A method for inversion of grassland aboveground biomass based on vegetation index reconstruction using relative height

Abstract

The invention discloses a grassland aboveground biomass inversion method based on relative height reconstruction of vegetation index, which comprises the following steps: step 1: using a multispectral and laser radar sensor carried by an unmanned aerial vehicle to obtain an unmanned aerial vehicle remote sensing image of a grassland to be monitored; step 2: calculating the vegetation index of a target area through the unmanned aerial vehicle multispectral image; obtaining a digital surface model DSM of the target area through the unmanned aerial vehicle laser radar data; step 3: obtaining the grassland vegetation canopy height, vegetation coverage and aboveground biomass in the study area on the spot; step 4: performing threshold segmentation on the vegetation index to obtain the specific positions of vegetation and bare soil in the grassland, and obtaining the grassland vegetation canopy relative height based on DSM; step 5: constructing a reconstructed vegetation index using the canopy relative height in combination with the vegetation index; step 6: inverting the grassland aboveground biomass based on the reconstructed vegetation index; the invention develops a grassland vegetation canopy relative height model, and constructs a new grassland aboveground biomass estimation method based on the reconstructed vegetation index.

Description

Technical Field

[0001] The invention relates to the field of quantitative remote sensing applications, and in particular to a grassland aboveground biomass inversion method based on relative height reconstruction of vegetation index. Background Art

[0002] The aboveground biomass of natural grasslands is one of the key indicators reflecting the health of grassland ecosystems. Its accurate monitoring is crucial for the rational use and management of grassland resources and the optimization of animal husbandry production. UAVs can quickly obtain different vegetation parameters by carrying sensors such as multispectral, hyperspectral and lidar. Compared with satellite remote sensing, drones can provide higher spatial resolution data, and with flexible operation methods, they can reduce the impact of factors such as spatial heterogeneity and vegetation growth period on the estimation of aboveground biomass.

[0003] At present, the inversion method of aboveground biomass in natural grassland is mainly based on empirical model, which establishes mathematical relationship between ground observation data and remote sensing parameters. The most commonly used remote sensing parameter is vegetation index. This method is simple and practical. Under the premise of appropriate ground measured data and remote sensing parameters, while paying attention to the phase and spatial scale, a more ideal estimation model can be constructed. However, vegetation index is prone to saturation in areas with high vegetation density, resulting in underestimation of aboveground biomass.

[0004] Natural grassland vegetation has relatively complex vertical structural information. Introducing three-dimensional structural information plays an important role in accurately estimating aboveground biomass. Relevant studies have shown that the inversion accuracy can be improved by adding information such as vegetation coverage and vegetation canopy height. Some studies have combined natural grassland three-dimensional spatial information to carry out remote sensing inversion of aboveground biomass, but most of them use vegetation index, vegetation height and vegetation coverage as independent input variables, fail to organically combine three-dimensional spatial information, lack in-depth theoretical understanding, and reduce the accuracy and interpretability of the model. Summary of the invention

[0005] The purpose of the present invention is to overcome the above-mentioned deficiencies in the prior art and to propose a grassland aboveground biomass inversion method based on relative height reconstruction of vegetation index.

[0006] The technical solution adopted by the present invention is as follows:

[0007] The grassland aboveground biomass inversion method based on relative height reconstruction vegetation index includes the following steps:

[0008] Step 1: Use a multispectral sensor and a lidar sensor mounted on a drone to obtain a drone remote sensing image of the grassland to be monitored;

[0009] Step 2: Calculate the vegetation index of the target area through the multispectral image of the UAV; obtain the digital surface model DSM of the target area through the UAV lidar data;

[0010] Step 3: Obtain the vegetation canopy height, vegetation coverage and aboveground biomass of the sample plots in the study area on site;

[0011] Step 4: Obtain the vegetation coverage information of the sample plot by threshold segmentation of the vegetation index, extract the specific location of grassland vegetation and bare soil, and obtain the relative height of the vegetation canopy of the sample plot based on the digital surface model DSM;

[0012] Step 5: Generate a reconstructed vegetation index based on the relative height of the vegetation canopy in the target area sample plot and the vegetation index in the target area;

[0013] Step 6: Use the reconstructed vegetation index of the sample area and the actual aboveground biomass data to build a linear regression model, evaluate the model accuracy, and select the optimal reconstructed vegetation index to estimate the aboveground biomass of the grassland.

[0014] Preferably, step 2 comprises the following steps:

[0015] Step 2.1: Preprocess the multispectral data acquired by the drone, including radiation correction, geometric correction, stitching and cropping;

[0016] Step 2.2: Calculate six vegetation indices using UAV multispectral images, the six vegetation indices are normalized difference vegetation index NDVI, green normalized difference vegetation index GNDVI, ratio vegetation index RVI, difference vegetation index DVI, soil adjusted vegetation index SAVI, and optimized soil adjusted vegetation index OSAVI; wherein, threshold division is performed on the optimized soil adjusted vegetation index OSAVI to extract vegetation coverage information;

[0017] Step 2.3: Preprocessing the UAV laser radar data, wherein the preprocessing includes point cloud reconstruction, point cloud classification and point cloud editing;

[0018] Step 2.4: Generate a digital surface model (DSM) of the study area based on the dense point cloud data obtained by the UAV lidar, and set the image spatial resolution to the same as the multispectral data.

[0019] Preferably, in step 3, a grassland plot is selected in the study area, and the plot size is 0.5m×0.5m; the vegetation canopy height of the plot is manually measured by a nine-point sampling method, and the average value is taken; the vegetation coverage is obtained by photographing with a camera and visual interpretation; the aboveground biomass of the grassland is obtained by obtaining all plants in the plot by mowing flush with the ground, and weighing them to obtain the fresh weight, thereby obtaining the aboveground biomass of the grassland.

[0020] Preferably, step 4 includes the following sub-steps:

[0021] Step 4.1: Perform threshold segmentation on OSAVI and observe that the bare soil distribution is close to the actual situation, and classify bare soil and vegetation;

[0022] Step 4.2: Use the bare soil mask layer to extract the minimum value of the digital surface model DSM layer;

[0023] Step 4.3: Use the formula to obtain the relative height of the vegetation canopy of the sample plot. The formula is as follows:

[0024] CRHM(i,j)=DSM(i,j)-MIN(DSM soil )

[0025] Where CRHM(i,j) is the relative height of the canopy at position (i,j) in the sample plot, DSM(i,j) is the DSM of the pixel at position (i,j) in the sample plot, and DSM soil is the DSM utilizing the bare soil mask area.

[0026] Preferably, step 5 includes the following sub-steps:

[0027] Step 5.1: Resample the vegetation index image and the plot vegetation canopy relative height data to a uniform spatial resolution;

[0028] Step 5.2: Normalize the plot vegetation canopy relative height data to obtain normalized canopy relative height data;

[0029] Step 5.3: Multiply the vegetation index image and the normalized canopy relative height data to obtain the reconstructed vegetation index. The specific formula is as follows:

[0030]

[0031] In the formula, ReVI represents the reconstructed vegetation index, VI represents the original vegetation index, CRHM represents the canopy relative height based on the DSM model, and CRHM max Indicates the maximum value based on CRHM, CRHM min Indicates the minimum value based on CRHM.

[0032] Preferably, the specific steps of step 6 are as follows:

[0033] Step 6.1: Using the aboveground biomass of the sample plot as the dependent variable, and the vegetation index and reconstructed vegetation index as the independent variables, the aboveground biomass estimation model was established using the linear regression method;

[0034] Step 6.2: Divide the aboveground biomass estimation model into training samples and validation samples with a ratio of 7:3, and repeat the cycle 100 times;

[0035] Step 6.3: Calculate the coefficient of determination R 2 The root mean square error RMSE is used to evaluate the model effect, and the determination coefficient R 2 The calculation formula of root mean square error RMSE is as follows:

[0036]

[0037]

[0038] In the formula, n represents the number of samples, y i represents the observed value, represents the predicted value, represents the mean of the observed values;

[0039] Step 6.4: Compare the prediction effects of vegetation index and reconstructed vegetation index, and select appropriate features and models to estimate grassland aboveground biomass based on the aboveground biomass fitting effect.

[0040] The beneficial effects of the grassland aboveground biomass inversion method based on relative height reconstruction vegetation index of the present invention are as follows:

[0041] The present invention utilizes the canopy relative height model to achieve more accurate acquisition of grassland vegetation height information, and overcomes the limitation that the vegetation index only contains planar information by reconstructing the vegetation index, thereby more accurately estimating the aboveground biomass of the grassland. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a flow chart of the grassland aboveground biomass inversion method based on relative height reconstruction vegetation index of the present invention;

[0043] Figure 2 It is a schematic diagram of the relative height acquisition process of the present invention. DETAILED DESCRIPTION

[0044] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0045] The grassland aboveground biomass inversion method based on relative height reconstruction vegetation index includes the following steps:

[0046] Step 1: Use a multispectral sensor and a lidar sensor mounted on a drone to obtain a drone remote sensing image of the grassland to be monitored;

[0047] Step 2: Calculate the vegetation index of the target area through the multispectral image of the UAV; obtain the digital surface model DSM of the target area through the UAV lidar data;

[0048] Step 3: Obtain the vegetation canopy height, vegetation coverage and aboveground biomass of the sample plots in the study area on site;

[0049] Step 4: Obtain the vegetation coverage information of the sample plot by threshold segmentation of the vegetation index, extract the specific location of grassland vegetation and bare soil, and obtain the relative height of the vegetation canopy of the sample plot based on the digital surface model DSM;

[0050] Step 5: Generate a reconstructed vegetation index based on the relative height of the vegetation canopy in the target area sample plot and the vegetation index in the target area;

[0051] Step 6: Use the reconstructed vegetation index of the sample area and the actual aboveground biomass data to build a linear regression model, evaluate the model accuracy, and select the optimal reconstructed vegetation index to estimate the aboveground biomass of the grassland.

[0052] Step 2 of this embodiment includes the following steps:

[0053] Step 2.1: Preprocess the multispectral data acquired by the drone, including radiation correction, geometric correction, stitching and cropping;

[0054] Step 2.2: Calculate six vegetation indices using UAV multispectral images, the six vegetation indices are normalized difference vegetation index NDVI, green normalized difference vegetation index GNDVI, ratio vegetation index RVI, difference vegetation index DVI, soil adjusted vegetation index SAVI, and optimized soil adjusted vegetation index OSAVI; wherein, threshold division is performed on the optimized soil adjusted vegetation index OSAVI to extract vegetation coverage information;

[0055] Step 2.3: Preprocessing the UAV laser radar data, wherein the preprocessing includes point cloud reconstruction, point cloud classification and point cloud editing;

[0056] Step 2.4: Generate a digital surface model (DSM) of the study area based on the dense point cloud data obtained by the UAV lidar, and set the image spatial resolution to the same as the multispectral data.

[0057] In step 3 of this implementation plan, a grassland plot is selected in the study area, and the plot size is 0.5m×0.5m; the vegetation canopy height of the plot is manually measured by the nine-point sampling method, and the average value is taken; the vegetation coverage is obtained by photographing with a camera and visual interpretation; the aboveground biomass of the grassland is obtained by obtaining all the plants in the plot by mowing at the ground level, and weighing them to obtain the fresh weight, thereby obtaining the aboveground biomass of the grassland.

[0058] Step 4 of this embodiment includes the following sub-steps:

[0059] Step 4.1: Perform threshold segmentation on OSAVI and observe that the bare soil distribution is close to the actual situation, and classify bare soil and vegetation;

[0060] Step 4.2: Use the bare soil mask layer to extract the minimum value of the digital surface model DSM layer;

[0061] Step 4.3: Use the formula to obtain the relative height of the grassland sample vegetation canopy. The formula is as follows:

[0062] CRHM(i,j)=DSM(i,j)-MIN(DSM soil )

[0063] Where CRHM(i,j) is the relative height of the canopy at position (i,j) in the sample plot, DSM(i,j) is the DSM of the pixel at position (i,j) in the sample plot, and DSM soil is the DSM utilizing the bare soil mask area.

[0064] Step 5 of this embodiment includes the following sub-steps:

[0065] Step 5.1: Resample the vegetation index image and grassland sample canopy relative height data to a uniform spatial resolution;

[0066] Step 5.2: Normalize the grassland sample plot canopy relative height data to obtain normalized canopy relative height data;

[0067] Step 5.3: Multiply the vegetation index image and the normalized canopy relative height data to obtain the reconstructed vegetation index. The specific formula is as follows:

[0068]

[0069] In the formula, ReVI represents the reconstructed vegetation index, VI represents the original vegetation index, CRHM represents the canopy relative height based on the DSM model, and CRHM max Indicates the maximum value based on CRHM, CRHM min Indicates the minimum value based on CRHM.

[0070] The specific steps of step 6 of this implementation scheme are as follows:

[0071] Step 6.1: Using the aboveground biomass of the sample plot as the dependent variable, and the vegetation index and reconstructed vegetation index as the independent variables, the aboveground biomass estimation model was established using the linear regression method;

[0072] Step 6.2: Divide the aboveground biomass estimation model into training samples and validation samples with a ratio of 7:3, and repeat the cycle 100 times;

[0073] Step 6.3: Calculate the coefficient of determination R 2 The root mean square error RMSE is used to evaluate the model effect, and the determination coefficient R 2 The calculation formula of root mean square error RMSE is as follows:

[0074]

[0075]

[0076] In the formula, n represents the number of samples, y i represents the observed value, represents the predicted value, represents the mean of the observed values;

[0077] Step 6.4: Compare the prediction effects of vegetation index and reconstructed vegetation index, and select appropriate features and models to estimate grassland aboveground biomass based on the aboveground biomass fitting effect.

[0078] When this embodiment is implemented, the present invention provides a grassland aboveground biomass inversion method based on relative height reconstruction vegetation index, such as Figure 1 As shown, the specific steps are as follows:

[0079] Step 1: Use a multispectral sensor and a lidar sensor mounted on a drone to obtain a drone remote sensing image of the grassland to be monitored;

[0080] Step 2: Calculate the vegetation index of the target area through the multispectral image of the UAV; obtain the digital surface model DSM of the target area through the UAV lidar data;

[0081] Step 3: Obtain the vegetation canopy height, vegetation coverage and aboveground biomass of the sample plots in the study area on site;

[0082] Step 4: Obtain the vegetation coverage information of the sample plot by threshold segmentation of the vegetation index, extract the specific location of grassland vegetation and bare soil, and obtain the relative height of the vegetation canopy of the sample plot based on the digital surface model DSM;

[0083] Step 5: Generate a reconstructed vegetation index based on the relative height of the vegetation canopy and the vegetation index of the target area sample plot;

[0084] Step 6: Use the reconstructed vegetation index of the sample area and the actual aboveground biomass data to build a linear regression model, evaluate the model accuracy, and select the optimal reconstructed vegetation index to estimate the aboveground biomass of the grassland. In a specific implementation, step 1 also includes:

[0085] The experiment was conducted on mowed grassland in natural grassland. The experimental area was about 120 meters × 120 meters. Targets were set at four corner points as ground control points for later image orthorectification. The UAV flight time was selected from the period of sufficient and stable light from 12 noon to 3 pm. The flight altitude was 50 meters when the UAV multispectral data was acquired, and the image spatial resolution was 2 cm. The flight altitude was 30 meters when the UAV lidar data was acquired, and the average point cloud density was not less than 3000 points / m 2 ;

[0086] In a specific implementation, step 2 further includes:

[0087] Step 2.1: Preprocess the multispectral data acquired by the drone, including radiation correction, geometric correction, stitching and cropping;

[0088] Step 2.2: Use UAV multispectral images to calculate six vegetation indices, including the normalized difference vegetation index NDVI, the green normalized difference vegetation index GNDVI, the ratio vegetation index RVI, the difference vegetation index DVI, the soil adjusted vegetation index SAVI, and the optimized soil adjusted vegetation index OSAVI; the optimized soil adjusted vegetation index OSAVI is thresholded to extract vegetation coverage information;

[0089] Step 2.3: Preprocess the UAV lidar data, including point cloud reconstruction, point cloud classification and point cloud editing;

[0090] Step 2.4: Generate DSM data of the study area based on the dense point cloud data acquired by the lidar, and set the image spatial resolution to the same as the multispectral data, which is uniformly set to 2 cm here;

[0091] In a specific implementation, step 3 further includes:

[0092] Grassland plots with a size of 0.5m×0.5m were selected in the study area. The vegetation height was measured manually by the nine-point sampling method and the average value was taken. The vegetation coverage was obtained by photographing with a camera and visually interpreted. All plants in the plot were obtained by mowing at the ground level and weighed to obtain the fresh weight, thereby obtaining the aboveground biomass of the plot.

[0093] In a specific implementation, step 4 further includes:

[0094] Step 4.1: Threshold OSAVI and observe that the bare soil distribution is close to the actual situation, and classify bare soil and vegetation;

[0095] Step 4.2: Use the bare soil mask layer to extract the minimum value of the DSM layer;

[0096] Step 4.3: Use the formula to extract the relative height of the grassland vegetation canopy for each pixel in the sample plot. The formula is as follows:

[0097] CRHM(i,j)=DSM(i,j)-MIN(DSM soil )

[0098] Where CRHM(i,j) is the relative height of the canopy at position (i,j) in the sample plot, DSM(i,j) is the DSM of the pixel at position (i,j) in the sample plot, and DSM soil To utilize the DSM of the bare soil mask area;

[0099] In a specific implementation, the OSAVI coverage extraction threshold for mowed grassland is 0.48.

[0100] In a specific implementation, step 5 further includes:

[0101] Step 5.1: Resample the vegetation index image and canopy relative height data to a uniform spatial resolution;

[0102] Step 5.2: Normalize the canopy relative height data;

[0103] Step 5.3: Multiply the vegetation index image and the normalized canopy relative height data to obtain the reconstructed vegetation index. The specific formula is as follows:

[0104]

[0105] In the formula, ReVI represents the reconstructed vegetation index, VI represents the original vegetation index, CRHM represents the canopy relative height based on the DSM model, and CRHM max 、CRHM min Indicates the minimum and maximum values ​​based on CRHM.

[0106] In a specific implementation, step 6 further includes:

[0107] Step 6.1: Using the aboveground biomass of the sample plot as the dependent variable, and the original vegetation index and the reconstructed vegetation index as the independent variables, the linear regression method was used to establish the grassland aboveground biomass estimation model;

[0108] Step 6.2: The ratio of training samples to validation samples is 7:3 to avoid selection errors, and the cycle is repeated 100 times;

[0109] Step 6.3: Calculate the coefficient of determination R 2 And root mean square error RMSE evaluates the model effect, R 2 The calculation formula of RMSE is as follows:

[0110]

[0111]

[0112] In the formula, n represents the number of samples, y i represents the observed value, represents the predicted value, represents the mean of the observed values;

[0113] Step 6.4: Compare the prediction results of vegetation index and reconstructed vegetation index, and select appropriate features and models to estimate grassland aboveground biomass based on the aboveground biomass fitting results.

[0114] In this paper, an experiment was conducted on a certain plot of mowed grassland to obtain the grassland vegetation canopy height, vegetation coverage and aboveground biomass, and the aboveground biomass of the grassland was estimated using the proposed algorithm. The results showed that the correlation between CRHM and the measured height was R 2 The value of NDVI linear regression model R is 0.61, which is higher than other height extraction parameters and can more accurately obtain grassland vegetation height information. 2 The R value of the reconstructed NDVI linear regression model is 0.39. 2 The value is 0.69, which overcomes the limitation that the vegetation index only contains planar information and effectively improves the estimation accuracy of grassland aboveground biomass.

Claims

1. A grassland aboveground biomass inversion method based on relative height reconstruction of vegetation index, characterized in that: The following steps are involved: Step 1: Use a multispectral sensor and a lidar sensor mounted on a drone to obtain a drone remote sensing image of the grassland to be monitored; Step 2: Calculate the vegetation index of the target area through the multispectral image of the UAV; obtain the digital surface model DSM of the target area through the UAV lidar data; Step 3: Obtain the vegetation canopy height, vegetation coverage and aboveground biomass of the sample plots in the study area on site; Step 4: Obtain the vegetation coverage information of the sample plot by threshold segmentation of the vegetation index, extract the specific location of grassland vegetation and bare soil, and obtain the relative height of the vegetation canopy of the sample plot based on the digital surface model DSM; Step 5: Generate a reconstructed vegetation index based on the relative height of the vegetation canopy in the target area sample plot and the vegetation index in the target area; Step 6: Use the reconstructed vegetation index of the sample area and the actual aboveground biomass data to build a linear regression model, evaluate the model accuracy, and select the optimal reconstructed vegetation index to estimate the aboveground biomass of the grassland; The step 2 comprises the following steps: Step 2.1: Preprocess the multispectral data acquired by the drone, including radiation correction, geometric correction, stitching and cropping; Step 2.2: Calculate six vegetation indices using UAV multispectral images, the six vegetation indices are normalized difference vegetation index NDVI, green normalized difference vegetation index GNDVI, ratio vegetation index RVI, difference vegetation index DVI, soil adjusted vegetation index SAVI, and optimized soil adjusted vegetation index OSAVI; wherein, threshold division is performed on the optimized soil adjusted vegetation index OSAVI to extract vegetation coverage information; Step 2.3: Preprocessing the UAV laser radar data, wherein the preprocessing includes point cloud reconstruction, point cloud classification and point cloud editing; Step 2.4: Generate the digital surface model DSM of the study area based on the dense point cloud data obtained by the UAV lidar, and set the image spatial resolution to the same as the multispectral data; In step 3, a grassland sample plot is selected in the study area, and the sample plot size is 0.5m×0.5m; the vegetation canopy height of the sample plot is manually measured by the nine-point sampling method, and the average value is taken; the vegetation coverage is obtained by taking pictures with a camera and visually interpreting; the aboveground biomass of the grassland is obtained by obtaining all plants in the sample plot by mowing them flush with the ground, and weighing them to obtain the fresh weight, thereby obtaining the aboveground biomass of the grassland; The step 4 includes the following sub-steps: Step 4.1: Perform threshold segmentation on OSAVI and observe that the bare soil distribution is close to the actual situation, and classify bare soil and vegetation; Step 4.2: Use the bare soil mask layer to extract the minimum value of the digital surface model DSM layer; Step 4.3: Use the formula to obtain the relative height of the vegetation canopy of the sample plot. The formula is as follows: CRHM(i,j)=DSM(i,j)-MIN(DSM soil ) Where CRHM(i, j) is the relative height of the canopy at position (i, j) in the sample plot, DSM(i, j) is the DSM of the pixel at position (i, j) in the sample plot, and DSM soil To utilize the DSM of the bare soil mask area; The step 5 comprises the following sub-steps: Step 5.1: Resample the vegetation index image and the plot vegetation canopy relative height data to a uniform spatial resolution; Step 5.2: Normalize the plot vegetation canopy relative height data to obtain normalized canopy relative height data; Step 5.3: Multiply the vegetation index image and the normalized canopy relative height data to obtain the reconstructed vegetation index. The specific formula is as follows: In the formula, ReVI represents the reconstructed vegetation index, VI represents the original vegetation index, CRHM represents the canopy relative height based on the DSM model, and CRHM max Indicates the maximum value based on CRHM, CRHM min represents the minimum value based on CRHM; The specific steps of step 6 are as follows: Step 6.1: Using the aboveground biomass of the sample plot as the dependent variable, and the vegetation index and reconstructed vegetation index as the independent variables, the aboveground biomass estimation model was established using the linear regression method; Step 6.2: Divide the aboveground biomass estimation model into training samples and validation samples with a ratio of 7:3, and repeat the cycle 100 times; Step 6.3: Calculate the coefficient of determination R 2 The root mean square error RMSE is used to evaluate the model effect, and the determination coefficient R 2 The calculation formula of root mean square error RMSE is as follows: In the formula, n represents the number of samples, y i represents the observed value, Represents the predicted value, y i represents the mean of the observed values; Step 6.4: Compare the prediction effects of vegetation index and reconstructed vegetation index, and select appropriate features and models to estimate grassland aboveground biomass based on the aboveground biomass fitting effect.

Images