Forest land health state analysis method and system based on multi-source remote sensing image analysis

Abstract

The invention relates to the technical field of remote sensing, and discloses a forest land health state analysis method and system based on multi-source remote sensing image analysis. The system comprises a multi-source remote sensing acquisition module, a feature extraction and fusion module, a health assessment module, a traceability analysis module, a strategy matching module, a visual reconstruction module, an early warning decision module and an execution feedback module, and constructs a multi-modal forest land observation data set by fusing multi-source remote sensing data of multispectrum, hyperspectrum, radar and thermal infrared. The comprehensive extraction of multi-dimensional information such as vegetation coverage, canopy biochemical characteristics, under-forest structures and surface thermal environments is realized, the limitation of single data source analysis is overcome, and the comprehensiveness and accuracy of forest land health condition evaluation are improved; through dynamic comparison of multi-stage remote sensing images and health risk level mapping, early identification and early warning of forest growth abnormity, degeneration trend and pest and disease risk are realized.

Description

Technical Field

[0001] This invention relates to the field of remote sensing technology, specifically to a method and system for analyzing the health status of forest land based on multi-source remote sensing image analysis. Background Technology

[0002] Forestry remote sensing is an application technology that uses sensors to acquire spectral information based on the characteristics of objects reflecting or radiating electromagnetic waves, and analyzes and identifies forest resources and their environment. It is mainly used in areas such as monitoring tree growth, assessing resources, and providing early warnings of pests, diseases, and fires. In forestry work, the earliest and most widespread application of remote sensing technology is in forest resource surveys. After aerial photography was adopted for military purposes, it was quickly introduced into forestry surveying work.

[0003] Currently, forest health status analysis based on remote sensing technology mainly relies on remote sensing image data of a single type or limited sources for assessment. It is difficult to comprehensively utilize multi-source remote sensing information such as multispectral, thermal infrared, and radar, resulting in insufficient detection capabilities for vegetation physiological status, understory environment, water stress, and early pests and diseases.

[0004] Therefore, a method and system for analyzing forest health status based on multi-source remote sensing image analysis is proposed to solve the above problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for analyzing forest health status based on multi-source remote sensing image analysis. This solves the problem mentioned in the background technology that it is difficult to comprehensively utilize multi-source remote sensing information such as multispectral, thermal infrared, and radar, resulting in insufficient detection capabilities for vegetation physiological status, understory environment, water stress, and early pests and diseases.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method and system for analyzing the health status of forest land based on multi-source remote sensing image analysis, wherein the method includes the following steps:

[0007] S1. Acquire multimodal remote sensing image data of the target forest area through a multi-source remote sensing image acquisition platform to generate the original forest observation dataset;

[0008] S2. Perform multi-scale feature extraction and fusion processing on the original forest observation dataset to generate a forest state feature vector set;

[0009] S3. Based on the forest land state feature vector set, perform a quantitative assessment of the abnormal health status risk and generate health risk level assessment data;

[0010] S4. Based on the health risk level assessment data, perform dynamic grading of the forest area's health status to generate health status grading identifier data.

[0011] S5. When the health status classification identifier data is in an abnormal state, forest degradation and disaster source tracing analysis are triggered to generate multi-dimensional degradation source tracing and positioning data;

[0012] S6. Based on the multi-dimensional degradation source tracing and location data and the historical case database, the disposal strategy is matched to generate target ecological restoration strategy feature data;

[0013] S7. Construct multi-period monitoring summary data to visualize and reconstruct the health status of forest areas, and generate a three-dimensional dynamic health status map.

[0014] S8. Based on the three-dimensional dynamic health status map and the preset ecological security threshold, make forest land safety early warning decisions and generate a graded early warning instruction set;

[0015] S9. Execute the ecological regulation and restoration operations corresponding to the hierarchical early warning instruction set through the ecological monitoring IoT platform.

[0016] Preferably, the acquisition platform in S1 acquires multimodal remote sensing image data of the target forest area, including the following steps:

[0017] S11. Collect vegetation cover index data using a multispectral remote sensing sensor;

[0018] S12. Obtain biochemical parameter data of forest canopy using a hyperspectral imager;

[0019] S13. Collect surface micro-topography and understory structure data using synthetic aperture radar sensors;

[0020] S14. Obtain surface temperature distribution data in forest areas using thermal infrared sensors;

[0021] S15. Integrate the vegetation cover index data, canopy biochemical parameter data, forest understory structure data, and surface temperature data to construct the original forest observation dataset.

[0022] Preferably, the multi-scale feature extraction and fusion processing in S2 includes the following steps:

[0023] S21. Calculate vegetation index and extract spatial texture features from multispectral image data to generate vegetation state feature vectors.

[0024] S22. Perform dimensionality reduction of spectral features and extraction of absorption peak parameters on hyperspectral images to generate canopy biochemical feature vectors.

[0025] S23. Perform scattering mechanism analysis and topographic relief calculation on radar images to generate forest understory structure feature vectors;

[0026] S24. Detect temperature anomaly areas and extract thermal inertia features from thermal infrared images to generate surface thermal environment feature vectors.

[0027] S25. Integrate the vegetation state feature vector, canopy biochemical feature vector, forest understory structure feature vector, and surface thermal environment feature vector to construct a forest land state feature vector set.

[0028] Preferably, the quantitative assessment of abnormal health status risk in S3 includes the following steps:

[0029] S31. Construct a health risk assessment model based on a multilayer perceptron, wherein the input layer receives the set of forest land state feature vectors.

[0030] S32. Perform multi-source feature nonlinear relationship analysis through hidden layers to generate feature response data;

[0031] S33. Perform health risk probability mapping at the output layer to generate health risk level assessment data, wherein the risk levels include: healthy level, sub-healthy level, and diseased level.

[0032] Preferably, the dynamic grading of forest area health status in S4 includes the following steps:

[0033] S41. Receive the health risk level assessment data and establish a four-dimensional hierarchical coordinate system including vegetation cover dimension, canopy biochemistry dimension, understory structure dimension and thermal environment dimension;

[0034] S42. In the vegetation cover dimension, three levels of cover status are divided based on vegetation index characteristics: lush, general and sparse.

[0035] S43. In the canopy biochemical dimension, the biochemical state is divided into three levels based on characteristics such as chlorophyll content and water content: normal level, decline level, and abnormal level.

[0036] S44. In the understory structure dimension, the structure is divided into three levels based on scattering characteristics and topographic features: intact level, disturbed level, and damaged level.

[0037] S45. In the thermal environment dimension, the thermal environment is divided into three levels based on the abnormality of temperature distribution: equilibrium level, fluctuation level, and abnormal level.

[0038] S46. Establish a dynamic hierarchical decision-making strategy, generate health status hierarchical identification data, and map the hierarchical results to a three-dimensional dynamic health status map in real time through a feedback mechanism.

[0039] Preferably, the generation of multi-dimensional degradation source tracing and location data in step S5 includes the following steps:

[0040] S51. When the health status classification identifier data is in an abnormal state, activate the source tracing analysis engine to locate the spatial range and time node where the abnormality occurred.

[0041] S52. Identify the dominant degradation factors through multi-period image comparison and parameter coupling analysis;

[0042] S53. Integrate spatial range, time nodes, and dominant degradation factors to construct multi-dimensional degradation source tracing and positioning data.

[0043] Preferably, the matching of the handling strategy in S6 includes the following steps:

[0044] S61. Establish a knowledge graph of historical ecological restoration cases, whose nodes include degradation type, location of occurrence, restoration measures and effect evaluation;

[0045] S62. Perform structural similarity matching between multi-dimensional degradation source tracing and localization data and historical case knowledge graph;

[0046] S63. When the matching similarity reaches the preset threshold, the corresponding ecological restoration strategy is output as the target strategy feature data.

[0047] S64. When the matching similarity is lower than the preset threshold, the adaptive strategy generation mechanism is activated to generate a new recovery strategy.

[0048] Preferably, the visual reconstruction of the forest area health status in S7 includes the following steps:

[0049] S71. By integrating multi-period vegetation index distribution data, canopy parameter data, surface temperature data, and topographic data, a three-dimensional dynamic health status model of the forest area is constructed.

[0050] S72. Use pseudo-color rendering technology to map health level data onto the model surface;

[0051] S73. Characterize the growth and degradation trends of forest trees through dynamic change simulation technology;

[0052] S74. Integrate and generate a three-dimensional dynamic health status map that includes the distribution of health status, degraded areas, and changing trends.

[0053] Preferably, the forest land safety early warning decision in S8 includes the following steps:

[0054] S81. Set up a multi-level early warning triggering mechanism, including blue observation level, yellow warning level, and red emergency level;

[0055] S82. When the coverage rate of abnormal areas in the health status map exceeds the set threshold, a corresponding early warning instruction is triggered.

[0056] S83. When a key biochemical parameter deviates from the baseline value for multiple consecutive periods, an early warning instruction is triggered.

[0057] S84. Generate a hierarchical early warning instruction set that includes the warning level, spatial range, and handling recommendations.

[0058] Preferably, the system includes:

[0059] The multi-source remote sensing acquisition module uses satellite remote sensing units, airborne remote sensing units, and ground-based IoT sensing units to acquire multimodal forest image data, and generates standardized observation datasets through the data preprocessing unit;

[0060] The feature extraction and fusion module receives the observation dataset, extracts multi-dimensional features through the vegetation analysis unit, canopy inversion unit, terrain construction unit and thermal environment analysis unit, and generates a forest land state feature vector set through the feature fusion unit;

[0061] The health assessment module receives the feature vector set, generates health level data through the risk assessment unit, and outputs a health status classification identifier with the help of the status classification unit.

[0062] The source tracing analysis module is activated when the health status is abnormal, and outputs degradation source tracing location data through the spatiotemporal positioning unit and the factor analysis unit;

[0063] The strategy matching module receives the source tracing and positioning data, and extracts and generates recovery strategies from the historical case database through the case retrieval unit and the similarity matching unit.

[0064] The visualization reconstruction module integrates multi-source data and evaluation results, and generates a visual map of health status through 3D modeling units and dynamic simulation units.

[0065] The early warning decision module receives the visualization map and outputs graded early warning instructions through the threshold judgment unit and the early warning generation unit;

[0066] The execution feedback module executes early warning commands through the ecological IoT control platform and continuously monitors and provides feedback on the effectiveness of the measures.

[0067] Beneficial effects

[0068] Compared with existing technologies, this invention provides a method and system for analyzing forest health status based on multi-source remote sensing image analysis, which has the following beneficial effects:

[0069] 1. In this invention, by integrating multi-source remote sensing data such as multispectral, hyperspectral, radar and thermal infrared, a multimodal forest observation dataset is constructed, which enables the comprehensive extraction of multi-dimensional information such as vegetation cover, canopy biochemical characteristics, understory structure and surface thermal environment. This overcomes the limitations of single data source analysis, improves the comprehensiveness and accuracy of forest health assessment, and provides more reliable decision support for forestry management.

[0070] 2. In this invention, by dynamically comparing and mapping health risk levels of multiple remote sensing images, early identification and warning of abnormal forest growth, degradation trends and pest and disease risks are achieved. Furthermore, the three-dimensional dynamic health status map is used for visualization, which improves the response speed and monitoring timeliness of forest land status changes and enhances the proactive management capability of forest resource protection.

[0071] 3. In this invention, through multi-scale feature extraction, multi-level grading and historical case matching mechanisms, fine zoning, hierarchical health assessment and degradation cause tracing of forest land in different regions are realized, reducing subjective errors of manual interpretation, improving the automation and repeatability of forest land health status analysis under large-scale and complex terrain conditions, and providing a scientific basis for the formulation of ecological restoration strategies. Attached Figure Description

[0072] Figure 1 This is a flowchart of the forest health status analysis method based on multi-source remote sensing image analysis according to the present invention;

[0073] Figure 2 This is a schematic diagram of the forest health status analysis system based on multi-source remote sensing image analysis of the present invention. Detailed Implementation

[0074] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0075] Specific embodiment: A method and system for analyzing forest health status based on multi-source remote sensing image analysis, the method including the following steps:

[0076] S1. Acquire multimodal remote sensing image data of the target forest area through a multi-source remote sensing image acquisition platform to generate the original forest observation dataset;

[0077] S2. Perform multi-scale feature extraction and fusion processing on the original forest observation dataset to generate a forest state feature vector set;

[0078] S3. Quantitatively assess the risk of abnormal health status based on the forest land status feature vector set, and generate health risk level assessment data;

[0079] S4. Based on the health risk level assessment data, perform dynamic grading of the health status of the forest area and generate health status grading label data.

[0080] S5. When the health status classification identifier data is in an abnormal state, forest degradation and disaster source tracing analysis are triggered to generate multi-dimensional degradation source tracing and positioning data.

[0081] S6. Based on multi-dimensional degradation source tracing and location data and historical case database, match treatment strategies to generate target ecological restoration strategy feature data;

[0082] S7. Construct multi-period monitoring summary data to visualize and reconstruct the health status of forest areas, and generate a three-dimensional dynamic health status map.

[0083] S8. Based on the three-dimensional dynamic health status map and the preset ecological security threshold, make forest land safety early warning decisions and generate a graded early warning instruction set;

[0084] S9. Execute ecological regulation and restoration operations corresponding to the hierarchical early warning instruction set through the ecological monitoring IoT platform.

[0085] The acquisition platform in S1 obtains multimodal remote sensing image data of the target forest area, including the following steps:

[0086] S11. Collect vegetation cover index data using a multispectral remote sensing sensor;

[0087] S12. Obtain biochemical parameter data of forest canopy using a hyperspectral imager;

[0088] S13. Collect surface micro-topography and understory structure data using synthetic aperture radar sensors;

[0089] S14. Obtain surface temperature distribution data in forest areas using thermal infrared sensors;

[0090] S15. Integrate vegetation cover index data, canopy biochemical parameter data, forest understory structure data, and surface temperature data to construct a primitive forest observation dataset.

[0091] Multi-scale feature extraction and fusion processing in S2 includes the following steps:

[0092] S21. Calculate vegetation index and extract spatial texture features from multispectral image data to generate vegetation state feature vectors.

[0093] Vegetation index calculation:

[0094] ;

[0095] in, The vegetation index, Indicates near-infrared reflectivity. Indicates the reflectivity in the red light band;

[0096] S22. Perform dimensionality reduction of spectral features and extraction of absorption peak parameters on hyperspectral images to generate canopy biochemical feature vectors.

[0097] S23. Perform scattering mechanism analysis and topographic relief calculation on radar images to generate forest understory structure feature vectors;

[0098] The scattering mechanism analysis uses the Cloude-Pottier decomposition method to process the fully polarimetric SAR data, calculates the coherence matrix, and then obtains three parameters—entropy, anisotropy, and average scattering angle—through eigenvalue decomposition.

[0099] Calculation of terrain relief:

[0100] ;

[0101] in, Indicates the degree of terrain relief. This represents the elevation value of the i-th pixel. n represents the average elevation within the calculation window. This represents the total number of pixels within the window.

[0102] S24. Detect temperature anomaly areas and extract thermal inertia features from thermal infrared images to generate surface thermal environment feature vectors.

[0103] S25. By integrating vegetation state feature vectors, canopy biochemical feature vectors, forest understory structure feature vectors, and surface thermal environment feature vectors, a forest land state feature vector set is constructed.

[0104] The quantitative assessment of abnormal health status risk in S3 includes the following steps:

[0105] S31. Construct a health risk assessment model based on a multilayer perceptron, whose input layer receives a set of forest land state feature vectors;

[0106] The health risk assessment model employs a multilayer perceptron structure based on an attention mechanism. The input layer receives a feature vector consisting of 16 feature indicators, including 4 vegetation index features, 4 spatial texture features, 3 scattering mechanism features, 3 terrain features, and 2 thermal environment features. The first hidden layer contains 32 neurons and uses the ReLU activation function. The second hidden layer contains 16 neurons and also uses the ReLU activation function. The output layer contains 3 neurons, corresponding to the probability outputs of healthy, sub-healthy, and diseased levels, respectively, and uses the Softmax activation function. The model is trained using the Adam optimizer with a learning rate of 0.001, a batch size of 64, and 100 training epochs.

[0107] S32. Perform multi-source feature nonlinear relationship analysis through hidden layers to generate feature response data;

[0108] S33. Perform health risk probability mapping in the output layer to generate health risk level assessment data. The risk levels include: healthy level, sub-healthy level, and diseased level.

[0109] The health risk probability mapping converts the model output into specific probability values ​​using the following formula:

[0110] ;

[0111] in, This represents the probability of the k-th risk level, where k=1,2,3 correspond to the healthy level, sub-healthy level, and diseased level, respectively. This represents the original output value of the k-th neuron in the output layer, where i is the summation index. In practice, the probability values ​​of three risk levels are calculated for each pixel, and the level corresponding to the highest probability is taken as the final risk level of that pixel.

[0112] The dynamic grading of forest health status in S4 forest areas includes the following steps:

[0113] S41. Receive health risk level assessment data and establish a four-dimensional hierarchical coordinate system that includes vegetation cover dimension, canopy biochemistry dimension, understory structure dimension and thermal environment dimension;

[0114] S42. In the vegetation cover dimension, three levels of cover status are divided based on vegetation index characteristics: lush, general and sparse.

[0115] S43. In the canopy biochemical dimension, the biochemical state is divided into three levels based on characteristics such as chlorophyll content and water content: normal level, decline level, and abnormal level.

[0116] S44. In the understory structure dimension, the structure is divided into three levels based on scattering characteristics and topographic features: intact level, disturbed level, and damaged level.

[0117] S45. In the thermal environment dimension, the thermal environment is divided into three levels based on the abnormality of temperature distribution: equilibrium level, fluctuation level, and abnormal level.

[0118] S46. Establish a dynamic hierarchical decision-making strategy, generate health status hierarchical identification data, and map the hierarchical results to a three-dimensional dynamic health status map in real time through feedback.

[0119] The dynamic hierarchical decision-making strategy employs a multi-level threshold determination method:

[0120] First, the grading thresholds for each dimension are set: In the vegetation cover dimension, a vegetation index > 0.6 is considered lush, 0.3 < vegetation index ≤ 0.6 is considered average, and vegetation index ≤ 0.3 is considered sparse; In the canopy biochemistry dimension, chlorophyll content > 40 μg / cm² is considered normal, 20-40 μg / cm² is considered declining, and < 20 μg / cm² is considered abnormal; In the understory structure dimension, scattering entropy < 0.6 is considered intact, 0.6-0.8 is considered disturbed, and > 0.8 is considered damaged; In the thermal environment dimension, temperature variation coefficient < 0.1 is considered balanced, 0.1-0.3 is considered fluctuating, and > 0.3 is considered abnormal; When making decisions, a weighted summation method is used to calculate the comprehensive health index:

[0121] ;

[0122] in, For health index, , , , Weights for each dimension and , , , , The scores are for each dimension; HI>80 indicates a healthy state, 60-80 indicates a sub-healthy state, and <60 indicates a diseased state.

[0123] Generating multi-dimensional degradation source location data in S5 includes the following steps:

[0124] S51. When the health status classification identifier data is in an abnormal state, activate the source tracing analysis engine to locate the spatial range and time node where the abnormality occurred.

[0125] S52. Identify the dominant degradation factors through multi-period image comparison and parameter coupling analysis;

[0126] S53. Integrate spatial range, time nodes, and dominant degradation factors to construct multi-dimensional degradation source tracing and positioning data.

[0127] S6's handling strategy matching includes the following steps:

[0128] S61. Establish a knowledge graph of historical ecological restoration cases, whose nodes include degradation type, location of occurrence, restoration measures and effect evaluation;

[0129] The knowledge graph of historical ecological restoration cases is constructed using the Neo4j graph database. The node types include: degradation type nodes, geographical location nodes, restoration measure nodes, and effect evaluation nodes. The edge relationships include: occurred in, adopted, and generated. Each node contains attribute fields: degradation type nodes contain severity indicators, geographical location nodes contain altitude, slope, and soil type information, restoration measure nodes contain implementation time and cost information, and effect evaluation nodes contain effect scores and monitoring periods.

[0130] S62. Perform structural similarity matching between multi-dimensional degradation source tracing and localization data and historical case knowledge graph;

[0131] Structural similarity matching employs a graph isomorphism algorithm, calculating the similarity between two knowledge graphs using the following formula:

[0132] ;

[0133] in, Indicates the similarity between two graphs. , These represent the node sets of the two graphs, , Let each represent the set of edges in the two graphs;

[0134] S63. When the matching similarity reaches the preset threshold, the corresponding ecological restoration strategy is output as the target strategy feature data.

[0135] The ecological restoration strategy is generated using a case-based reasoning method. When similar historical cases are found, the system extracts the restoration measures and their effect evaluations from these cases, and selects the three measures with the highest applicability scores to form a restoration strategy recommendation.

[0136] S64. When the matching similarity is lower than the preset threshold, the adaptive strategy generation mechanism is activated to generate a new recovery strategy.

[0137] When the matching similarity is lower than a preset threshold, an adaptive policy generation mechanism based on reinforcement learning is activated. The system uses the forest environment status as the state space, including 10 parameters such as soil moisture, vegetation cover, and slope aspect; it uses possible restoration measures as the action space, containing 12 standard restoration operations; and it uses the degree of improvement in the health index as the reward function.

[0138] ;

[0139] in, Indicates the reward value. Indicates the health index after the implementation of the measures. The health index before the implementation of the measures is represented; the Q-learning algorithm is used for policy optimization, with a learning rate α=0.1 and a discount factor γ=0.9. The optimal policy is obtained through 500 iterations of training.

[0140] Visual reconstruction of the health status of forest areas in S7 includes the following steps:

[0141] S71. By integrating multi-period vegetation index distribution data, canopy parameter data, surface temperature data, and topographic data, a three-dimensional dynamic health status model of the forest area is constructed.

[0142] The three-dimensional dynamic health status model of the forest area was built using the Unity3D engine. First, a digital elevation model was imported as the terrain base. Then, the two-dimensional health data was mapped to three-dimensional space using the following formula:

[0143] ;

[0144] in, Represents the three-dimensional health status value. , , , These represent two-dimensional data for vegetation, biochemistry, structure, and thermal environment, respectively. , , , The weighting coefficient is used. During implementation, elevation offset technology is used to visualize the health status data. Healthy areas are displayed in green and maintain their original elevation, while diseased areas are displayed in red and show elevation subsidence. The subsidence depth is proportional to the severity of the disease. This simulates changes in forest growth. Particle density is proportional to vegetation coverage, and particle size is proportional to the forest growth status.

[0145] S72. Use pseudo-color rendering technology to map health level data onto the model surface;

[0146] S73. Characterize the growth and degradation trends of forest trees through dynamic change simulation technology;

[0147] S74. Integrate and generate a three-dimensional dynamic health status map that includes the distribution of health status, degraded areas, and changing trends.

[0148] S8 forest land safety early warning decision-making includes the following steps:

[0149] S81. Set up a multi-level early warning triggering mechanism, including blue observation level, yellow warning level, and red emergency level;

[0150] S82. When the coverage rate of abnormal areas in the health status map exceeds the set threshold, a corresponding early warning instruction is triggered.

[0151] S83. When a key biochemical parameter deviates from the baseline value for multiple consecutive periods, an early warning instruction is triggered.

[0152] S84. Generate a hierarchical early warning instruction set that includes the warning level, spatial range, and handling recommendations.

[0153] The system includes:

[0154] The multi-source remote sensing acquisition module uses satellite remote sensing units, airborne remote sensing units, and ground-based IoT sensing units to acquire multimodal forest image data, and generates standardized observation datasets through the data preprocessing unit;

[0155] The feature extraction and fusion module receives the observation dataset, extracts multi-dimensional features through the vegetation analysis unit, canopy inversion unit, terrain construction unit and thermal environment analysis unit, and generates a forest land state feature vector set through the feature fusion unit;

[0156] The health assessment module receives a set of feature vectors, generates health level data through the risk assessment unit, and outputs a health status classification identifier with the help of the status classification unit.

[0157] The source tracing analysis module is activated when the health status is abnormal, and outputs degradation source tracing location data through the spatiotemporal positioning unit and the factor analysis unit;

[0158] The strategy matching module receives source tracing and location data, and extracts and generates recovery strategies from the historical case database through the case retrieval unit and similarity matching unit.

[0159] The visualization reconstruction module integrates multi-source data and evaluation results, and generates a visual map of health status through 3D modeling units and dynamic simulation units.

[0160] The early warning decision module receives the visual map and outputs graded early warning instructions through the threshold judgment unit and the early warning generation unit;

[0161] The execution feedback module executes early warning commands through the ecological IoT control platform and continuously monitors and provides feedback on the effectiveness of the measures.

[0162] The operation steps of this method and system are as follows:

[0163] This method first acquires multimodal remote sensing image data of the target forest area through a multi-source remote sensing image acquisition platform, integrating multispectral, hyperspectral, radar, and thermal infrared information to generate a raw forest observation dataset. Subsequently, multi-scale feature extraction and fusion processing is performed on the raw dataset to extract vegetation status, canopy biochemical characteristics, understory structure, and surface thermal environment features, which are then fused to generate a comprehensive forest status feature vector set. Based on this feature vector set, anomaly risk quantification assessment is performed using a health risk assessment model, generating risk assessment data including levels of health, sub-health, and disease.

[0164] Based on the assessment results, the system establishes a multi-dimensional hierarchical coordinate system for dynamic grading of forest health status. It classifies status levels from four dimensions: vegetation cover, canopy biochemistry, understory structure, and thermal environment, generating health status grading identifier data. When an abnormal status is identified, a source tracing analysis mechanism is triggered. Through multi-period image comparison and parameter coupling analysis, the spatial range and time points of the anomaly are located, the dominant degradation factors are identified, and multi-dimensional degradation source tracing and location data are generated.

[0165] The system matches the source data with a historical case database, obtains and generates corresponding ecological restoration strategies through structural similarity comparison, and integrates monitoring data from multiple periods to construct a three-dimensional dynamic health status model. It then uses visualization technology to generate dynamic maps reflecting health distribution, degraded areas, and changing trends. Finally, based on the map data and preset ecological security thresholds, the system makes multi-level safety early warning decisions, generates a tiered early warning instruction set, and executes corresponding ecological regulation and restoration operations through the ecological IoT platform, forming a complete closed-loop management process from monitoring, assessment, early warning to execution.

[0166] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0167] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing forest health status based on multi-source remote sensing image analysis, characterized in that: The method includes the following steps: S1. Acquire multimodal remote sensing image data of the target forest area through a multi-source remote sensing image acquisition platform to generate the original forest observation dataset; S2. Perform multi-scale feature extraction and fusion processing on the original forest observation dataset to generate a forest state feature vector set; S3. Based on the forest land state feature vector set, perform a quantitative assessment of the abnormal health status risk and generate health risk level assessment data; S4. Based on the health risk level assessment data, perform dynamic grading of the forest area's health status to generate health status grading identifier data. S5. When the health status classification identifier data is in an abnormal state, forest degradation and disaster source tracing analysis are triggered to generate multi-dimensional degradation source tracing and positioning data; S6. Based on the multi-dimensional degradation source tracing and location data and the historical case database, the disposal strategy is matched to generate target ecological restoration strategy feature data; S7. Construct multi-period monitoring summary data to visualize and reconstruct the health status of forest areas, and generate a three-dimensional dynamic health status map. S8. Based on the three-dimensional dynamic health status map and the preset ecological security threshold, make forest land safety early warning decisions and generate a graded early warning instruction set; S9. Execute the ecological regulation and restoration operations corresponding to the hierarchical early warning instruction set through the ecological monitoring IoT platform.

2. The forest land health status analysis method based on multi-source remote sensing image analysis according to claim 1, characterized in that: The acquisition platform in S1 obtains multimodal remote sensing image data of the target forest area, including the following steps: S11. Collect vegetation cover index data using a multispectral remote sensing sensor; S12. Obtain biochemical parameter data of forest canopy using a hyperspectral imager; S13. Collect surface micro-topography and understory structure data using synthetic aperture radar sensors; S14. Obtain surface temperature distribution data in forest areas using thermal infrared sensors; S15. Integrate the vegetation cover index data, canopy biochemical parameter data, forest understory structure data, and surface temperature data to construct the original forest observation dataset.

3. The forest land health status analysis method based on multi-source remote sensing image analysis according to claim 1, characterized in that: The multi-scale feature extraction and fusion process in S2 includes the following steps: S21. Calculate vegetation index and extract spatial texture features from multispectral image data to generate vegetation state feature vectors. S22. Perform dimensionality reduction of spectral features and extraction of absorption peak parameters on hyperspectral images to generate canopy biochemical feature vectors. S23. Perform scattering mechanism analysis and topographic relief calculation on radar images to generate forest understory structure feature vectors; S24. Detect temperature anomaly areas and extract thermal inertia features from thermal infrared images to generate surface thermal environment feature vectors. S25. Integrate the vegetation state feature vector, canopy biochemical feature vector, forest understory structure feature vector, and surface thermal environment feature vector to construct a forest land state feature vector set.

4. The forest land health status analysis method based on multi-source remote sensing image analysis according to claim 1, characterized in that: The quantitative assessment of abnormal health status risk in S3 includes the following steps: S31. Construct a health risk assessment model based on a multilayer perceptron, wherein the input layer receives the set of forest land state feature vectors. S32. Perform multi-source feature nonlinear relationship analysis through hidden layers to generate feature response data; S33. Perform health risk probability mapping at the output layer to generate health risk level assessment data, wherein the risk levels include: healthy level, sub-healthy level, and diseased level.

5. The forest land health status analysis method based on multi-source remote sensing image analysis according to claim 1, characterized in that: The dynamic grading of forest area health status in S4 includes the following steps: S41. Receive the health risk level assessment data and establish a four-dimensional hierarchical coordinate system including vegetation cover dimension, canopy biochemistry dimension, understory structure dimension and thermal environment dimension; S42. In the vegetation cover dimension, three levels of cover status are divided based on vegetation index characteristics: lush, general and sparse. S43. In the canopy biochemical dimension, the biochemical state is divided into three levels based on characteristics such as chlorophyll content and water content: normal level, decline level, and abnormal level. S44. In the understory structure dimension, the structure is divided into three levels based on scattering characteristics and topographic features: intact level, disturbed level, and damaged level. S45. In the thermal environment dimension, the thermal environment is divided into three levels based on the abnormality of temperature distribution: equilibrium level, fluctuation level, and abnormal level. S46. Establish a dynamic hierarchical decision-making strategy, generate health status hierarchical identification data, and map the hierarchical results to a three-dimensional dynamic health status map in real time through a feedback mechanism.

6. The method for analyzing forest health status based on multi-source remote sensing image analysis according to claim 1, characterized in that: The generation of multi-dimensional degradation source tracing and location data in S5 includes the following steps: S51. When the health status classification identifier data is in an abnormal state, activate the source tracing analysis engine to locate the spatial range and time node where the abnormality occurred. S52. Identify the dominant degradation factors through multi-period image comparison and parameter coupling analysis; S53. Integrate spatial range, time nodes, and dominant degradation factors to construct multi-dimensional degradation source tracing and positioning data.

7. The method for analyzing forest health status based on multi-source remote sensing image analysis according to claim 1, characterized in that: The matching of the handling strategy in S6 includes the following steps: S61. Establish a knowledge graph of historical ecological restoration cases, whose nodes include degradation type, location of occurrence, restoration measures and effect evaluation; S62. Perform structural similarity matching between multi-dimensional degradation source tracing and localization data and historical case knowledge graph; S63. When the matching similarity reaches the preset threshold, the corresponding ecological restoration strategy is output as the target strategy feature data. S64. When the matching similarity is lower than the preset threshold, the adaptive strategy generation mechanism is activated to generate a new recovery strategy.

8. The method for analyzing forest health status based on multi-source remote sensing image analysis according to claim 1, characterized in that: The visualization reconstruction of forest health status in S7 includes the following steps: S71. By integrating multi-period vegetation index distribution data, canopy parameter data, surface temperature data, and topographic data, a three-dimensional dynamic health status model of the forest area is constructed. S72. Use pseudo-color rendering technology to map health level data onto the model surface; S73. Characterize the growth and degradation trends of forest trees through dynamic change simulation technology; S74. Integrate and generate a three-dimensional dynamic health status map that includes the distribution of health status, degraded areas, and changing trends.

9. The method for analyzing forest health status based on multi-source remote sensing image analysis according to claim 1, characterized in that: The forest safety early warning decision-making in S8 includes the following steps: S81. Set up a multi-level early warning triggering mechanism, including blue observation level, yellow warning level, and red emergency level; S82. When the coverage rate of abnormal areas in the health status map exceeds the set threshold, a corresponding early warning instruction is triggered. S83. When a key biochemical parameter deviates from the baseline value for multiple consecutive periods, an early warning instruction is triggered. S84. Generate a hierarchical early warning instruction set that includes the warning level, spatial range, and handling recommendations.

10. A forest land health status analysis system based on multi-source remote sensing image analysis, used to implement the forest land health status analysis method based on multi-source remote sensing image analysis as described in any one of claims 1-9, characterized in that: The system includes: The multi-source remote sensing acquisition module uses satellite remote sensing units, airborne remote sensing units, and ground-based IoT sensing units to acquire multimodal forest image data, and generates standardized observation datasets through the data preprocessing unit; The feature extraction and fusion module receives the observation dataset, extracts multi-dimensional features through the vegetation analysis unit, canopy inversion unit, terrain construction unit and thermal environment analysis unit, and generates a forest land state feature vector set through the feature fusion unit; The health assessment module receives the feature vector set, generates health level data through the risk assessment unit, and outputs a health status classification identifier with the help of the status classification unit. The source tracing analysis module is activated when the health status is abnormal, and outputs degradation source tracing location data through the spatiotemporal positioning unit and the factor analysis unit; The strategy matching module receives the source tracing and positioning data, and extracts and generates recovery strategies from the historical case database through the case retrieval unit and the similarity matching unit. The visualization reconstruction module integrates multi-source data and evaluation results, and generates a visual map of health status through 3D modeling units and dynamic simulation units. The early warning decision module receives the visualization map and outputs graded early warning instructions through the threshold judgment unit and the early warning generation unit; The execution feedback module executes early warning commands through the ecological IoT control platform and continuously monitors and provides feedback on the effectiveness of the measures.

Images