Karst fragile area ecological restoration method and system based on measurement characteristics

Abstract

The invention relates to the technical field of data processing, and discloses a karst fragile area ecological restoration method and system based on measurement characteristics. The method comprises the following steps: detecting the content of carbon, nitrogen, phosphorus, potassium, calcium and magnesium in karst plants through a multispectral sensor, carrying out standardization processing to obtain a stoichiometric characteristic matrix, inputting a deep convolutional neural network to identify a plant conservation and resource acquisition adaptation strategy, quantitatively analyzing microtopography and niche environmental influence weight, and calculating the content of carbon, nitrogen, phosphorus, potassium, calcium and magnesium in the karst plants. And optimizing a restrictive nutrient factor ratio based on a calcium-magnesium regulation mechanism, and generating a differential nutrient regulation scheme in combination with an adaptive strategy to form a karst fragile region accurate ecological restoration implementation scheme. The technical problems that in ecological restoration of the karst fragile region, nutrient limiting factors cannot be accurately recognized based on plant stoichiometric characteristics, the quantitative analysis capability of environmental heterogeneity is lacked, and personalized nutrient proportion optimization is difficult to achieve are solved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a method and system for ecological restoration in karst vulnerable areas based on metering characteristics. Background Art

[0002] As an important global geological landscape type, the unique carbonate rock geological structure of karst landforms has formed complex and diverse landforms and ecological environments. Under such special geological conditions, vegetation growth faces severe challenges such as poor soil, serious nutrient loss, and high pH value. Traditional ecological restoration methods mainly rely on empirical plant selection and extensive nutrient supplementation, improving soil conditions by spreading chemical fertilizers or organic fertilizers over a large area, and at the same time using conventional vegetation configuration models for ecological reconstruction. These methods have achieved certain results in ecological restoration in plain areas and provided basic technical support for ecosystem restoration.

[0003] However, the existing ecological restoration technologies have shown obvious limitations and deficiencies in the application in karst vulnerable areas. Firstly, there is a lack of in-depth analysis of plant stoichiometric characteristics, and it is impossible to accurately identify the true nutrient requirements and limiting factors of different plants under specific environmental conditions, resulting in blindness and inefficiency in nutrient supplementation. Secondly, the influence of the spatial heterogeneity of microtopography and microhabitat on plant adaptation strategies is ignored, and it is difficult to adapt to the complex and changeable environmental conditions in karst areas by adopting a unified restoration mode. Thirdly, there is a lack of understanding of the action mechanisms of key regulatory elements such as Ca and Mg, and it is impossible to make full use of the rich calcium and magnesium resources in karst areas for precise regulation. Finally, there is a lack of an intelligent decision support system, and it is impossible to achieve dynamic optimization and precise implementation of the restoration plan.

[0004] Based on the deficiencies of the existing technology and the actual needs of ecological restoration in karst vulnerable areas, it is urgent to solve the technical problems of how to accurately identify plant stoichiometric characteristics and determine limiting nutrient factors, how to formulate a differential nutrient regulation plan based on plant adaptation strategies and the degree of influence of environmental factors, and how to establish a precise ecological restoration decision-making system integrating multi-source data and intelligent algorithms. Traditional restoration methods cannot accurately quantify the coupling relationship between plant nutrient requirements and environmental conditions, and it is difficult to achieve personalized nutrient ratios based on stoichiometric characteristics. The existing technology lacks the ability to quantitatively analyze the spatial variability of microtopography and microhabitat, and it is impossible to adjust the restoration strategy according to environmental heterogeneity. The current ecological restoration lacks an intelligent data processing and decision optimization mechanism, and it is difficult to generate a precise restoration plan with multi-factor coordination. Summary of the Invention

[0005] The present application provides a method and system for ecological restoration in karst fragile areas based on stoichiometric characteristics, which solves the technical problems in ecological restoration of karst fragile areas, such as the inability to accurately identify nutrient limiting factors based on plant stoichiometric characteristics, the lack of quantitative analysis ability for environmental heterogeneity, and the difficulty in realizing personalized nutrient ratio optimization.

[0006] In the first aspect, the present application provides a method for ecological restoration in karst fragile areas based on stoichiometric characteristics. The method for ecological restoration in karst fragile areas based on stoichiometric characteristics includes: detecting nutrient elements of plants in karst fragile areas through a multispectral sensor to obtain carbon, nitrogen, phosphorus, potassium, calcium, and magnesium content data, and obtaining a plant stoichiometric characteristic matrix through standardization processing; inputting the plant stoichiometric characteristic matrix into a deep convolutional neural network for pattern recognition processing, outputting a conservative strategy label and a resource acquisition strategy label, and obtaining a plant adaptation strategy classification result; performing weighted quantitative analysis on environmental impact factors according to microtopography parameters and microhabitat data, and calculating to obtain a microtopography impact weight value and a microhabitat impact weight value; optimizing the ratio calculation of limiting nutrient factors according to the calcium and magnesium element regulation mechanism, and generating a targeted nutrient regulation plan in combination with the plant adaptation strategy classification result to obtain an implementation plan for ecological restoration in karst fragile areas.

[0007] In the second aspect, the present application provides a system for ecological restoration in karst fragile areas based on stoichiometric characteristics. The system for ecological restoration in karst fragile areas based on stoichiometric characteristics includes: a detection module, configured to detect nutrient elements of plants in karst fragile areas through a multispectral sensor to obtain carbon, nitrogen, phosphorus, potassium, calcium, and magnesium content data, and obtaining a plant stoichiometric characteristic matrix through standardization processing; an identification module, configured to input the plant stoichiometric characteristic matrix into a deep convolutional neural network for pattern recognition processing, outputting a conservative strategy label and a resource acquisition strategy label, and obtaining a plant adaptation strategy classification result; a quantification module, configured to perform weighted quantitative analysis on environmental impact factors according to microtopography parameters and microhabitat data, and calculating to obtain a microtopography impact weight value and a microhabitat impact weight value; a ratio module, configured to optimize the ratio calculation of limiting nutrient factors according to the calcium and magnesium element regulation mechanism, and generating a targeted nutrient regulation plan in combination with the plant adaptation strategy classification result to obtain an implementation plan for ecological restoration in karst fragile areas.

[0008] In the third aspect, there is provided an ecological restoration device for karst fragile areas based on stoichiometric characteristics, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory to enable the ecological restoration device for karst fragile areas based on stoichiometric characteristics to execute the above-mentioned method for ecological restoration in karst fragile areas based on stoichiometric characteristics.

[0009] In a fourth aspect, a computer-readable storage medium is provided. Instructions are stored in the computer-readable storage medium, and when they run on a computer, the computer is made to execute the above-mentioned method for ecological restoration of karst vulnerable areas based on stoichiometric characteristics.

[0010] In the technical solution provided by this application, the present invention obtains data on the contents of carbon, nitrogen, phosphorus, potassium, calcium, and magnesium by detecting nutrient elements of plants in karst vulnerable areas through a multispectral sensor and obtains a plant stoichiometric characteristic matrix through standardization processing, achieving an accurate quantitative characterization of the internal nutrient status of plants, solving the problem of inaccurate identification of nutrient requirements caused by traditional methods relying on empirical judgment. At the same time, the plant stoichiometric characteristic matrix is input into a deep convolutional neural network for pattern recognition processing to output conservative strategy labels and resource acquisition strategy labels to obtain a plant adaptation strategy classification result. The multi-scale feature extraction ability of the deep convolutional neural network and the deep learning characteristics of the residual network structure enable the algorithm to automatically identify complex patterns and potential correlation relationships in stoichiometric data. Compared with traditional statistical analysis methods, the convolutional neural network can capture the spatial correlation between stoichiometric ratios through local receptive fields and weight sharing mechanisms, while residual connections and batch normalization techniques ensure the effective training and feature transfer of the deep network, significantly improving the accuracy and robustness of plant adaptation strategy recognition. Furthermore, based on microtopography parameters and niche data, a weight quantification analysis of environmental impact factors is carried out to obtain a microtopography impact weight value and a niche impact weight value through calculation. The time series modeling ability of the long short-term memory network and the weight allocation strategy of the attention mechanism enable the algorithm to accurately quantify the relative importance and spatio-temporal variation law of different environmental factors, overcoming the limitation of traditional methods that cannot quantitatively evaluate the impact of environmental heterogeneity. Finally, based on the calcium and magnesium element regulation mechanism, the limiting nutrient factors are optimized and proportioned, and combined with the plant adaptation strategy classification result, a targeted nutrient regulation plan is generated to obtain an ecological restoration implementation plan for karst vulnerable areas. The adversarial training mechanism of the generative adversarial network and the non-linear mapping ability of the multi-layer fully connected network enable the algorithm to learn the complex regulation relationship between calcium and magnesium elements and limiting nutrient factors, and continuously optimize the rationality of the proportioning plan through the game process between the generator and the discriminator, realizing a fundamental transformation from extensive unified fertilization to precise personalized nutrient regulation. The overall solution deeply integrates the pattern recognition, feature learning, and optimization decision-making capabilities of artificial intelligence algorithms with the professional knowledge of karst ecological restoration. The core contribution of the algorithm features to the solution is reflected in automatically discovering the internal laws between stoichiometric characteristics and adaptation strategies through deep learning, accurately quantifying the weights of environmental factors through time series networks and attention mechanisms, and realizing the intelligent optimization of nutrient ratios through generative adversarial networks, thereby constructing an intelligent ecological restoration technology system combining data-driven and knowledge-guided. Description of the Drawings

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0012] Figure 1 It is a schematic diagram of an embodiment of the ecological restoration method for karst vulnerable areas based on metering characteristics in the embodiments of the present application; Figure 2 It is a schematic diagram of an embodiment of the ecological restoration system for karst vulnerable areas based on metering characteristics in the embodiments of the present application; Figure 3 It is a structural schematic block diagram of the ecological restoration equipment for karst vulnerable areas based on metering characteristics in the embodiments of the present invention. Detailed implementation manners

[0013] The embodiments of the present application provide an ecological restoration method and system for karst vulnerable areas based on metering characteristics. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned accompanying drawings of the present application are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0014] For ease of understanding, the specific processes of the embodiments of the present application will be described below. Please refer to Figure 1 , an embodiment of the ecological restoration method for karst vulnerable areas based on metering characteristics in the embodiments of the present application includes: Step S101: Detect the nutrient elements of plants in karst vulnerable areas through a multispectral sensor, obtain the content data of carbon, nitrogen, phosphorus, potassium, calcium, and magnesium, and obtain a plant stoichiometric characteristic matrix through standardization processing; Step S102: Input the plant stoichiometric characteristic matrix into a deep convolutional neural network for pattern recognition processing, output a conservative strategy label and a resource acquisition strategy label, and obtain a plant adaptation strategy classification result; Step S103: Perform weighted quantification analysis on environmental impact factors according to microtopography parameters and microhabitat data, and calculate the microtopography impact weight value and the microhabitat impact weight value; Step S104: Optimally proportion and calculate the limiting nutrient factors according to the calcium and magnesium element regulation mechanism, generate a targeted nutrient regulation plan in combination with the classification results of plant adaptation strategies, and obtain an ecological restoration implementation plan for the karst fragile area.

[0015] It can be understood that the execution entity of this application can be an ecological restoration system for karst fragile areas based on measurement characteristics, or a terminal or a server. Specifically, no limitation is made here. In the embodiments of this application, the server is taken as an example of the execution entity for illustration.

[0016] Specifically, a multispectral sensor is used to obtain plant spectral reflectance data, and then these spectral data are input into an element content inversion algorithm for quantitative analysis. This algorithm calculates the original content values of carbon, nitrogen, phosphorus, potassium, calcium, and magnesium elements of a single plant based on the regression relationship between spectral bands and element content. Then, stoichiometric ratio calculations are performed on the original element content values, including fifteen stoichiometric ratio combinations such as carbon-nitrogen ratio, nitrogen-phosphorus ratio, and calcium-magnesium ratio. These stoichiometric ratio data form a multi-dimensional data vector to form a plant individual stoichiometric data set. Subsequently, outlier detection and missing value filling processing are performed on the data set, and the processed data is reconstructed into a classified stoichiometric data matrix according to the functional group category. Finally, through zero-mean normalization and maximum-minimum normalization processing, combined with micro-topographic position tags and micro-habitat type tags, a sample feature vector is constructed to obtain a plant stoichiometric feature matrix.

[0017] The plant stoichiometric feature matrix is used as input data and sent into a deep convolutional neural network. This network first performs feature extraction through multi-scale convolutional layers. Among them, the 3×3 convolutional kernel is responsible for extracting local stoichiometric relationship features, and the 7×7 convolutional kernel is responsible for extracting global nutrient distribution features. The two convolutional kernels perform convolutional operations on the stoichiometric feature matrix respectively to generate feature maps of different scales. Then, attention mechanism weighted fusion processing is performed on the multi-dimensional feature maps, and the local stoichiometric relationship features and global nutrient distribution features are combined and calculated according to the weight coefficients to obtain a fused feature vector. This fused feature vector is input into a residual network structure for deep feature learning. The input features are added to the intermediate layer features through skip connections, and at the same time, batch normalization operations are applied to standardize the output of each layer. Finally, a high-level semantic feature representation is obtained. Based on this feature representation, a softmax classifier calculation is performed, and the features are mapped into conservative strategy probability values and resource acquisition strategy probability values, and the classification result of plant adaptation strategies is output.

[0018] Perform weighted quantification analysis on microtopographic parameters and niche data. First, collect data on slope, aspect, altitude, and soil pH parameters in the karst vulnerable area, and construct a microtopographic parameter matrix with the collected data. At the same time, perform numerical coding on sunny slopes, shady slopes, semi-shady slopes, and flatland types to obtain a niche type vector. Then, input the microtopographic parameter matrix and the niche type vector into a long short-term memory network for temporal feature learning. This network controls which historical information needs to be forgotten through the forget gate, determines the importance of the current input information through the input gate, and controls the flow of output information through the output gate. The three gating mechanisms work together to extract the temporal feature sequence of environmental factors. Then, perform bidirectional recurrent calculation on the temporal feature sequence, and perform weighted summation of the forward hidden state and the backward hidden state according to the time step to obtain the bidirectional environmental impact feature representation. Finally, apply the attention weight allocation mechanism to the feature representation for importance evaluation, calculate the attention scores of each environmental factor through the softmax function, and obtain the microtopographic impact weight value and the niche impact weight value.

[0019] Optimize the ratio calculation of limiting nutrient factors according to the calcium and magnesium element regulation mechanism. First, perform limiting analysis on the nitrogen, phosphorus, and potassium element contents in the plant stoichiometric characteristic matrix, and identify the nutrient limiting factors through the minimum value determination algorithm. This algorithm compares the relative contents of nitrogen, phosphorus, and potassium to determine the limiting factor corresponding to the minimum value. At the same time, extract the calcium and magnesium element contents as regulatory variables to form a nutrient limiting factor dataset. Then, input this dataset into a generative adversarial network for ratio optimization. The generator network calculates the optimal ratio relationship between calcium and magnesium elements and the limiting nutrient factors through multiple fully connected layers, and the discriminator network evaluates the authenticity of the generated scheme. The two networks obtain the initial nutrient ratio scheme through adversarial training. Then, perform strategy matching on the initial ratio scheme based on the plant adaptation strategy classification results, and use the conservative strategy label and the resource acquisition strategy label as constraints to adjust the ratio parameters to obtain the strategy-matched nutrient ratio scheme. Finally, perform spatial differentiation processing on the ratio scheme according to the microtopographic impact weight value and the niche impact weight value, calculate the dedicated ratio coefficients according to different terrain conditions and habitat types, and form an ecological restoration implementation plan for the karst vulnerable area.

[0020] In a specific embodiment, the process of executing step S101 may specifically include the following steps: Collect spectral data of different functional group plants in the karst vulnerable area, input the spectral reflectance data into the element content inversion algorithm for quantitative analysis, and obtain the original content values of carbon, nitrogen, phosphorus, potassium, calcium, and magnesium elements of individual plants; Calculate the stoichiometric ratios between elements based on the original element content values to obtain the plant individual stoichiometric dataset; Outlier detection and missing value imputation are performed on the individual plant stoichiometry dataset, and the processed data is reconstructed into a matrix according to functional group categories to obtain a classified stoichiometry data matrix. The classified stoichiometry data matrix is subjected to zero-mean normalization and min-max normalization, and sample feature vectors are constructed by combining microtopography position tags and niche type tags to obtain a plant stoichiometry feature matrix.

[0021] Specifically, the spectral data collection of plants in different functional groups in the karst fragile area is to obtain the spectral reflectance information of plant leaves within a specific wavelength range through a multispectral sensor. Different functional group plants refer to plant groups with different ecological functions such as trees, shrubs, and herbs. The spectral reflectance data reflects the light energy reflection characteristics of plant leaves at different wavelengths. There are specific mathematical relationships between these reflectance values and the element contents inside the plants. The element content inversion algorithm is calculated based on the regression model between spectral bands and element concentrations. This algorithm converts the spectral reflectance values into specific element content values through a pre-established calibration equation, so as to obtain the original content values of six key elements of carbon, nitrogen, phosphorus, potassium, calcium, and magnesium in a single plant. These content values are expressed in milligrams per gram of dry weight. The calculation of stoichiometric ratios based on the obtained original element content values involves performing ratio operations on different element contents. Stoichiometric ratios refer to the relative proportional relationships of the contents of two or more elements, specifically including fifteen different element ratio combinations such as carbon-nitrogen ratio, carbon-phosphorus ratio, nitrogen-phosphorus ratio, calcium-magnesium ratio, and potassium-calcium ratio. These ratios are obtained through simple division operations. For example, the carbon-nitrogen ratio is equal to the carbon content divided by the nitrogen content. The combination of all stoichiometric ratio values of each plant individual forms a multi-dimensional data vector, and the set of multi-dimensional data vectors of all plant individuals constitutes the individual plant stoichiometry dataset, which contains the complete element proportion information of each plant.

[0022] Outlier detection uses statistical methods to identify values in a dataset that deviate from the normal range. Usually, the quartile method or the standard deviation method is used to determine the threshold range of outliers. Data points outside this range are marked as outliers and removed from the dataset. Missing value imputation processing addresses the possible data missing situations during data collection. Mean imputation or regression imputation methods are used to estimate and fill in the missing data. The processed data is grouped and sorted according to functional group categories such as trees, shrubs, and herbs. The stoichiometric data of each functional group is arranged in rows to form a submatrix, and the submatrices of all functional groups are combined by columns to form a classified stoichiometric data matrix. The rows of this matrix represent different plant individuals, and the columns represent different stoichiometric ratio indicators. Zero-mean normalization processing is to subtract the mean value of each indicator in the classified stoichiometric data matrix from the value of the indicator and then divide by the standard deviation, so that the mean value of each indicator becomes zero and the standard deviation is one, eliminating the differences in dimension and value range between different indicators. Min-max normalization processing is to subtract the minimum value of each indicator from the value of the indicator and then divide by the difference between the maximum value and the minimum value, so that all values are compressed into the range from zero to one. Micro-topography position labels include the coding identifiers of topographic position types such as upper slope position, middle slope position, lower slope position, and depression. Microhabitat type labels include the coding identifiers of habitat conditions such as sunny slope, shady slope, semi-shady slope, and flat ground. The stoichiometric data after standardization and normalization processing is combined by columns with the corresponding topographic position labels and microhabitat type labels to construct a sample feature vector containing stoichiometric characteristics and environmental label information. The feature vectors of all samples are arranged in rows to form a plant stoichiometric characteristic matrix, which contains both the intrinsic stoichiometric characteristics of plants and the information of external environmental conditions.

[0023] In a specific embodiment, the process of executing step S102 may specifically include the following steps: Input the plant stoichiometric characteristic matrix into a multi-scale convolutional layer for feature extraction processing. Extract local stoichiometric relationship features through a 3×3 convolutional kernel and global nutrient distribution features through a 7×7 convolutional kernel to obtain a multi-dimensional feature map; Perform attention mechanism weighted fusion processing on the multi-dimensional feature map, and combine and calculate the local stoichiometric relationship features and global nutrient distribution features according to the weight coefficients to obtain a fused feature vector; Input the fused feature vector into a residual network structure for deep feature learning processing, and optimize the gradient propagation process through skip connections and batch normalization operations to obtain a high-level semantic feature representation; Perform softmax classifier calculation processing based on the high-level semantic feature representation, map the feature representation to the probability values of conservative strategy and resource acquisition strategy, and obtain the classification result of plant adaptation strategy.

[0024] Specifically, in the multi-scale convolutional layer feature extraction process, the phyto-stoichiometric feature matrix is used as input data and fed into two parallel convolutional branches for feature extraction at different scales. The 3×3 convolutional kernel captures the local correlation relationships between adjacent elements through a sliding window operation on the stoichiometric feature matrix. This convolutional kernel covers a data area of three rows and three columns each time. After performing a dot product operation between the weight matrix and the input data and then summing them up, an output value is obtained. Such a small-sized convolutional kernel mainly identifies the fine-grained correlations between stoichiometric ratios, such as the direct correlation between the carbon-nitrogen ratio and the nitrogen-phosphorus ratio. The 7×7 convolutional kernel covers a larger receptive field range and processes a data area of seven rows and seven columns each time, capable of capturing the global distribution patterns and long-range dependence relationships among more extensive nutrient elements, such as the complex regulatory relationships between calcium and magnesium elements and other multiple elements. The two convolutional kernels at different scales perform convolutional operations on the stoichiometric feature matrix respectively, generating corresponding local stoichiometric relationship feature maps and global nutrient distribution feature maps. These feature maps are stacked along the channel dimension to form a multi-dimensional feature map spectrum. The attention mechanism weighted fusion process optimizes the feature combination by calculating the importance weights of different feature channels. First, a global average pooling operation is performed on the multi-dimensional feature map spectrum to compress the two-dimensional feature map of each feature channel into a scalar value. Then, these scalar values are input into a two-layer fully connected network for non-linear transformation. The first fully connected network compresses the input dimension to one-sixteenth of the original. After being processed by the ReLU activation function, it is input into the second fully connected network to restore to the original number of channels. Finally, the output value is restricted between zero and one through the sigmoid activation function as the attention weight for each channel. The local stoichiometric relationship features and the global nutrient distribution features are multiplied element-wise with the corresponding attention weights respectively, and then the weighted features are added along the channel dimension to obtain a fused feature vector, which synthesizes the important information of features at different scales.

[0025] The residual network structure depth feature learning process adopts a skip connection mechanism to solve the problem of gradient disappearance in the training of deep networks. The fused feature vector is first input into the first residual block, which contains two convolutional layers. The first convolutional layer performs a convolutional operation on the input features to obtain intermediate features, and the second convolutional layer continues to perform convolutional processing on the intermediate features. Then, the output of the second convolutional layer and the original input of the residual block are added element-wise through a skip connection. Such a skip connection allows the gradient to directly backpropagate from the output layer to the input layer. Batch normalization operations are performed after each convolutional layer to standardize the features by calculating the mean and variance of the current batch of data, so that the input distribution of each layer remains stable. After cascading multiple residual blocks, a high-level semantic feature representation is obtained, which contains the deep abstract information of phyto-stoichiometric features.

[0026] The softmax classifier calculates and processes the high-level semantic feature representation, mapping it through a fully connected layer to two output nodes, corresponding to the conservative strategy and the resource acquisition strategy respectively. The fully connected layer multiplies the weight matrix with the feature vector and adds the bias vector to obtain two raw score values. The softmax function performs exponential operations on these two score values and then normalizes them so that the sum of the two output values equals one. The specific calculation process is to take the exponent of each score value and divide it by the sum of the exponents of all score values, obtaining the probability value of the conservative strategy and the probability value of the resource acquisition strategy. The strategy category with the larger probability value is used as the final adaptation strategy classification result for the plant. The entire classification process makes decisions based on the deep learning representation of the plant stoichiometric characteristics.

[0027] In a specific embodiment, the process of inputting the fused feature vector into the residual network structure for deep feature learning processing may specifically include the following steps: Input the fused feature vector into the first residual block for non-linear transformation processing. Through the convolutional layer and the activation function, calculate the first intermediate feature, and perform a skip connection addition operation with the fused feature vector to obtain the first residual output feature; Perform batch normalization processing on the first residual output feature, calculate the mean and variance in the batch dimension for the feature data for standardization transformation, and obtain the normalized first residual feature; Input the normalized first residual feature into the second residual block for deep feature abstraction processing. Through the combination of multiple convolutional layers and non-linear activation functions, calculate the second intermediate feature, and perform a skip connection addition operation with the normalized first residual feature to obtain the second residual output feature; Perform global average pooling processing on the second residual output feature, compress the two-dimensional feature map into a one-dimensional feature vector, and obtain the high-level semantic feature representation through the average value calculation in the spatial dimension.

[0028] Specifically, the first residual block non-linear transformation process performs feature learning on the fused feature vector through two consecutive convolutional layers. The first convolutional layer uses a 3x3 convolutional kernel to perform convolution operations on the fused feature vector, performing a dot product summation operation with the local region of the input features through weight parameters to generate a preliminary non-linear feature representation. Then, after passing through the ReLU activation function, negative values are set to zero and positive values are retained to enhance the non-linear expression ability of the features. The second convolutional layer continues to perform convolution processing on the activated features, using a convolutional kernel of the same size to further extract deeper feature information. The concatenated processing of the two convolutional layers obtains the first intermediate feature, which contains the abstracted information after two non-linear transformations. The skip connection addition operation adds the first intermediate feature and the original fused feature vector element by element. This direct addition mechanism ensures that the original information is not lost during the propagation of the deep network, and at the same time allows the network to learn the residual mapping rather than the complete mapping. After addition, the first residual output feature is obtained, which contains both the new features after non-linear transformation and the original feature information of the input.

[0029] The batch normalization process performs statistical standardization operations on the first residual output feature. First, the mean of all samples in the current batch in each feature dimension is calculated by adding the feature values of all samples in that dimension and then dividing by the number of samples. Then, the variance of the feature values in that dimension is calculated by averaging the squared differences between each sample feature value and the mean. The standardization transformation subtracts the mean of the corresponding dimension from each feature value and then divides by the standard deviation, where the standard deviation is equal to the square root of the variance. This standardization operation makes the data distribution of each feature dimension have a mean of zero and a variance of one, eliminating the differences in the numerical ranges between different feature dimensions. Batch normalization also introduces learnable scaling parameters and translation parameters, allowing the network to further adjust the feature distribution on the basis of standardization to obtain the normalized first residual feature, which has stable statistical characteristics and is convenient for the learning of subsequent network layers.

[0030] The second residual block depth feature abstraction process uses a more complex network structure to perform further feature learning on the normalized first residual features. This residual block contains three convolutional layers to form a deeper feature extraction path. The first convolutional layer uses a 1x1 convolutional kernel to perform dimensional transformation on the input features. Its main function is to adjust the number of feature channels without changing the spatial dimensions. The second convolutional layer uses a 3x3 convolutional kernel for spatial feature extraction to capture the spatial correlation relationships between features. The third convolutional layer uses a 1x1 convolutional kernel again for dimensional recovery. ReLU activation functions are inserted between all three convolutional layers to enhance the non-linear expression ability. The combined calculation of multiple convolutional layers and activation functions enables the network to learn more abstract and complex feature representations, obtaining the second intermediate feature. This intermediate feature contains highly abstract semantic information after three non-linear transformations. The skip connection addition operation adds the second intermediate feature and the normalized first residual feature element by element, maintaining the continuity of feature transmission and obtaining the second residual output feature.

[0031] The global average pooling process converts the second residual output feature from a two-dimensional feature map into a one-dimensional feature vector. This process calculates the average value of the spatial dimension for each channel of the feature map separately. The specific operation is to add up the values at all positions in the two-dimensional feature map of each channel and then divide by the total number of positions to obtain the global average value of that channel. The global average values of all channels are arranged in order to form a one-dimensional feature vector. This pooling operation effectively compresses the spatial information of the features while retaining the semantic information in the channel dimension, eliminating the influence of position changes in the feature map on the classification result and making the network invariant to spatial transformations of the input. The high-level semantic feature representation obtained through the calculation of the spatial dimension average value is a compact vector containing the global information of all channels, which captures the final abstract representation of the plant stoichiometric features after deep network learning.

[0032] In a specific embodiment, the process of executing step S103 may specifically include the following steps: Collect data on the slope, aspect, altitude, and soil pH parameters in the karst fragile area, construct the collected data into a micro-topography parameter matrix, and perform encoding processing on the sunny slope, shady slope, semi-shady slope, and flat land types to obtain the niche type vector. Input the micro-topography parameter matrix and the niche type vector into a long short-term memory network for temporal feature learning processing, and control the information flow through the forget gate, input gate, and output gate to obtain the environmental factor temporal feature sequence. Perform bidirectional cyclic calculation processing based on the environmental factor temporal feature sequence, and perform weighted summation of the forward hidden state and the backward hidden state at each time step to obtain the bidirectional environmental impact feature representation. Apply the attention weight allocation mechanism to the two-way environmental impact feature representation for importance assessment processing. Calculate the attention scores of each environmental factor through the softmax function to obtain the microtopography impact weight value and the niche impact weight value.

[0033] Specifically, in the data collection process, key environmental parameters of the karst fragile area are obtained through topographic measurement equipment and soil detection instruments. The slope parameter is measured by a clinometer to obtain a value in degrees representing the surface tilt angle. The aspect parameter is determined by a compass or GPS device to obtain an azimuth angle value representing the slope aspect. The elevation parameter is obtained by an elevation measurement device to get the height value above sea level. The soil pH parameter is measured by a pH meter to obtain the pH value representing the acidity and alkalinity of the soil solution. These collected values are arranged in the order of sampling point location and time to construct a microtopography parameter matrix. The rows of this matrix represent different sampling points, and the columns represent four parameters: slope, aspect, elevation, and soil pH. The niche type coding process converts the qualitative habitat description into a numerical identifier. The sunny slope is coded as one, the shady slope is coded as two, the semi-shady slope is coded as three, and the flat ground is coded as four. Each sampling point corresponds to a coded value, and the coded values of all sampling points are arranged in order to form a niche type vector. This vector corresponds one-to-one with the microtopography parameter matrix in terms of spatial position.

[0034] The long short-term memory network time series feature learning process takes the microtopography parameter matrix and the niche type vector as input sequences for time series modeling. The network regards the environmental parameters of each sampling point as a time step in the time series. The forget gate calculates the forgetting probability through the sigmoid activation function. This probability determines which information in the previous cell state needs to be discarded. The input of the forget gate includes the environmental parameters at the current time step and the hidden state at the previous time step. After transformation by the weight matrix, it outputs a forgetting coefficient between zero and one through the sigmoid function. The input gate also uses the sigmoid function to calculate the input probability, determining which parts of the current input information need to be stored in the cell state. The input gate also includes a tanh layer for creating candidate update values, which non-linearly transforms the current input and the previous hidden state. The output gate controls which information in the cell state needs to be output to the hidden state. After calculating the output probability through the sigmoid function, it multiplies the cell state activated by tanh. The cooperation of the three gating mechanisms enables the network to selectively remember and forget environmental information at different times, obtaining an environmental factor time series feature sequence, which contains a comprehensive representation of the environmental parameters at each time step.

[0035] The bidirectional cyclic calculation process uses recurrent neural networks in both forward and backward directions to simultaneously process the time-series feature sequence of environmental factors. The forward recurrent network calculates in the order from the first sampling point to the last sampling point according to the time step. The hidden state at each time step is calculated based on the current input and the hidden state at the previous moment. The backward recurrent network calculates in the reverse time order from the last sampling point to the first sampling point. The hidden state at each time step is calculated based on the current input and the hidden state at the next moment. The forward hidden state and the backward hidden state respectively capture the historical and future dependencies of environmental factors. The weighted sum operation is performed at each time step to linearly combine the forward hidden state and the backward hidden state at the corresponding time step through learnable weight coefficients. The weight coefficients are automatically learned through network training. After the weighted sum, a bidirectional environmental impact feature representation is obtained, which integrates the bidirectional dependency information of environmental factors in time series.

[0036] The importance evaluation process of the attention weight allocation mechanism calculates the importance weights of the impacts of different environmental factors on plant adaptation strategies. First, the bidirectional environmental impact feature representation is input into the fully connected layer for linear transformation to obtain the importance scores of each environmental factor. These scores reflect the impact intensity of the corresponding environmental factors on plant growth and adaptation. The softmax function normalizes the importance scores of all environmental factors. The exponential value of each score is calculated through the exponential function, and then each exponential value is divided by the sum of all exponential values to obtain the attention scores of each environmental factor. The value range of the attention scores is between zero and one, and the sum of all scores is equal to one. The microtopography impact weight value is obtained by weighted averaging the attention scores corresponding to the three microtopography parameters of slope, aspect, and elevation. The niche impact weight value corresponds to the attention score of the niche type parameter. These two weight values quantitatively reflect the relative importance of the impacts of microtopography and niche on plant stoichiometric characteristics.

[0037] In a specific embodiment, the process of executing step S104 may specifically include the following steps: Perform a restrictive analysis process on the nitrogen, phosphorus, and potassium element contents in the plant stoichiometric characteristic matrix. Identify nutrient limiting factors through the minimum value determination algorithm, and extract the calcium and magnesium element contents as adjustment variables to obtain a nutrient limiting factor dataset; Input the nutrient limiting factor dataset into the generative adversarial network for ratio optimization processing. Calculate the optimal ratio relationship between calcium, magnesium elements and the limiting nutrient factors through the deep fully connected layer to obtain an initial nutrient ratio scheme; Perform a strategy matching process on the initial nutrient ratio scheme based on the plant adaptation strategy classification results. Use the conservative strategy label and the resource acquisition strategy label as constraint conditions to adjust the ratio parameters to obtain a strategy matching nutrient ratio scheme; Spatially differentiate the strategy-matching nutrient ratio plan according to the microtopography influence weight value and the niche influence weight value, calculate the special ratio coefficients respectively according to different topographic conditions and habitat types, and obtain the ecological restoration implementation plan for the karst fragile area.

[0038] Specifically, perform a limiting analysis on the nitrogen, phosphorus, and potassium element contents in the plant stoichiometric characteristics matrix. First, extract three columns of data on the nitrogen element content, phosphorus element content, and potassium element content from the stoichiometric characteristics matrix. The limiting analysis is based on the principle of Liebig's law of the minimum, which states that the growth rate of plants is limited by the scarcest nutrient element. The minimum value determination algorithm determines the limiting factor by comparing the relative contents of nitrogen, phosphorus, and potassium. The specific calculation process is to divide the nitrogen, phosphorus, and potassium contents of each plant by their respective plant demand thresholds to obtain the relative sufficiency. The demand threshold is a standard value determined according to the minimum nutrient concentration required for normal plant growth. The element with the smallest relative sufficiency is the limiting nutrient factor for that plant. At the same time, extract the calcium element content and magnesium element content from the stoichiometric characteristics matrix as adjustment variables. Calcium and magnesium elements have unique regulatory effects in the karst area. Calcium elements participate in cell wall construction and signal transduction processes, and magnesium elements are the core atoms of chlorophyll molecules and participate in various enzyme reactions. Combine the identified limiting nutrient factor types, the corresponding element content values, and the calcium and magnesium element content values to form a nutrient limiting factor dataset, which contains the limiting factor identification and relevant element concentration information for each plant.

[0039] The generative adversarial network ratio optimization process uses the nutrient limiting factor dataset as input for optimal ratio calculation. The generative adversarial network consists of two neural networks, a generator and a discriminator. The generator is responsible for generating the nutrient ratio plan, and the discriminator is responsible for evaluating the rationality of the ratio plan. The generator first concatenates the nutrient limiting factor data with a random noise vector to expand the input dimension, and then performs a linear transformation through the first-layer deep fully connected layer. This layer multiplies the input vector by a weight matrix and adds a bias vector, and then passes through the ReLU activation function to introduce non-linear characteristics. The second fully connected layer continues to transform and abstract the features. The third fully connected layer outputs the ratio coefficients of calcium and magnesium elements to the limiting nutrient factors. The ratio coefficient reflects the quantitative relationship between the added amounts of calcium and magnesium elements and the missing amounts of the limiting factors. The discriminator network receives the generated ratio plan and evaluates its authenticity. Through the adversarial training process, the generator learns to generate more reasonable ratio plans, and the discriminator learns to distinguish real ratios from generated ratios. The two networks play against each other until they reach a balanced state, and an initial nutrient ratio plan is obtained, which contains the optimal addition ratios of calcium and magnesium elements for different limiting factors.

[0040] Performing strategy matching processing based on the classification results of plant adaptation strategies is to combine the initial nutrient ratio scheme with the plant survival strategy for personalized adjustment. The conservative strategy label corresponds to plants that adopt a cautious resource utilization method. Such plants usually grow slowly but have strong stress resistance. The resource acquisition strategy label corresponds to plants that actively acquire nutrient resources. Such plants grow rapidly but have higher requirements for environmental conditions. The strategy matching processing adjusts the ratio parameters according to different strategy labels. For conservative strategy plants, the addition concentration of calcium and magnesium elements is reduced to avoid physiological stress caused by excessive nutrients. For resource acquisition strategy plants, the addition amount of calcium and magnesium elements is appropriately increased to meet their nutritional requirements for rapid growth. The adjustment process is achieved through preset strategy coefficients. The conservative strategy coefficient is less than one and is used to reduce the ratio concentration. The resource acquisition strategy coefficient is greater than one and is used to increase the ratio concentration. The strategy coefficient is multiplied by the calcium and magnesium addition amounts in the initial ratio scheme to obtain the adjusted ratio parameters, forming a strategy matching nutrient ratio scheme.

[0041] Performing spatial differentiation processing according to the microtopography influence weight value and the microhabitat influence weight value takes into account the influence of environmental heterogeneity on the nutrient ratio scheme. The microtopography influence weight value reflects the influence degree of topographic factors such as slope, aspect, and altitude on the nutrient requirements of plants. The microhabitat influence weight value reflects the action intensity of small-scale environmental factors such as light and water. The spatial differentiation processing calculates special ratio coefficients according to different topographic conditions and habitat types. For the upper slope position, due to poor soil and water loss, the nutrient addition amount needs to be increased, and the ratio coefficient is set to a value greater than one. For the lower slope position and depression, due to nutrient enrichment and sufficient water, the addition amount can be appropriately reduced, and the ratio coefficient is set to a value less than one. For the sunny slope, due to strong sunlight and large evaporation, the calcium and magnesium regulation effect needs to be enhanced. For the shady slope, due to the relatively mild environment, the regulation intensity can be reduced. The microtopography weight value is multiplied by the topographic coefficient to obtain the topographic regulation factor. The microhabitat weight value is multiplied by the habitat coefficient to obtain the habitat regulation factor. The final ratio scheme is calculated by multiplying the strategy matching ratio scheme by the topographic regulation factor and the habitat regulation factor, forming a differentiated nutrient regulation scheme for different spatial positions, which constitutes the ecological restoration implementation scheme for the karst fragile area.

[0042] In a specific embodiment, the process of performing ratio optimization processing by inputting the nutrient limitation factor dataset into the generative adversarial network may specifically include the following steps: Performing random noise vector splicing processing on the nutrient limitation factor dataset, combining the nitrogen, phosphorus, and potassium limitation factor values with the Gaussian noise vector by dimension to obtain the generator input data; Inputting the generator input data into the first fully connected layer for linear transformation processing, calculating the output of the first hidden layer through weight matrix multiplication and bias vector addition, and obtaining the first non-linear feature through activation function processing; Perform forward propagation processing on the multi-layer fully connected network based on the first non-linear feature, calculate the features sequentially through the second fully connected layer and the third fully connected layer, and apply batch normalization to the output of each layer to obtain a candidate calcium-magnesium ratio scheme; Perform discriminator authenticity evaluation processing on the candidate calcium-magnesium ratio scheme, calculate the difference degree between the generated scheme and the true ratio through the adversarial loss function, and adjust the network parameters according to gradient backpropagation to obtain the initial nutrient ratio scheme.

[0043] Specifically, extract the nitrogen, phosphorus, and potassium limiting factor values from the nutrient limiting factor dataset. These values include the type identifier of the limiting factor and the corresponding element content values. The nitrogen limiting factor value represents the relative deficiency degree of nitrogen element, the phosphorus limiting factor value represents the relative deficiency degree of phosphorus element, and the potassium limiting factor value represents the relative deficiency degree of potassium element. The Gaussian noise vector is a numerical sequence randomly sampled from the standard normal distribution. The dimension of this noise vector is usually set to be the same as or higher than the dimension of the limiting factor data. The introduction of Gaussian noise enables the generator to produce diverse output results rather than a fixed mapping relationship. The dimension merging operation concatenates the nitrogen, phosphorus, and potassium limiting factor values with the Gaussian noise vector in the feature dimension. Specifically, the limiting factor data vector and the noise vector are arranged column by column to form a longer vector. The length of the concatenated vector is equal to the sum of the original limiting factor data dimension and the noise vector dimension, obtaining the generator input data which contains the original nutrient deficiency information and random perturbation information, providing rich input features for the subsequent ratio generation process. The linear transformation processing of the first fully connected layer performs mathematical operations on the generator input data through the weight matrix and the bias vector. The weight matrix is a parameter matrix learned during the network training process, and its dimension is the input data length multiplied by the number of neurons in the first hidden layer. The weight matrix multiplication operation multiplies the input data vector with the weight matrix. After each input element is multiplied by the corresponding weight and summed, the pre-activation value of the hidden layer neuron is obtained. The bias vector addition operation adds the result of the weight matrix multiplication to the bias vector element by element. The dimension of the bias vector is equal to the number of neurons in the first hidden layer, and each bias value corresponds to a hidden layer neuron. The mathematical expression of the linear transformation is that the hidden layer output is equal to the input data multiplied by the weight matrix plus the bias vector, obtaining the first hidden layer output which is a numerically transformed vector. The activation function processing transforms the first hidden layer output through a non-linear activation function. Commonly used activation functions include the ReLU function, the Sigmoid function, or the Tanh function. The role of the activation function is to introduce non-linearity so that the network can learn complex non-linear mapping relationships, obtaining the first non-linear feature which contains the feature representation after linear transformation and non-linear activation.

[0044] The forward propagation process of the multi-layer fully connected network continues to perform deep feature learning based on the first non-linear features. The second fully connected layer receives the first non-linear features as input, performs a linear transformation by multiplying with the second layer weight matrix and adding the second layer bias vector, and then obtains the second non-linear output after being processed by the activation function. The third fully connected layer continues to perform the same linear transformation and non-linear activation processing on the output of the second layer. The calculation process of each layer follows the same pattern of weight matrix multiplication plus bias vector addition and then passing through the activation function. Batch normalization is applied after the output of each layer. Batch normalization first calculates the mean and variance of the current batch of data on each feature dimension, then normalizes each feature value by subtracting the mean and dividing by the standard deviation, and then adjusts the normalized result through learnable scaling parameters and translation parameters. The role of batch normalization is to stabilize the training process and accelerate network convergence. The cascaded processing of the multi-layer network makes the features abstract and transform layer by layer. Finally, the output of the third fully connected layer corresponds to the candidate values of the calcium-magnesium ratio, and the obtained candidate calcium-magnesium ratio scheme contains the recommended addition ratios of calcium and magnesium elements for the current input conditions.

[0045] The discriminator authenticity evaluation process judges the authenticity of the generated candidate calcium-magnesium ratio scheme through an independent discriminator network. The discriminator network receives the candidate ratio scheme as input and calculates the authenticity probability score through a multi-layer fully connected network structure. This score reflects the rationality and credibility of the ratio scheme. The adversarial loss function calculates the difference degree between the generated scheme and the real ratio data. The loss function usually adopts the form of binary cross-entropy or least square error. The goal of the generator is to minimize the probability that the discriminator identifies the generated data as fake, and the goal of the discriminator is to maximize the ability to distinguish real data and generated data. The gradient backpropagation adjusts the network parameters by calculating the gradient of the loss function with respect to the network weights to update the parameter values. The gradient calculation uses the chain rule to propagate layer by layer from the output layer to the input layer. The weight update amount of each layer is equal to the learning rate multiplied by the corresponding gradient value. The generator and the discriminator alternately update the parameters. The generator adjusts its generation strategy according to the feedback of the discriminator, and the discriminator adjusts its discrimination ability according to the new generated data. After multiple rounds of adversarial training, the network reaches an equilibrium state. The generator can generate a ratio scheme close to the real distribution, and the obtained initial nutrient ratio scheme is the optimal addition ratio of calcium and magnesium elements optimized through adversarial training.

[0046] The above describes the method for ecological restoration in karst vulnerable areas based on measurement features in the embodiments of the present application. Next, the ecological restoration system in karst vulnerable areas based on measurement features in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the ecological restoration system in karst vulnerable areas based on measurement features in the embodiments of the present application includes: The detection module 201 is used to detect nutrient elements of plants in the karst vulnerable area through a multispectral sensor, obtain data on the contents of carbon, nitrogen, phosphorus, potassium, calcium, and magnesium, and obtain a plant stoichiometric characteristic matrix through standardization processing; The recognition module 202 is used to input the plant stoichiometric characteristic matrix into a deep convolutional neural network for pattern recognition processing, output conservative strategy labels and resource acquisition strategy labels, and obtain a plant adaptation strategy classification result; The quantification module 203 is used to perform weighted quantification analysis on environmental impact factors according to microtopography parameters and niche data, and calculate the microtopography impact weight value and the niche impact weight value; The proportioning module 204 is used to optimize the proportion calculation of limiting nutrient factors according to the calcium and magnesium element regulation mechanism, combine the plant adaptation strategy classification result to generate a targeted nutrient regulation plan, and obtain an ecological restoration implementation plan for the karst vulnerable area.

[0047] above Figure 2 The karst vulnerable area ecological restoration system based on stoichiometric characteristics in the embodiments of the present invention is described in detail from the perspective of modular functional entities. Next, the karst vulnerable area ecological restoration device based on stoichiometric characteristics in the embodiments of the present invention is described in detail from the perspective of hardware processing.

[0048] Refer to Figure 3 , and an ecological restoration device for the karst vulnerable area based on stoichiometric characteristics is further provided in the embodiments of the present invention. The ecological restoration device for the karst vulnerable area based on stoichiometric characteristics can be a server, and its internal structure can be as Figure 3 shown. The ecological restoration device for the karst vulnerable area based on stoichiometric characteristics includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of this computer design is used to provide computing and control capabilities. The memory of the ecological restoration device for the karst vulnerable area based on stoichiometric characteristics includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the ecological restoration device for the karst vulnerable area based on stoichiometric characteristics is used to store the corresponding data in this embodiment. The network interface of the ecological restoration device for the karst vulnerable area based on stoichiometric characteristics is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0049] Those skilled in the art can understand that Figure 3 the structure shown in

[0050] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are run on a computer, the computer is caused to execute the steps of the method for ecological restoration of karst vulnerable areas based on measurement features.

[0051] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, system, and unit can refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0052] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions to cause a device for ecological restoration of karst vulnerable areas based on measurement features (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0053] The above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An ecological restoration method for karst vulnerable areas based on measurement characteristics, characterized in that, The method includes: Detecting nutrient elements of plants in the karst vulnerable area through a multispectral sensor, obtaining carbon, nitrogen, phosphorus, potassium, calcium, and magnesium content data, and obtaining a plant stoichiometric characteristic matrix through standardization processing; Inputting the plant stoichiometric characteristic matrix into a deep convolutional neural network for pattern recognition processing, outputting conservative strategy labels and resource acquisition strategy labels, and obtaining a plant adaptation strategy classification result; Performing weighted quantification analysis on environmental impact factors according to microtopography parameters and niche data, and obtaining a microtopography impact weight value and a niche impact weight value through calculation; Optimizing the ratio calculation of limiting nutrient factors according to the calcium and magnesium element regulation mechanism, and generating a targeted nutrient regulation plan in combination with the plant adaptation strategy classification result, and obtaining an ecological restoration implementation plan for the karst vulnerable area.

2. The ecological restoration method for Karst vulnerable areas based on measurement features according to claim 1, characterized in that, The detecting nutrient elements of plants in the karst vulnerable area through a multispectral sensor, obtaining carbon, nitrogen, phosphorus, potassium, calcium, and magnesium content data, and obtaining a plant stoichiometric characteristic matrix through standardization processing includes: Collecting spectral data of plants in different functional groups in the karst vulnerable area, inputting the spectral reflectance data into an element content inversion algorithm for quantitative analysis, and obtaining the original content values of carbon, nitrogen, phosphorus, potassium, calcium, and magnesium elements of a single plant; Calculating the stoichiometric ratio between elements based on the original element content values to obtain a plant individual stoichiometric data set; Performing outlier detection and missing value filling processing on the plant individual stoichiometric data set, and reconstructing the processed data into a matrix according to the functional group category to obtain a classified stoichiometric data matrix; Performing zero-mean standardization and maximum-minimum normalization processing on the classified stoichiometric data matrix, and constructing a sample feature vector in combination with the microtopography position label and the niche type label to obtain a plant stoichiometric characteristic matrix.

3. The ecological restoration method for karst vulnerable areas based on measurement features according to claim 1, characterized in that The inputting the plant stoichiometric characteristic matrix into a deep convolutional neural network for pattern recognition processing, outputting conservative strategy labels and resource acquisition strategy labels, and obtaining a plant adaptation strategy classification result includes: Inputting the plant stoichiometric characteristic matrix into a multi-scale convolutional layer for feature extraction processing, extracting local stoichiometric relationship features through a 3×3 convolutional kernel and global nutrient distribution features through a 7×7 convolutional kernel, and obtaining a multi-dimensional feature map; Performing attention mechanism weighted fusion processing on the multi-dimensional feature map, combining and calculating the local stoichiometric relationship features and the global nutrient distribution features according to the weight coefficient, and obtaining a fused feature vector; Inputting the fused feature vector into a residual network structure for deep feature learning processing, and optimizing the gradient propagation process through skip connection and batch normalization operations to obtain a high-level semantic feature representation; Performing softmax classifier calculation processing based on the high-level semantic feature representation, mapping the feature representation to a conservative strategy probability value and a resource acquisition strategy probability value, and obtaining a plant adaptation strategy classification result.

4. The ecological restoration method for Karst fragile areas based on measurement characteristics according to claim 3, characterized in that, The inputting the fused feature vector into a residual network structure for deep feature learning processing, and optimizing the gradient propagation process through skip connection and batch normalization operations to obtain a high-level semantic feature representation includes: Input the fused feature vector into the first residual block for non-linear transformation processing. Calculate the first intermediate feature through a convolutional layer and an activation function, and perform a skip connection addition operation with the fused feature vector to obtain the first residual output feature; Perform batch normalization processing on the first residual output feature, calculate the mean and variance of the feature data along the batch dimension for standardization transformation, and obtain the normalized first residual feature; Input the normalized first residual feature into the second residual block for deep feature abstraction processing. Calculate the second intermediate feature through a combination of multiple convolutional layers and non-linear activation functions, and perform a skip connection addition operation with the normalized first residual feature to obtain the second residual output feature; Perform global average pooling processing on the second residual output feature, compress the two-dimensional feature map into a one-dimensional feature vector, and calculate the high-level semantic feature representation through the average value of the spatial dimension.

5. The ecological restoration method for Karst vulnerable areas based on measurement features according to claim 1, characterized in that, The weight quantization analysis of the environmental impact factors according to the microtopography parameters and niche data, and calculating the microtopography impact weight value and the niche impact weight value, includes: Collect data on the slope, aspect, elevation, and soil pH parameters in the karst vulnerable area, construct the collected data into a microtopography parameter matrix, and perform coding processing on the sunny slope, shady slope, semi-shady slope, and flat land types to obtain the niche type vector; Input the microtopography parameter matrix and the niche type vector into a long short-term memory network for temporal feature learning processing, and control the information flow through the forget gate, input gate, and output gate to obtain the environmental factor temporal feature sequence; Perform bidirectional recurrent calculation processing based on the environmental factor temporal feature sequence, perform weighted summation of the forward hidden state and the backward hidden state at each time step, and obtain the bidirectional environmental impact feature representation; Apply the attention weight allocation mechanism to the bidirectional environmental impact feature representation for importance evaluation processing, calculate the attention scores of each environmental factor through the softmax function, and obtain the microtopography impact weight value and the niche impact weight value.

6. The ecological restoration method for Karst vulnerable areas based on measurement characteristics according to claim 1, characterized in that, The optimized ratio calculation of the limiting nutrient factors according to the calcium and magnesium element regulation mechanism, and generating a targeted nutrient regulation plan in combination with the plant adaptation strategy classification result to obtain the ecological restoration implementation plan for the karst vulnerable area, includes: Perform limiting analysis processing on the nitrogen, phosphorus, and potassium element contents in the plant stoichiometry feature matrix, identify the nutrient limiting factors through the minimum value determination algorithm, and extract the calcium and magnesium element contents as the regulation variables to obtain the nutrient limiting factor data set; Input the nutrient limiting factor data set into a generative adversarial network for ratio optimization processing, calculate the optimal ratio relationship between calcium and magnesium elements and the limiting nutrient factors through a deep fully connected layer, and obtain the initial nutrient ratio plan; Perform strategy matching processing on the initial nutrient ratio plan based on the plant adaptation strategy classification result, use the conservative strategy label and the resource acquisition strategy label as constraint conditions to adjust the ratio parameters, and obtain the strategy matching nutrient ratio plan; Perform spatial differentiation processing on the strategy matching nutrient ratio plan according to the microtopography impact weight value and the niche impact weight value, calculate the dedicated ratio coefficients for different terrain conditions and habitat types respectively, and obtain the ecological restoration implementation plan for the karst vulnerable area.

7. The method for ecological restoration in karst fragile areas based on measurement characteristics according to claim 6, characterized in that, Input the nutrient limitation factor dataset into the generative adversarial network for ratio optimization processing, and calculate the optimal ratio relationship between calcium and magnesium elements and the limiting nutrient factors through a deep fully connected layer to obtain an initial nutrient ratio scheme, including: Perform random noise vector splicing processing on the nutrient limitation factor dataset, and merge the nitrogen, phosphorus, and potassium limitation factor values with the Gaussian noise vector by dimension to obtain the input data for the generator; Input the input data for the generator into the first fully connected layer for linear transformation processing, calculate the output of the first hidden layer through weight matrix multiplication and bias vector addition, and obtain the first non-linear feature through activation function processing; Based on the first non-linear feature, perform forward propagation processing of a multi-layer fully connected network, calculate the features sequentially through the second fully connected layer and the third fully connected layer, and apply batch normalization to the output of each layer to obtain a candidate calcium-magnesium ratio scheme; Perform discriminator authenticity evaluation processing on the candidate calcium-magnesium ratio scheme, calculate the difference degree between the generated scheme and the true ratio through the adversarial loss function, and adjust the network parameters according to gradient backpropagation to obtain the initial nutrient ratio scheme.

8. An ecological restoration system for karst vulnerable areas based on measurement characteristics, characterized in that, For implementing the metrology feature-based ecological restoration method for karst vulnerable areas as described in any one of claims 1-7, the metrology feature-based ecological restoration system for karst vulnerable areas includes: A detection module for detecting nutrient element content of plants in karst vulnerable areas through a multispectral sensor, obtaining carbon, nitrogen, phosphorus, potassium, calcium, and magnesium content data, and obtaining a plant stoichiometric feature matrix through standardization processing; An identification module for inputting the plant stoichiometric feature matrix into a deep convolutional neural network for pattern recognition processing, outputting conservative strategy labels and resource acquisition strategy labels, and obtaining a plant adaptation strategy classification result; A quantification module for performing weighted quantification analysis on environmental impact factors according to microtopography parameters and microhabitat data, and obtaining a microtopography impact weight value and a microhabitat impact weight value through calculation; A ratio module for performing optimized ratio calculation on the limiting nutrient factors according to the calcium and magnesium element regulation mechanism, and generating a targeted nutrient regulation scheme in combination with the plant adaptation strategy classification result to obtain an ecological restoration implementation scheme for karst vulnerable areas.

9. An ecological restoration device for karst vulnerable areas based on measurement characteristics, characterized in that, It includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the metrology feature-based ecological restoration method for karst vulnerable areas as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, the processor is caused to execute the metrology feature-based ecological restoration method for karst vulnerable areas as described in any one of claims 1 to 7.

Images