = leaf area of ​​each leaf layer per single stem × total number of single stems The leaf area of ​​each leaf is calculated as follows: Leaf area = leaf length × leaf width × 0.83 Among them, the blade length refers to the distance from the base to the tip of the blade, and the blade width refers to the widest part of the blade.

Images