Rural ecological restoration wasteland identification method

Abstract

The invention provides a rural ecological restoration wasteland identification method, which comprises the steps of determining a boundary range of a wasteland to be monitored, and collecting multi-dimensional data of the wasteland to be monitored, the multi-dimensional data comprising electrical data, smell data and voiceprint data; preprocessing the multi-dimensional data, performing feature extraction on the preprocessed multi-dimensional data, and constructing a three-dimensional tensor matrix; constructing a tensor decomposition and matrix regression fusion model, and predicting the restoration progress of the wasteland to be monitored based on the three-dimensional tensor matrix to obtain a prediction result; based on the comparison of the prediction result and the restoration requirement of the to-be-monitored wasteland, the restoration strategy of the to-be-monitored wasteland is dynamically adjusted, the limitation of a traditional single data source can be broken through by fusing the plant rhizosphere electric signals, the soil odor fingerprints and the voiceprint data, the recognition precision is greatly improved, and a tensor decomposition and matrix regression fusion model is introduced, so that the recognition precision is improved. The problem of multi-dimensional heterogeneous data fusion can be effectively solved, a scientific basis is provided for restoration decision making, and sustainable restoration of a rural ecosystem is promoted.

Description

Technical Field

[0001] This invention relates to the field of ecological restoration technology, and in particular to a method for identifying wasteland in rural ecological restoration. Background Technology

[0002] Rural ecological restoration is an important measure to improve the rural living environment and promote sustainable development. Accurate identification of wasteland is a prerequisite for restoration work. Traditional wasteland identification relies mainly on remote sensing image interpretation or on-site sampling surveys. The former is easily affected by vegetation cover and weather conditions, making it difficult to distinguish plots with similar degrees of degradation; the latter is time-consuming and labor-intensive, and relies on human experience, making it difficult to achieve large-scale dynamic monitoring. With the development of ecological restoration technology, a single data source can no longer meet the accuracy requirements of wasteland identification. Especially in complex rural topography, the collaborative analysis of multi-dimensional information such as soil quality, vegetation vitality, and biological activity has become the key to improving the accuracy of identification. In existing technologies, detection methods based on soil physicochemical properties require frequent sampling and testing, resulting in poor timeliness; identification methods relying on plant appearance characteristics are significantly affected by seasonal changes. Meanwhile, new data sources such as plant rhizosphere electrical signals, soil odor fingerprints, and acoustic signatures have not been fully utilized. Rhizosphere electrical signals can reflect plant metabolic activity and soil stress response, soil volatile organic compound composition can indicate microbial activity and pollution status, and acoustic signatures are associated with the level of regional biodiversity. These data characterize the ecological state of wasteland from different dimensions, but their collaborative application faces problems such as difficulty in data fusion and weak feature correlation. Therefore, how to integrate multidimensional heterogeneous data and construct accurate and efficient wasteland identification methods has become a technical problem that urgently needs to be solved in the field of rural ecological restoration. Summary of the Invention

[0003] In view of this, the present invention proposes a method for identifying wasteland in rural ecological restoration, which integrates multi-dimensional data of electricity, odor and voiceprint, and can accurately identify the progress of wasteland restoration so as to dynamically adjust the restoration strategy.

[0004] The technical solution of this invention is implemented as follows: A method for identifying wasteland for rural ecological restoration includes the following steps: Step S1: Determine the boundary range of the wasteland to be monitored and collect multidimensional data of the wasteland to be monitored, including electrical data, odor data and voiceprint data; Step S2: Preprocess the multidimensional data, extract features from the preprocessed multidimensional data, and construct a three-dimensional tensor matrix; Step S3: Construct a fusion model of tensor decomposition and matrix regression, and predict the restoration progress of the wasteland to be monitored based on the three-dimensional tensor matrix to obtain the prediction results; Step S4: Based on the comparison between the prediction results and the restoration requirements of the wasteland to be monitored, dynamically adjust the restoration strategy of the wasteland to be monitored.

[0005] Preferably, the specific steps for determining the boundary range of the wasteland to be monitored in step S1 are as follows: Step S11: Determine the boundary of the wasteland to be monitored by combining on-site surveys with remote sensing images.

[0006] Preferably, the specific steps for collecting electrical data of the wasteland to be monitored in step S1 are as follows: Step S12: In the wasteland to be monitored, select representative plants based on their growth status and distribution density. Step S13: Place the implanted microelectrode array in the soil around the rhizosphere of a representative plant and collect the changes in electrical signals in the rhizosphere microdomain as electrical data.

[0007] Preferably, the specific steps for collecting odor data of the wasteland to be monitored in step S1 are as follows: Step S14: Grid the wasteland to be monitored, and select the grid on one of the diagonals of the wasteland to be monitored as the first collection point array; Step S15: Extend from the first collection point array to the other two corners of the wasteland to be monitored, and select a second collection point array composed of several grids at certain intervals, parallel to the first collection point array; Step S16: Collect odor samples from the headspace of the grid soil on the first and second collection point arrays as odor data.

[0008] Preferably, the specific steps for collecting voiceprint data of the wasteland to be monitored in step S1 are as follows: Step S17: Distribute a distributed microphone array evenly in the wasteland to be monitored; Step S18: Collect voiceprint data through a distributed microphone array.

[0009] Preferably, the specific steps for preprocessing the multidimensional data in step S2 are as follows: Step S21: Wavelet transform is used to process the electrical data to remove high-frequency noise and low-frequency interference. Moving average method is used to smooth the denoised electrical data, and linear interpolation method is used to fill in missing values. Step S22: Remove baseline drift in odor data, correct the baseline using polynomial fitting, extract characteristic peaks from odor data using partial least squares discriminant analysis, and standardize the characteristic peaks. Step S23: Use Fourier transform to convert the voiceprint data to the frequency domain, remove environmental noise, and then perform frame processing. Add a Hanning window to each frame of data, and use the endpoint detection method to detect the start and end points of the voiceprint signal, and remove invalid silent parts.

[0010] Preferably, the specific steps in step S2 for feature extraction from the preprocessed multidimensional data and construction of a three-dimensional tensor matrix are as follows: Step S24: Extract time-domain features, frequency-domain features, and time-frequency-domain features from the preprocessed electrical data, and construct an electrical column matrix; Step S25: Perform cluster analysis on the preprocessed odor data, use the K-means clustering algorithm to cluster, and construct a soil odor fingerprint database. Extract the position and intensity of the characteristic peaks from the soil odor fingerprint database and construct an odor column matrix. Step S26: Process the voiceprint data using a Mel filter, extract the Mel cepstral coefficients, calculate the soundscape complexity index of the voiceprint data, extract the spectral centroid of the voiceprint data, and construct the voiceprint column matrix. Step S27: Construct a three-dimensional tensor matrix from the electrical column matrix, the odor column matrix, and the voiceprint column matrix.

[0011] Preferably, the tensor decomposition and matrix regression fusion model includes a tensor decomposition part and a matrix regression part. The tensor decomposition part is used to extract latent features from a three-dimensional tensor matrix, and the matrix regression part is used to predict the restoration progress of the wasteland to be monitored based on the extracted latent features.

[0012] Preferably, the specific steps of step S3 are as follows: Step S31: The tensor decomposition part uses Tucker decomposition to decompose the three-dimensional tensor matrix into core tensors, factor matrices, feature dimensions, and data type dimensions. Step S32: Extract shared features across data types from the factor matrix, obtain feature interaction information from the core tensor, and form a structured feature matrix; Step S33: The matrix regression part has a kernel matrix regression model, which takes the structured feature matrix as input and outputs a repair progress label. The repair progress is predicted based on the repair progress label.

[0013] Preferably, the specific steps of step S4 are as follows: Step S41: Obtain the restoration requirements of the wasteland to be monitored, and determine the expected restoration result of the wasteland to be monitored at the current moment based on the restoration requirements; Step S42: If the prediction result is worse than the expected repair result, determine the repair period based on the prediction result; Step S43: Dynamically select one of the following restoration strategies based on the different restoration periods: soil improvement, vegetation planting, irrigation management, vegetation optimization, pest and disease control, soil monitoring and regulation, ecological maintenance, and biodiversity conservation.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention provides a method for identifying wasteland in rural ecological restoration. By integrating plant rhizosphere electrical signals, soil odor fingerprints, and voiceprint data, it comprehensively characterizes the ecological state of wasteland from multiple dimensions, including plant growth vitality, soil microbial activity, and biodiversity. This method breaks through the limitations of traditional single data sources and significantly improves identification accuracy. It can accurately distinguish adjacent wastelands with similar degrees of degradation by observing subtle changes in the characteristic peaks of volatile organic compounds in soil odor fingerprints. With the help of advanced sensor technology and data processing algorithms, it can achieve real-time and dynamic monitoring of large areas of rural wasteland. Compared with traditional field sampling surveys, it is more efficient and can detect plant growth stress in a timely manner, providing a basis for early intervention. Furthermore, the tensor decomposition and matrix regression fusion model effectively solves the problem of multidimensional heterogeneous data fusion, automatically extracts key features, reduces human subjectivity, and is not dependent on specific geographical environments or climate conditions. It is applicable to rural wastelands with different landforms and vegetation types, and has strong adaptability. At the same time, the accurate identification results can provide a scientific basis for restoration decisions, help formulate targeted strategies, improve restoration efficiency, reduce costs, and promote the sustainable restoration of rural ecosystems. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of a method for identifying wasteland in rural ecological restoration according to the present invention; Figure 2 This is a flowchart of step S1 of a method for identifying wasteland in rural ecological restoration according to the present invention; Figure 3 This is a flowchart of step S2 of the rural ecological restoration wasteland identification method of the present invention; Figure 4 This is a flowchart of step S3 of a method for identifying wasteland in rural ecological restoration according to the present invention; Figure 5 This is a flowchart of step S4 of a method for identifying wasteland in rural ecological restoration according to the present invention. Detailed Implementation

[0017] To better understand the technical content of this invention, a specific embodiment is provided below, and the invention will be further described in conjunction with the accompanying drawings.

[0018] See Figures 1 to 5 The present invention provides a method for identifying wasteland in rural ecological restoration, comprising the following steps: Step S1: Determine the boundary range of the wasteland to be monitored and collect multidimensional data of the wasteland to be monitored, including electrical data, odor data and voiceprint data; Step S2: Preprocess the multidimensional data, extract features from the preprocessed multidimensional data, and construct a three-dimensional tensor matrix; Step S3: Construct a fusion model of tensor decomposition and matrix regression, and predict the restoration progress of the wasteland to be monitored based on the three-dimensional tensor matrix to obtain the prediction results; Step S4: Based on the comparison between the prediction results and the restoration requirements of the wasteland to be monitored, dynamically adjust the restoration strategy of the wasteland to be monitored.

[0019] This invention discloses a method for identifying the restoration progress of rural wasteland in ecological restoration. The method first requires determining the specific location of the wasteland to be monitored. Since the methods for ecological restoration vary significantly depending on the type of wasteland, it is necessary to define the boundary range of the wasteland to be monitored and to establish the boundaries of different types of wasteland to ensure that the collected data corresponds to the wasteland being monitored. After determining the boundary range of the wasteland to be monitored, multidimensional data can be collected. This multidimensional data includes electrical data, odor data, and voiceprint data. Electrical data involves electrical signals generated by plant rhizosphere during growth; odor data involves odor changes caused by soil changes during wasteland restoration; and voiceprint data involves sound data produced by birds and insects at different stages of wasteland restoration. Combining data from different dimensions allows for accurate analysis of the wasteland restoration progress. After acquiring the multidimensional data, preprocessing and feature extraction can be performed. The extracted feature vectors can be used to construct a three-dimensional tensor matrix, which is then... The tensor matrix can be used as input to the constructed tensor decomposition and matrix regression fusion model. This model can decompose and regress the three-dimensional tensor matrix to predict the restoration progress of the monitored wasteland and obtain the prediction results. The restoration progress can be displayed as a percentage, for example, 50% restoration progress means that the restoration progress of the monitored wasteland has reached half. Then, the predicted restoration progress is compared with the restoration requirements of the monitored wasteland to determine whether the prediction results meet the restoration requirements. This allows for dynamic adjustment of the restoration strategy for the monitored wasteland, ensuring that the restoration of the monitored wasteland can be completed within the specified time, promoting the sustainable restoration of the rural ecosystem. Because multidimensional data is used to assess the restoration progress, the limitations of traditional single-dimensional data can be overcome, improving the accuracy of restoration progress identification. At the same time, the introduced tensor decomposition and matrix regression fusion model can effectively solve the problem of multidimensional heterogeneous data fusion, further ensuring the accuracy of the prediction results.

[0020] Preferably, the specific steps for determining the boundary range of the wasteland to be monitored in step S1 are as follows: Step S11: Determine the boundary of the wasteland to be monitored by combining on-site surveys with remote sensing images.

[0021] Before a restoration plan is developed for the wasteland to be monitored, a corresponding planning document can be generated. However, the boundary of the wasteland to be monitored may change during the planning process. Therefore, based on the planning document, on-site surveys and remote sensing imagery can be conducted to accurately determine the boundary of the wasteland to be monitored and ensure that the collected multidimensional data belong to the same wasteland to be monitored.

[0022] Preferably, the specific steps for collecting electrical data of the wasteland to be monitored in step S1 are as follows: Step S12: In the wasteland to be monitored, select representative plants based on their growth status and distribution density. Step S13: Place the implanted microelectrode array in the soil around the rhizosphere of a representative plant and collect the changes in electrical signals in the rhizosphere microdomain as electrical data.

[0023] Multidimensional data includes electrical data, which mainly includes electrical signals generated by ecological restoration plants planted on the wasteland to be monitored at different growth stages, such as their impact on soil temperature and humidity. Among these, the rhizosphere of plants has the greatest impact on the soil due to the absorption of water and nutrients. Therefore, when collecting electrical data, it is necessary to collect electrical signals from the plant rhizosphere. In order to ensure the representativeness of the samples, representative plants with suitable growth conditions and distribution densities are selected as samples from the wasteland to be monitored. Then, an implantable microelectrode array is buried in the soil and located in the soil around the plant rhizosphere. The changes in electrical signals in the micro-domain of the plant rhizosphere can be collected as electrical data through the implantable microelectrode array.

[0024] Preferably, the specific steps for collecting odor data of the wasteland to be monitored in step S1 are as follows: Step S14: Grid the wasteland to be monitored, and select the grid on one of the diagonals of the wasteland to be monitored as the first collection point array; Step S15: Extend from the first collection point array to the other two corners of the wasteland to be monitored, and select a second collection point array composed of several grids at certain intervals, parallel to the first collection point array; Step S16: Collect odor samples from the headspace of the grid soil on the first and second collection point arrays as odor data.

[0025] Odor data primarily consists of odor samples from the headspace near the soil. To ensure the representativeness of the odor data collection, sampling points need to be determined. The wasteland to be monitored can be approximated as a quadrilateral. After gridding the wasteland, each grid becomes a sampling point. Then, two corners of the wasteland are connected to form a diagonal, and the grids on the diagonal serve as the first sampling point array. Then, using the first sampling point array as the center, the grid is moved towards the other two corners of the wasteland, and grids on straight lines parallel to the first sampling point array are selected at equal intervals during the movement to form the second sampling point array. The grids on the second sampling point array also serve as odor sampling points for the soil. Finally, the odor in the headspace of the soil corresponding to the grids on the first and second sampling point arrays can be collected using odor sampling equipment, and the collected odor samples are used as odor data.

[0026] Preferably, the specific steps for collecting voiceprint data of the wasteland to be monitored in step S1 are as follows: Step S17: Distribute a distributed microphone array evenly in the wasteland to be monitored; Step S18: Collect voiceprint data through a distributed microphone array.

[0027] Voiceprint data mainly includes the sounds emitted by birds and insects, so a distributed microphone array can be evenly deployed to collect voiceprint data generated by birds and insects on the wasteland to be monitored.

[0028] Preferably, the specific steps for preprocessing the multidimensional data in step S2 are as follows: Step S21: Wavelet transform is used to process the electrical data to remove high-frequency noise and low-frequency interference. Moving average method is used to smooth the denoised electrical data, and linear interpolation method is used to fill in missing values. Step S22: Remove baseline drift in odor data, correct the baseline using polynomial fitting, extract characteristic peaks from odor data using partial least squares discriminant analysis, and standardize the characteristic peaks. Step S23: Use Fourier transform to convert the voiceprint data to the frequency domain, remove environmental noise, and then perform frame processing. Add a Hanning window to each frame of data, and use the endpoint detection method to detect the start and end points of the voiceprint signal, and remove invalid silent parts.

[0029] The main purpose of preprocessing multidimensional data is to remove the noise it contains and to fill in the missing data. In addition, the data of different dimensions are not uniform, so corresponding standardization processing is required.

[0030] Preferably, the specific steps in step S2 for feature extraction from the preprocessed multidimensional data and construction of a three-dimensional tensor matrix are as follows: Step S24: Extract time-domain features, frequency-domain features, and time-frequency-domain features from the preprocessed electrical data, and construct an electrical column matrix; Step S25: Perform cluster analysis on the preprocessed odor data, use the K-means clustering algorithm to cluster, and construct a soil odor fingerprint database. Extract the position and intensity of the characteristic peaks from the soil odor fingerprint database and construct an odor column matrix. Step S26: Process the voiceprint data using a Mel filter, extract the Mel cepstral coefficients, calculate the soundscape complexity index of the voiceprint data, extract the spectral centroid of the voiceprint data, and construct the voiceprint column matrix. Step S27: Construct a three-dimensional tensor matrix from the electrical column matrix, the odor column matrix, and the voiceprint column matrix.

[0031] After preprocessing, feature extraction is required for the three types of data. For electrical data, time-domain, frequency-domain, and time-frequency-domain features can be extracted and constructed into an electrical column matrix. For odor data, K-means clustering algorithm is used to cluster odors of the same type and construct a corresponding soil odor database. Then, the position and intensity of feature peaks can be extracted from the soil odor database to construct an odor column matrix. For voiceprint data, Mel filter is used to extract its Mel cepstral coefficients and the soundscape complexity index of the heating data is calculated at the same time. After extracting the spectral centroid of the heating data, a voiceprint array is constructed. Finally, the three column matrices are constructed into a three-dimensional tensor matrix. The number of elements in the three column matrices may differ, and missing elements are padded with zeros.

[0032] Preferably, the tensor decomposition and matrix regression fusion model includes a tensor decomposition part and a matrix regression part. The tensor decomposition part is used to extract latent features from the three-dimensional tensor matrix, and the matrix regression part is used to predict the restoration progress of the wasteland to be monitored based on the extracted latent features. The specific steps of step S3 are as follows: Step S31: The tensor decomposition part uses Tucker decomposition to decompose the three-dimensional tensor matrix into core tensors, factor matrices, feature dimensions, and data type dimensions. Step S32: Extract shared features across data types from the factor matrix, obtain feature interaction information from the core tensor, and form a structured feature matrix; Step S33: The matrix regression part has a kernel matrix regression model, which takes the structured feature matrix as input and outputs a repair progress label. The repair progress is predicted based on the repair progress label.

[0033] The tensor decomposition and matrix regression model is divided into two parts: tensor decomposition and matrix regression. These two parts perform different functions. The tensor decomposition part uses Tucker decomposition to decompose the three-dimensional tensor matrix into a core tensor, factor matrix, feature dimension, and data type dimension. Then, latent features are extracted from the decomposed factor matrix and core tensor to form a structured feature matrix. This structured feature matrix serves as input to the kernel matrix regression model within the matrix regression part. The kernel matrix regression model processes this data and outputs it to a restoration progress label. The restoration progress label represents the restoration progress at 100%, while a label of 50% indicates that the restoration is halfway complete. This fusion model of tensor decomposition and matrix regression solves the problem of multi-dimensional heterogeneous data fusion, accurately predicts restoration progress, provides a scientific basis for subsequent restoration decisions, and promotes the sustainable restoration of rural ecosystems.

[0034] Preferably, the specific steps of step S4 are as follows: Step S41: Obtain the restoration requirements of the wasteland to be monitored, and determine the expected restoration result of the wasteland to be monitored at the current moment based on the restoration requirements; Step S42: If the prediction result is worse than the expected repair result, determine the repair period based on the prediction result; Step S43: Dynamically select one of the following restoration strategies based on the different restoration periods: soil improvement, vegetation planting, irrigation management, vegetation optimization, pest and disease control, soil monitoring and regulation, ecological maintenance, and biodiversity conservation.

[0035] After obtaining the prediction results, the current restoration progress can be assessed. The restoration plan for wasteland includes restoration requirements, which specify the restoration expectations for different periods. Therefore, based on the restoration requirements, the expected restoration results for the wasteland to be monitored at the current moment can be determined. The expected restoration results can be compared with the prediction results. If the prediction results indicate that the current restoration progress is greater than the expected restoration progress, it means that the restoration of the wasteland to be monitored has exceeded expectations and no adjustment of the restoration strategy is needed. However, if the prediction results indicate that the current restoration progress is slower, the restoration strategy can be adjusted. The restoration strategies adopted at different restoration stages are also different. The prediction results can be used to determine the current restoration stage of the wasteland to be monitored, such as the initial, middle, and late stages. In the initial stage, restoration strategies such as soil improvement, vegetation planting, and irrigation management can be adopted. In the middle stage, restoration strategies such as vegetation optimization, pest and disease control, and soil monitoring and regulation can be adopted. In the late stage, restoration strategies such as ecological maintenance and biodiversity conservation can be adopted. By dynamically adjusting the restoration strategies, the restoration effect of the wasteland to be monitored can be improved, and the restoration can be accelerated to achieve the expected restoration requirements.

[0036] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying wasteland in rural ecological restoration, characterized in that, Includes the following steps: Step S1: Determine the boundary range of the wasteland to be monitored and collect multidimensional data of the wasteland to be monitored, including electrical data, odor data and voiceprint data; Step S2: Preprocess the multidimensional data, extract features from the preprocessed multidimensional data, and construct a three-dimensional tensor matrix; Step S3: Construct a fusion model of tensor decomposition and matrix regression, and predict the restoration progress of the wasteland to be monitored based on the three-dimensional tensor matrix to obtain the prediction results; Step S4: Based on the comparison between the prediction results and the restoration requirements of the wasteland to be monitored, dynamically adjust the restoration strategy of the wasteland to be monitored.

2. The method for identifying wasteland for rural ecological restoration according to claim 1, characterized in that, The specific steps for determining the boundary range of the wasteland to be monitored in step S1 are as follows: Step S11: Determine the boundary of the wasteland to be monitored by combining on-site surveys with remote sensing images.

3. The method for identifying wasteland for rural ecological restoration according to claim 1, characterized in that, The specific steps for collecting electrical data of the wasteland to be monitored in step S1 are as follows: Step S12: In the wasteland to be monitored, select representative plants based on their growth status and distribution density. Step S13: Place the implanted microelectrode array in the soil around the rhizosphere of a representative plant and collect the changes in electrical signals in the rhizosphere microdomain as electrical data.

4. The method for identifying wasteland for rural ecological restoration according to claim 1, characterized in that, The specific steps for collecting odor data of the wasteland to be monitored in step S1 are as follows: Step S14: Grid the wasteland to be monitored, and select the grid on one of the diagonals of the wasteland to be monitored as the first collection point array; Step S15: Extend from the first collection point array to the other two corners of the wasteland to be monitored, and select a second collection point array composed of several grids at certain intervals, parallel to the first collection point array; Step S16: Collect odor samples from the headspace of the grid soil on the first and second collection point arrays as odor data.

5. The method for identifying wasteland for rural ecological restoration according to claim 1, characterized in that, The specific steps for collecting voiceprint data of the wasteland to be monitored in step S1 are as follows: Step S17: Distribute a distributed microphone array evenly in the wasteland to be monitored; Step S18: Collect voiceprint data through a distributed microphone array.

6. The method for identifying wasteland for rural ecological restoration according to claim 1, characterized in that, The specific steps for preprocessing the multidimensional data in step S2 are as follows: Step S21: Wavelet transform is used to process the electrical data to remove high-frequency noise and low-frequency interference. Moving average method is used to smooth the denoised electrical data, and linear interpolation method is used to fill in missing values. Step S22: Remove baseline drift in odor data, correct the baseline using polynomial fitting, extract characteristic peaks from odor data using partial least squares discriminant analysis, and standardize the characteristic peaks. Step S23: Use Fourier transform to convert the voiceprint data to the frequency domain, remove environmental noise, and then perform frame processing. Add a Hanning window to each frame of data, and use the endpoint detection method to detect the start and end points of the voiceprint signal, and remove invalid silent parts.

7. The method for identifying wasteland for rural ecological restoration according to claim 1, characterized in that, The specific steps in step S2 for feature extraction from the preprocessed multidimensional data and construction of a three-dimensional tensor matrix are as follows: Step S24: Extract time-domain features, frequency-domain features, and time-frequency-domain features from the preprocessed electrical data, and construct an electrical column matrix; Step S25: Perform cluster analysis on the preprocessed odor data, use the K-means clustering algorithm to cluster, and construct a soil odor fingerprint database. Extract the position and intensity of the characteristic peaks from the soil odor fingerprint database and construct an odor column matrix. Step S26: Process the voiceprint data using a Mel filter, extract the Mel cepstral coefficients, calculate the soundscape complexity index of the voiceprint data, extract the spectral centroid of the voiceprint data, and construct the voiceprint column matrix. Step S27: Construct a three-dimensional tensor matrix from the electrical column matrix, the odor column matrix, and the voiceprint column matrix.

8. The method for identifying wasteland for rural ecological restoration according to claim 1, characterized in that, The tensor decomposition and matrix regression fusion model includes a tensor decomposition part and a matrix regression part. The tensor decomposition part is used to extract latent features from a three-dimensional tensor matrix, and the matrix regression part is used to predict the restoration progress of the wasteland to be monitored based on the extracted latent features.

9. A method for identifying wasteland for rural ecological restoration according to claim 8, characterized in that, The specific steps of step S3 are as follows: Step S31: The tensor decomposition part uses Tucker decomposition to decompose the three-dimensional tensor matrix into core tensors, factor matrices, feature dimensions, and data type dimensions. Step S32: Extract shared features across data types from the factor matrix, obtain feature interaction information from the core tensor, and form a structured feature matrix; Step S33: The matrix regression part has a kernel matrix regression model, which takes the structured feature matrix as input and outputs a repair progress label. The repair progress is predicted based on the repair progress label.

10. The method for identifying wasteland for rural ecological restoration according to claim 1, characterized in that, The specific steps of step S4 are as follows: Step S41: Obtain the restoration requirements of the wasteland to be monitored, and determine the expected restoration result of the wasteland to be monitored at the current moment based on the restoration requirements; Step S42: If the prediction result is worse than the expected repair result, determine the repair period based on the prediction result; Step S43: Dynamically select one of the following restoration strategies based on the different restoration periods: soil improvement, vegetation planting, irrigation management, vegetation optimization, pest and disease control, soil monitoring and regulation, ecological maintenance, and biodiversity conservation.

Images