Method for constructing soil fertility index comprehensive evaluation model based on multiple soil indexes

Abstract

The invention provides a method for constructing a soil fertility index comprehensive evaluation model based on multiple soil indexes, and the method comprises the steps: (1) collecting soil sample image data of different regions, and carrying out the preprocessing of the data, so as to obtain a panoramic looking-around soil profile hyperspectral image; (2) extracting soil sample data and standard reflectivity plate data; (3) carrying out radiation correction and preprocessing on the soil sample data and the standard reflectivity plate data; (4) carrying out characteristic wavelength extraction on the spectral reflectivity data of the soil samples at different depths; (5) respectively measuring parameter data of the soil samples at different depths; (6) constructing a multi-parameter inversion model under different depths; (7) establishing a soil fertility comprehensive evaluation model; and (8) obtaining color distribution diagrams of different soil parameters and soil fertility indexes. According to the method, the spectrum data of the soil profiles at different depths can be quickly acquired in a low-loss manner, so that the soil index data is inverted, and the soil fertility evaluation efficiency is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of soil fertility evaluation, and in particular to a method for constructing a soil fertility index comprehensive evaluation model based on multiple soil indexes. BACKGROUND

[0002] Soil health refers to the ability of soil to sustain biological productivity, maintain air and water environmental quality, and promote plant, animal and human health as a living system within the scope of an ecosystem and land use. Soil fertility refers to the ability of soil to provide air, temperature, nutrients and non-toxic substances required for plant growth in a timely and coordinated manner. Soil fertility is a comprehensive reflection of various physical and chemical properties of soil, and is the main function and essential attribute of soil. Soil fertility is the result of the comprehensive action of the internal material, structure and physical and chemical properties of soil and the external environmental conditions.

[0003] Traditional soil analysis relies on laboratory chemical methods, which destroys the sample and is time-consuming and labor-intensive. Conventional soil profile spectral observation has long relied on laboratory sampling and analysis, and a few studies have made in-situ observations on soil by digging large square profiles (such as 2.2m long x 0.8m wide x 1.2m deep) for manual spectral measurement. This method has the problems of high profile excavation cost and low efficiency, and greatly destroys the soil structure, and cannot realize continuous dynamic monitoring.

[0004] Currently, soil parameter inversion based on spectral data is mostly single-index inversion, and the parameter inversion accuracy is not high. Due to the differences in physical and chemical properties of soil at different depths, the spectral characteristics are offset. Due to the lack of deep profile data, there are few studies on the changes in soil parameter content at different depths. SUMMARY

[0005] In view of the technical problems in the above background art, the present application provides a method for constructing a soil fertility index comprehensive evaluation model based on multiple soil indexes, which is reasonable in concept and can obtain soil profile spectral data at different depths by low-loss and rapid acquisition, thereby inverting soil index data such as total nitrogen, organic matter, calcium carbonate and water-soluble salt, and combining the calculation of the proposed soil fertility index (SFI) to greatly improve the soil fertility evaluation efficiency.

[0006] To solve the above technical problems, the present application provides a method for constructing a soil fertility index comprehensive evaluation model based on multiple soil indexes, which specifically includes the following steps: (1) Collecting soil sample image data in different regions, and then pre-processing the image data to obtain panoramic soil profile hyperspectral images; (2) According to the obtained soil profile hyperspectral image, different regions are divided according to the soil depth, and manual screening sampling is respectively performed to screen out abnormal regions and standard reflectance plate regions, then the data of the abnormal regions are removed, the soil sample data in each depth range is extracted, and the spectral data of the standard reflectance plate region is homogenized, and the standard reflectance plate data is extracted; (3) The extracted soil sample data and standard reflectance plate data are radiometrically corrected and converted into radiance data, then the soil sample radiance data is converted into soil sample spectral reflectance data based on the standard reflectance plate data, and the soil sample spectral reflectance data is preprocessed; (4) The soil sample spectral reflectance data at different depths is extracted as the input quantity of the multi-parameter inversion model; (5) The parameter data of the soil sample at different depths is measured by a chemical method as the output quantity of the multi-parameter inversion model training; (6) A deep learning network structure is designed and network training is performed to obtain a multi-parameter inversion model at different depths, and the multi-parameter inversion model is used for soil multi-parameter spectral quantitative inversion; (7) A soil fertility comprehensive evaluation model is established for each parameter data at different depths in different regions; (8) Each pixel point of the hyperspectral image is subjected to multi-parameter inversion and soil fertility index calculation through the established inversion algorithm model and soil fertility comprehensive evaluation model, and different soil parameters and soil fertility index color distribution maps are obtained.

[0007] The method for constructing the soil fertility index comprehensive evaluation model based on multiple soil indicators, wherein the specific process of step (1) is: using the completed spectral sensing module and the fiber image transmission beam to synthesize an imaging spectral equipment, and collecting sample spectral data for different regional soil; then, for the selected soil sampling area, a pit position for soil sample collection is dug, a standard reflectance plate is placed at the pit opening, and the imaging spectral equipment is placed at the pit opening and moved downward to push scan and obtain image data; the image data is sequentially subjected to a wavelet threshold denoising algorithm to remove honeycomb artifacts, then based on a Brown-Conrady model, image edge distortion is eliminated, image horizontal and vertical splicing is performed by using a cylindrical coordinate system projection, and a panoramic soil profile hyperspectral image is obtained.

[0008] The method for constructing the soil fertility index comprehensive evaluation model based on multiple soil indicators, wherein: in step (3), the extracted soil sample data and standard reflectance plate data are subjected to radiometric correction, and the original digital number (DN value) is converted into radiance value by using the calibration data measured in the laboratory, so as to more accurately reflect the radiation characteristics of the soil.

[0009] The method for constructing the soil fertility index comprehensive evaluation model based on multiple soil indicators, wherein the pre-processing of the soil sample spectral reflectance data in step (3) is specifically Savitzky-Golay filtering and multivariate scatter correction respectively; the Savitzky-Golay filtering is to smooth the data by local polynomial fitting of the soil sample spectral reflectance data to remove noise; and the multivariate scatter correction is used to reduce the changes caused by spectral scattering and eliminate the spectral changes caused by the differences in the physical properties of the samples.

[0010] The method for constructing the soil fertility index comprehensive evaluation model based on multiple soil indicators, wherein the specific process of step (4) is as follows: the competitive adaptive reweighted sampling algorithm is used to extract feature wavelengths from the spectral reflectance data of the soil samples at different depths, the number of Monte Carlo sampling is initialized to 1000, 80% of the samples are randomly selected as the training set samples, the remaining 20% of the samples are randomly selected as the validation set samples, the wavelength importance weight is calculated by the PLSR algorithm, iterative sampling is continuously performed, the wavelength retention frequency is counted, and the top 10% of the wavelengths with the highest frequency are selected as the input quantity of the multi-parameter inversion model.

[0011] The method for constructing the soil fertility index comprehensive evaluation model based on multiple soil indicators, wherein step (5) is to use the chemical method to respectively measure the soil parameter data of the organic matter, total nitrogen, salt content and calcium carbonate of each regional soil sample as the output quantity of the training of the multi-parameter inversion model.

[0012] The method for constructing the soil fertility index comprehensive evaluation model based on multiple soil indicators, wherein the specific process of step (6) is as follows: the inversion algorithm model for the four soil parameters, i.e., the organic matter, total nitrogen, salt content and calcium carbonate, under the condition of different depths is established, the deep learning network is designed, and the deep learning network is used to automatically learn the features from the hyperspectral data and perform the inversion of the soil parameters. The structure of the deep learning network includes an input layer, a feature extraction layer, multiple fully connected layers and an output layer; the input layer inputs the spectral reflectance data after feature extraction; the feature extraction layer extracts important features through a residual module and a channel attention module; the multiple fully connected layers gradually fuse the features through each fully connected layer, and finally output four soil parameters, i.e., the organic matter, total nitrogen, salt content and calcium carbonate, through linear transformation; and the output layer outputs the predicted values of the four soil parameters. The deep learning network needs to be artificially set parameters when training, that is, the convolution layer activation function is ReLU, the optimizer is Adam, the loss function is mean square error function, the batch size is 64, and the learning rate is 0.0001; for soil samples of different depths, 90% of the samples in each region are set as the training set, and 10% of the samples are set as the test set, the sample reflectivity data is input into the deep learning network for inversion model training, and the multi-parameter inversion network parameters in different depth ranges are obtained.

[0013] The method for constructing the soil fertility index comprehensive evaluation model based on multiple soil indicators, wherein the specific process of establishing the inversion algorithm model is: (6.1.1) Data preparation The input data, that is, the prepared hyperspectral reflectivity data of the soil sample, are used as the input of the inversion algorithm model; the output data, that is, the experimental measurement values of the corresponding four soil parameters, that is, organic matter, total nitrogen, salt content and calcium carbonate, in the soil sample, are used as the output of the inversion algorithm model; (6.1.2) Selection of inversion algorithm The PLSR inversion algorithm is selected to map the input data to a low-dimensional space through dimension reduction, and to predict the soil parameters through regression analysis; (6.1.3) Training and optimization The data are divided into k subsets using k-fold cross-validation, and each time the model is trained, k-1 subsets are used for training and the remaining one subset is used for validation, which is repeated k times; the model parameters are adjusted to improve the performance of the model; the optimal feature number is selected, and the prediction error is calculated to evaluate the accuracy of the model.

[0014] The method for constructing the soil fertility index comprehensive evaluation model based on multiple soil indicators, wherein the soil fertility comprehensive evaluation model established in step (7) is: ; In the above formula, represent the evaluation scores of the four soil parameters, that is, total nitrogen, organic matter, calcium carbonate and salt content, respectively; represent the proportions of the four soil parameters, that is, total nitrogen, organic matter, calcium carbonate and salt content, in the soil fertility comprehensive evaluation model, respectively.

[0015] The method for constructing the soil fertility index comprehensive evaluation model based on multiple soil indicators, wherein the specific process of step (8) is: based on the multi-parameter inversion model and the soil fertility index comprehensive evaluation model in different depth ranges, the pre-processed hyperspectral image of each pixel point is quantitatively inverted to obtain the inversion grayscale images of the four soil parameters of total nitrogen, organic matter, calcium carbonate and salt content and the soil fertility index, and the grayscale images are subjected to pseudo-color processing to obtain the visual color distribution map of the organic matter, total nitrogen, salt content, calcium carbonate content and soil fertility index on the soil panoramic profile.

[0016] By adopting the technical scheme, the present application has the following beneficial effects: The present application can obtain soil profile spectral data at different depths with low loss and high speed, thereby inverting soil index data such as total nitrogen, organic matter, calcium carbonate and water-soluble salt, and greatly improving the soil fertility evaluation efficiency in combination with the proposed soil fertility index (SFI) calculation.

[0017] The soil fertility index comprehensive evaluation model of the present application can significantly improve the efficiency of soil fertility evaluation through low-loss and fast spectral data acquisition and multi-parameter inversion, and has the following multiple beneficial effects and advantages: (1) efficient and low-loss acquisition The present application adopts an imaging spectral device composed of a spectral sensing module and a fiber image transmission bundle, which can quickly acquire soil profile spectral data at different depths without damaging the soil sample; compared with the traditional laboratory sampling and profile excavation method, the present application can greatly reduce the complexity of the soil damage and sampling process, and realize efficient and low-loss monitoring of a large range of soil.

[0018] (2) multi-parameter comprehensive inversion to improve accuracy The present application is not limited to inverting a single soil parameter (such as organic matter, total nitrogen, calcium carbonate, etc.), but combines multiple soil indicators to comprehensively evaluate soil fertility; through deep learning network inversion of multiple soil indicators, the accuracy of soil fertility evaluation is greatly improved, and the health status of soil can be more comprehensively and meticulously reflected.

[0019] (3) dynamic monitoring and continuous evaluation Traditional methods usually rely on static laboratory analysis and cannot provide continuous dynamic soil monitoring data; while the present application can realize long-term and continuous monitoring of soil fertility through non-destructive sampling and inversion technology based on spectral data, and provides real-time data support for farmland management.

[0020] (4) strong adaptability, suitable for different regions and soil types The method of the present application can be flexibly adapted to different soil types in different regions; by combining the specific properties of the soil with hyperspectral data for deep learning training, the model can be widely applied to different soil environments, and has strong universality.

[0021] (5) Reducing manual intervention and cost Compared with the traditional manual sampling and chemical experimental analysis method, the method of the present application reduces manual intervention, reduces labor cost and laboratory analysis cost. The automatic data acquisition and analysis system not only improves the efficiency, but also reduces the human error.

[0022] (6) Precise deep soil profile analysis Through spectral inversion technology, the present application can accurately analyze the soil composition at different depths, especially can overcome the problem that traditional technology is difficult to obtain deep soil data. Not only can the surface soil be evaluated, but also accurate fertility evaluation can be provided for deep soil, and more comprehensive soil health data support can be provided for agricultural production.

[0023] (7) Environmentally friendly, in line with the requirements of sustainable development The soil evaluation method of the present application adopts non-destructive spectral imaging technology, which avoids the damage to the environment caused by traditional soil sampling, meets the needs of sustainable agriculture and ecological environment protection, and helps to promote the green development of agricultural production.

[0024] (8) Improve the scientificity of soil fertility management and agricultural decision-making The present application provides more accurate and comprehensive soil fertility evaluation data to help agricultural producers, researchers and policy makers manage and utilize soil resources more scientifically. This not only helps to improve the accuracy of soil management, but also effectively reduces the amount of fertilization and reduces the environmental burden in agricultural production.

[0025] (9) Provide technical support for precision agriculture The present application provides technical support for soil management in precision agriculture, which can accurately guide soil improvement and fertilization plan, improve the utilization efficiency of agricultural resources, optimize the agricultural production process, and reduce resource waste and environmental pollution.

[0026] (10) Large-scale application potential The soil evaluation model of the present application can process large-scale regional data, is suitable for large-scale soil monitoring and evaluation, has broad application prospects, and can meet the needs of modern agriculture for large-scale soil monitoring and precise management. BRIEF DESCRIPTION OF DRAWINGS

[0027] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0028] Figure 1 The flowchart illustrates the method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators, as described in this invention. Detailed Implementation

[0029] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The present invention will be further explained below with reference to specific embodiments.

[0031] like Figure 1 As shown in the figure, the method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators provided in this embodiment specifically includes the following steps: S100. Soil sample image data from different regions are acquired using a hyperspectral imaging device, and then the image data is preprocessed to obtain a panoramic hyperspectral image of the soil profile; the specific process is as follows: Using a constructed spectral sensing module and fiber optic image bundle synthesized imaging spectrometer, spectral data of soil samples from different regions were collected. A cylindrical pit with a diameter of 0.5m and a height of 2m was excavated. A standard reflectance plate was placed at the pit opening, and the imaging spectrometer was positioned at the pit opening and moved downwards to acquire a batch of image data. Wavelet thresholding denoising algorithms were applied to the image data to remove honeycomb artifacts, and the Brown-Conrady model was used to eliminate image edge distortion. Lateral and vertical image stitching was performed using cylindrical coordinate system projection to obtain a panoramic hyperspectral image of the soil profile with a spectral range of 400-1000nm and a spectral resolution of 5nm.

[0032] S200. Divide the soil into different regions according to depth, manually screen for abnormal areas, extract 1000 spectral data points from each depth range, and homogenize the spectral data of the standard reflectance plate region, extracting one spectral data point from it; the specific process is as follows: For the obtained soil profile hyperspectral image, the longitudinal direction is the soil depth, the actual depth is 2m, and the hyperspectral image is divided into 20 regions according to the 10cm depth and the standard reflectance plate region at the pit opening; manually screen and sample at different depths in the soil profile hyperspectral image, and after removing the abnormal region data, 1000 soil sample data are extracted respectively, and the standard reflectance plate region in the soil profile hyperspectral image is framed, and the data in the framed region is processed by mean value, and 1 standard reflectance plate data is obtained.

[0033] S300, according to the laboratory calibration coefficient, the extracted spectral data is radiometrically corrected, converted into radiance data, and based on the standard reflectance radiance data, the soil sample radiance data is converted into reflectance, and pretreated; the specific process is: The soil sample data and the standard reflectance plate data are radiometrically corrected by the calibration data measured in the laboratory, the DN value is converted into radiance value, and based on the standard reflectance plate data, the soil sample radiance data is converted into soil sample spectral reflectance data; then, the soil sample reflectance data is pretreated, that is, Savitzky-Golay filtering and multivariate scatter correction are performed respectively, to reduce the noise interference and the baseline deviation and amplitude change caused by soil particle size and roughness.

[0034] S400, the characteristic wavelength of the soil sample spectral reflectance data at different depths is extracted by using the competitive adaptive reweighted sampling algorithm (CARS), the Monte Carlo sampling number is initialized to 1000, 80% is randomly selected as the training set sample, and the remaining 20% is randomly selected as the validation set sample, the wavelength importance weight is calculated by PLSR algorithm, the iteration sampling is continuously carried out, the wavelength retention frequency is counted, and the top 10% wavelength with the highest frequency is selected as the input quantity of the multi-parameter inversion model.

[0035] S500, the organic matter, total nitrogen, salt content and calcium carbonate of each region soil sample are measured by chemical method, which are used as the output quantity of the multi-parameter inversion model training.

[0036] S600, a deep learning network structure is designed and trained, and a multi-parameter inversion model at different depths is obtained, which is used for soil multi-parameter spectral quantitative inversion. Among them, the inversion algorithm model for four soil parameters (organic matter, total nitrogen, salt content and calcium carbonate) under different depth conditions is established, and a deep learning network is designed.

[0037] The specific process of establishing the inversion algorithm model is as follows: The goal of the inversion algorithm model is to extract soil parameters such as organic matter, total nitrogen, salinity, and calcium carbonate from spectral data; the inversion algorithm model usually uses methods such as regression analysis, least squares, or partial least squares regression (PLSR); the specific steps are as follows: S610, data preparation Input data: prepare hyperspectral reflectance data of soil samples (e.g., wavelength features extracted using the CARS algorithm), which are used as input for the inversion model; Output data: prepare experimental measurements of the corresponding four soil parameters (organic matter, total nitrogen, salinity, and calcium carbonate) in the soil samples as the output of the inversion model.

[0038] S620, select inversion algorithm Selecting the appropriate inversion algorithm is crucial. Common inversion algorithms include multiple linear regression, partial least squares regression (PLSR), and support vector machine regression (SVR). Among them, the PLSR algorithm is commonly used and suitable for high-dimensional data and multi-parameter inversion.

[0039] The PLSR inversion algorithm is a method that combines principal component analysis (PCA) and least squares regression. The goal of PLSR is to map the input data to a low-dimensional space through dimensionality reduction and predict soil parameters through regression analysis; the specific process is as follows: S621, standardize input data: standardize the spectral data to make the mean zero and the standard deviation one to eliminate dimensional differences.

[0040] S622, dimensionality reduction (PCA): perform principal component analysis (PCA) on the standardized spectral data to extract the most important principal components; usually select the first few principal components, which can explain most of the data variability.

[0041] S623, regression analysis: use the selected principal components and target soil parameters for regression analysis to establish an inversion model.

[0042] S630, training and optimization Cross-validation: to avoid overfitting and ensure the generalization ability of the model, use cross-validation techniques to evaluate the performance of the model. Usually use k-fold cross-validation, divide the data into k subsets, each time train the model, use k-1 subsets for training, the remaining one subset for validation, repeat k times.

[0043] Optimize the model: adjust the model parameters (such as the number of principal components in PLSR) to improve the performance of the model. Select the optimal number of features and calculate the prediction error (such as root mean square error RMSE) to evaluate the accuracy of the model.

[0044] The specific process of designing a deep learning network is as follows: When designing a deep learning network, the goal is to use the network to automatically learn features from hyperspectral data and perform soil parameter inversion. Common deep learning network structures include multi-layer perceptron (MLP), convolutional neural network (CNN), and recurrent neural network (RNN), but since the task is mainly a regression problem, a fully connected neural network (MLP) is usually used.

[0045] ① Network structure design The fully connected network (MLP) structure is suitable for regression tasks and has high flexibility and expressiveness.

[0046] Input layer: The number of nodes in the input layer is equal to the number of features in the reflectance data (i.e., the number of selected spectral wavelengths). For example, if 100 feature wavelengths are selected, the input layer will contain 100 nodes.

[0047] Hidden layer: Generally, two to three hidden layers are set. The number of neurons in each layer can be adjusted according to the complexity of the data. Generally, starting from 64 to 256 neurons can be tested and adjusted gradually.

[0048] Activation function: The hidden layer uses the ReLU activation function, which can improve the non-linear expression ability of the model and avoid the problem of gradient disappearance.

[0049] Output layer: Since this is a regression task, the number of nodes in the output layer is equal to the number of soil parameters to be predicted (e.g., four parameters: organic matter, total nitrogen, salt content, and calcium carbonate). The output layer usually uses a linear activation function (i.e., no activation function).

[0050] ② Network training Loss function: Since it is a regression task, the loss function usually chooses mean squared error (MSE), whose formula is: Optimizer: Use the Adam optimizer to optimize network parameters. The Adam optimizer combines momentum and adaptive learning rate, which can provide fast convergence in most problems: Batch size and learning rate: Usually use Batch Gradient Descent or Mini-batch Gradient Descent, set appropriate batch size (such as 64 or 128), and adjust the learning rate (e.g., 0.001) to ensure good convergence.

[0051] ③ Training process Data segmentation: Divide the dataset into a training set (usually 80%) and a test set (20%) for training and evaluating the performance of the model.

[0052] Training: Train the network using the training set data, update the network parameters step by step, and minimize the loss function.

[0053] Evaluation: Evaluate the performance of the model on the test set, calculate the root mean square error (RMSE) or other regression evaluation indicators, and ensure the generalization ability of the model.

[0054] The final structure of the deep learning network described above includes an input layer, a feature extraction layer, multiple fully connected layers, and an output layer. The input of the input layer is the spectral reflectance data extracted by the feature extraction layer; the feature extraction layer extracts important features through residual modules and channel attention modules; each fully connected layer in the multiple fully connected layers gradually fuses features, and finally outputs four soil parameters (organic matter, total nitrogen, salt content, and calcium carbonate) through linear transformation; the output of the output layer is the predicted value of the four soil parameters. This design enables the network to not only capture spectral features related to soil fertility, but also automatically learn and optimize the model, improving inversion accuracy.

[0055] The input layer of the deep learning network is the feature wavelength reflectance data selected in step S400, which is added to the residual module and channel attention module by the feature extraction layer of the deep learning network, and finally connected to four fully connected layers to obtain four soil parameter inversion data. The network hyperparameters (here, the network hyperparameters refer to parameters that need to be manually set when training the deep learning model. These parameters are not automatically optimized by the model through learning data, but need to be adjusted according to actual conditions. These hyperparameters are crucial to the training process of the model, affecting the convergence speed, learning effect, and final performance of the model) are as follows: the activation function of the convolutional layer is ReLU, the optimizer is Adam, the loss function is mean square error function (MSE), the batch size is 64, and the learning rate is 0.0001. For 20 different soil samples of different depths, 90% of the samples in each region are set as the training set and 10% of the samples are set as the test set. The sample reflectance data is input into the deep network for inversion model training, and the multi-parameter inversion network parameters are obtained under 20 different depth ranges.

[0056] S700, a soil fertility comprehensive evaluation model (SFI) is established for four types of parameter data (organic matter, total nitrogen, salt content, and calcium carbonate) in different regions and different soil depths. ; In the above formula, represent the evaluation scores of each soil parameter, i.e., scoring each type of parameter based on Table 1 and the data inverted by the model, represent the proportion of each type of parameter in the evaluation model. The proportion of each parameter in the evaluation model can be assigned by experts based on actual land types, and the comprehensive evaluation score of the soil sample (100 points) is obtained.

[0057] S800. Using an inversion algorithm model and an evaluation model, multi-parameter inversion and soil fertility index calculation are performed on each pixel of the hyperspectral image to obtain a color distribution map of different soil parameters and soil fertility indices; the specific process is as follows: Based on the multi-parameter inversion model and the comprehensive evaluation model of soil fertility index at different depth ranges, quantitative inversion is performed on each pixel of the pre-processed panoramic hyperspectral image to obtain inversion grayscale images of four soil parameters and soil fertility index. The grayscale images are then processed with pseudo-color to obtain a visualized color distribution map of organic matter, total nitrogen, salt content, calcium carbonate content and soil fertility index on the soil panoramic profile.

[0058] Table 1. Scores of Soil Fertility Parameters:

[0059] This invention can acquire spectral data of soil profiles at different depths with low loss and speed, thereby retrieving soil index data such as total nitrogen, organic matter, calcium carbonate, and water-soluble salts. Combined with the proposed Soil Fertility Index (SFI) calculation, it can significantly improve the efficiency of soil fertility evaluation.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators, characterized in that, Specifically, the following steps are included: (1) Collect soil sample image data from different regions, and then preprocess the image data to obtain a panoramic view of the soil profile hyperspectral image; (2) For the obtained soil profile hyperspectral image, different regions are divided according to soil depth and manual screening and sampling are carried out to screen out abnormal regions and standard reflectance plate regions. After removing the data of abnormal regions, soil sample data within each depth range is extracted. Then, the spectral data of the standard reflectance plate region is homogenized and the standard reflectance plate data is extracted. (3) Perform radiometric correction on the extracted soil sample data and standard reflectance plate data, convert them into radiance data, and then convert the soil sample radiance data into soil sample spectral reflectance data based on the standard reflectance plate data. Finally, preprocess the soil sample spectral reflectance data. (4) Extract characteristic wavelengths from the spectral reflectance data of soil samples at different depths and use them as inputs for the multi-parameter inversion model; (5) Use chemical methods to measure the parametric data of soil samples at different depths, and use them as the output of the multi-parameter inversion model training; (6) Design a deep learning network structure and train it to obtain a multi-parameter inversion model at different depths. This multi-parameter inversion model is used for multi-parameter spectral quantitative inversion of soil. (7) Establish a comprehensive evaluation model for soil fertility based on the data of various parameters at different soil depths in different regions; (8) By establishing the inversion algorithm model and the comprehensive evaluation model of soil fertility, multi-parameter inversion and soil fertility index calculation are performed on each pixel of the hyperspectral image to obtain color distribution maps of different soil parameters and soil fertility index.

2. The method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators as described in claim 1, characterized in that, The specific process of step (1) is as follows: using the constructed spectral sensing module and fiber optic image bundle to synthesize imaging spectral equipment, sample spectral data are collected for soil in different regions; then, for the selected soil sampling area, a pit is dug out, which is the location for soil sample collection, a standard reflectivity plate is placed at the pit opening, and the imaging spectral equipment is placed at the pit opening and moved downwards to obtain image data; wavelet threshold denoising algorithm is used to remove honeycomb artifacts from the image data in sequence, and then the image edge distortion is eliminated based on the Brown-Conrady model. The image is then stitched horizontally and vertically using cylindrical coordinate system projection to obtain a panoramic view of the soil profile hyperspectral image.

3. The method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators as described in claim 1, characterized in that: In step (3), the extracted soil sample data and standard reflectance plate data are subjected to radiation correction. This involves using calibration data measured in the laboratory to convert the original digital number, i.e., the DN value, into a radiance value, so as to more accurately reflect the radiation characteristics of the soil.

4. The method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators as described in claim 1, characterized in that: In step (3), the preprocessing of the soil sample spectral reflectance data specifically involves Savitzky-Golay filtering and multivariate scattering correction. Savitzky-Golay filtering smooths the data by performing local polynomial fitting on the soil sample spectral reflectance data to remove noise. Multivariate scattering correction is used to reduce changes caused by spectral scattering and eliminate spectral changes caused by differences in the physical properties of the samples.

5. The method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators as described in claim 1, characterized in that, The specific process of step (4) is as follows: the competitive adaptive reweighted sampling algorithm is used to extract the characteristic wavelengths of the spectral reflectance data of soil samples at different depths. The Monte Carlo sampling number is initialized to 1000. 80% are randomly selected as training set samples and the remaining 20% ​​are validation set samples. The wavelength importance weight is calculated by the PLSR algorithm. Iterative sampling is continuously performed, the wavelength retention frequency is statistically analyzed, and the top 10% of wavelengths with the highest frequency are selected as the input of the multi-parameter inversion model.

6. The method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators as described in claim 1, characterized in that, Step (5) involves using chemical methods to measure soil parameters such as organic matter, total nitrogen, salt content, and calcium carbonate in each region's soil sample, which are then used as outputs for training the multi-parameter inversion model.

7. The method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators as described in claim 1, characterized in that, The specific process of step (6) is as follows: establish an inversion algorithm model for four soil parameters, namely organic matter, total nitrogen, salinity and calcium carbonate, under different depth conditions, design a deep learning network and use the deep learning network to automatically learn features from hyperspectral data and perform soil parameter inversion; The deep learning network structure includes an input layer, a feature extraction layer, multiple fully connected layers, and an output layer. The input layer takes in the spectral reflectance data after feature extraction. The feature extraction layer extracts important features through a residual module and a channel attention module. The multiple fully connected layers fuse features step by step through each fully connected layer, and finally output four soil parameters—organic matter, total nitrogen, salinity, and calcium carbonate—through a linear transformation. The output layer outputs the predicted values ​​of the four soil parameters. During training, the deep learning network requires manual parameter setting: the convolutional layer activation function is ReLU, the optimizer is Adam, the loss function is mean squared error, the batch size is 64, and the learning rate is 0.0001. For soil samples at different depths, 90% of the samples in each region are set as the training set and 10% as the test set. The sample reflectance data is input into the deep learning network to train the inversion model and obtain the multi-parameter inversion network parameters for different depth ranges.

8. The method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators as described in claim 8, characterized in that, The specific process for establishing the inversion algorithm model is as follows: (6.1.1) Data preparation Input data, namely the hyperspectral reflectance data of the soil sample, is used as the input to the inversion algorithm model; output data, namely the experimental measurements of the four soil parameters corresponding to the soil sample, namely organic matter, total nitrogen, salinity and calcium carbonate, are used as the output of the inversion algorithm model. (6.1.2) Selecting an inversion algorithm The PLSR inversion algorithm is selected to map the input data to a low-dimensional space by dimensionality reduction and to predict soil parameters through regression analysis. (6.1.3) Training and Optimization Using k-fold cross-validation, the data is divided into k subsets. Each time the model is trained, k-1 subsets are used for training, and the remaining subset is used for validation. This process is repeated k times. The model performance is improved by adjusting the model parameters. The optimal number of features is selected, and the prediction error is calculated to evaluate the model's accuracy.

9. The method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators as described in claim 1, characterized in that, The comprehensive soil fertility evaluation model established in step (7) is as follows: ； In the above formula, These represent the evaluation scores for four soil parameters: total nitrogen, organic matter, calcium carbonate, and salinity, respectively. These represent the proportions of the four soil parameters—total nitrogen, organic matter, calcium carbonate, and salt content—in the comprehensive evaluation model of soil fertility.

10. The method for constructing a comprehensive evaluation model of soil fertility index based on multiple soil indicators as described in claim 1, characterized in that, The specific process of step (8) is as follows: Based on the multi-parameter inversion model and the comprehensive evaluation model of soil fertility index under different depth ranges, quantitative inversion is performed on each pixel of the pre-processed panoramic hyperspectral image to obtain the inversion grayscale images of the four soil parameters of total nitrogen, organic matter, calcium carbonate and salt content and soil fertility index respectively. The grayscale images are processed by pseudo-color to obtain the visualized color distribution map of organic matter, total nitrogen, salt content, calcium carbonate content and soil fertility index on the panoramic soil profile.

Images