A forestry intelligent surveying and mapping method and system based on regional feature feedback

Abstract

The application discloses a forestry intelligent surveying and mapping method and system based on regional feature feedback, relates to the technical field of forest area surveying and mapping, and provides the following scheme, which comprises the following steps: extracting spectral and texture features of a target forest area image through satellite remote sensing data, realizing identification and hierarchical division of feature regions by using an intelligent segmentation algorithm, and generating a priority monitoring grid containing quantitative evaluation indexes; scheduling a UAV to perform first scanning on the priority monitoring grid, and integrating image data of the UAV and the satellite; extracting spectral and texture features of a target forest area through satellite remote sensing data, generating a non-uniform priority monitoring grid containing a feature density index, scheduling a UAV to perform first scanning on a high-priority grid, fusing satellite multispectral data, UAV high-resolution images and LiDAR point cloud data, identifying tree species, and generating a health index heat map, so that a high-risk area can be quickly identified, the scanning range of the UAV can be significantly compressed, and the monitoring efficiency can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of forest area surveying and mapping, in particular to a forestry intelligent surveying and mapping method and system based on regional feature feedback. BACKGROUND

[0002] The current forestry monitoring mainly adopts a satellite remote sensing and unmanned aerial vehicle cooperative operation mode. The satellite provides large-range forest coverage data through multi-spectral images, combines with supervised classification, such as random forest to divide vegetation types, and the unmanned aerial vehicle realizes analysis on tree species through high-resolution images or LiDAR data, and detects disease and pest areas through the NDVI threshold method.

[0003] The existing fixed division strategy of the unmanned aerial vehicle scanning area based on satellite images needs to forcibly complete full forest coverage scanning to generate a health heat map, which causes the unmanned aerial vehicle to perform a large amount of invalid scanning in the healthy area, resulting in significant resource redundancy and energy consumption. In addition, since the feature feedback mechanism is not established in the first scanning, the identification of high-risk areas must wait for the completion of full forest data collection, which significantly prolongs the response cycle from data collection to risk warning. Therefore, the present application provides a forestry intelligent surveying and mapping method and system based on regional feature feedback to solve the above problems. SUMMARY

[0004] To solve the above technical problems, the present application provides a forestry intelligent surveying and mapping method and system based on regional feature feedback, which solves the problems of the existing fixed division strategy of the unmanned aerial vehicle scanning area based on satellite images, which needs to forcibly complete full forest coverage scanning to generate a health heat map, which causes the unmanned aerial vehicle to perform a large amount of invalid scanning in the healthy area, resulting in significant resource redundancy and energy consumption. In addition, since the feature feedback mechanism is not established in the first scanning, the identification of high-risk areas must wait for the completion of full forest data collection, which significantly prolongs the response cycle from data collection to risk warning.

[0005] To achieve the above purposes, the technical scheme adopted by the present application is as follows: the present application provides a forestry intelligent surveying and mapping method based on regional feature feedback, which comprises the following steps:

[0006] Step 1: Extract the spectral and texture features of the target forest area image through satellite remote sensing data, use an intelligent segmentation algorithm to realize the identification and hierarchical division of the feature area, and generate a priority monitoring grid containing quantitative evaluation indexes;

[0007] Step 2: Schedule the unmanned aerial vehicle to perform the first scanning on the priority monitoring grid, integrate the image data of the unmanned aerial vehicle and the satellite, fuse the spectral features and three-dimensional structure parameters to construct a tree species identification model, and establish a category system with tree species nouns as labels;

[0008] Step 3: Based on the threshold library of image features of healthy tree species, the deviation degree of each region from the healthy state is quantified, and the health index thermal map of the target forest area is generated through spatial statistical analysis;

[0009] Step 4: Establish a dynamic mapping relationship between the health index and the risk level, dynamically update the category system based on the scanning results of the unmanned aerial vehicle, and plan the secondary scanning of the target forest area by the unmanned aerial vehicle according to the health risk level.

[0010] Further, the spectral and texture features of the target forest area image are extracted from the satellite remote sensing data, the feature region is identified and classified by using an intelligent segmentation algorithm, and a priority monitoring grid containing quantitative evaluation indicators is generated, specifically including:

[0011] Obtain the multispectral satellite remote sensing image of the target forest area, extract the spectral reflectance after preprocessing by radiation calibration and atmospheric correction, and calculate the texture feature parameters based on the gray level co-occurrence matrix to generate the preliminary feature region;

[0012] The normalized spectral reflectance and texture feature parameter values are spliced into a feature vector, the preliminary feature region is clustered based on the vector, and a first-level cluster label is assigned;

[0013] The satellite remote sensing image and the cluster label are input into the semantic segmentation model, and after optimization by the cross-entropy loss function, the probability distribution map of each pixel belonging to different feature categories is output, and the non-uniform grid is divided based on the probability distribution density using an adaptive method, denoted as the monitoring grid;

[0014] The ratio of the sum of the pixel category probabilities in each grid to the grid area is calculated as the feature density index, and the priority area is divided according to the preset density threshold.

[0015] Further, the unmanned aerial vehicle is dispatched to perform the first scan on the priority monitoring grid, the image data of the unmanned aerial vehicle and the satellite are integrated, the tree species recognition model is constructed by fusing the spectral features and three-dimensional structure parameters, and a category system with tree species nouns as labels is established, specifically including:

[0016] Collect the hyperspectral imaging and LiDAR point cloud data of the area to be scanned, calculate the Euclidean distance of each pixel from the standard tree species spectral library as the spectral fingerprint difference, and generate DEM and DSM based on the point cloud, extract the leaf layer density gradient and crown concave-convex degree as vertical structure parameters;

[0017] Fuse the spectral fingerprint difference and the vertical structure parameters through a random forest model to dynamically distinguish tree species, and calculate the proportion of pixels of each tree species in the grid as its distribution area;

[0018] The unmanned aerial vehicle spectrum fingerprint is combined with satellite remote sensing images for mixed pixel decomposition, the proportion of each tree species in the satellite remote sensing image pixel is calculated, the tree species composition of the preliminary characteristic region is analyzed, and the first-level clustering label is refined into a second-level clustering label according to the tree species name;

[0019] The tree species in the non-preliminary characteristic region are clustered, and when the cumulative distribution area of a certain type of tree species reaches the clustering threshold, a second-level clustering label is assigned.

[0020] Further, the health state deviation degree of each region is quantified based on the health tree species image feature threshold library, and a health index heat map of the target forest area is generated through spatial statistical analysis, specifically including:

[0021] For each tree species in the second-level clustering label, the spectral reflectance threshold and the crown concave-convex degree threshold under the health state are obtained, and each threshold is adjusted according to the season;

[0022] The spectral reflectance difference and the crown concave-convex degree difference between the target tree species and the health threshold are calculated pixel by pixel, and the 3 Principle is used to remove abnormal difference data;

[0023] The distribution area of the tree species is taken as a weight coefficient to generate a health coefficient of the tree species by weighted summation;

[0024] The data missing area is compensated by the Kriging spatial interpolation method, the health coefficients of each tree species are summarized and weightedly averaged, and the health index heat map of the target forest area is output.

[0025] Further, the dynamic mapping relationship between the health index and the risk level is established, and the category system is dynamically updated based on the scanning results of the unmanned aerial vehicle, specifically including:

[0026] The health coefficient threshold interval is set, the second-level clustering label with a total health coefficient exceeding the upper limit of the threshold interval is marked as a dangerous label, the label within the interval is marked as a potential dangerous label, and the label below the lower limit of the threshold interval is marked as a low-dangerous label;

[0027] When the unmanned aerial vehicle scans a new to-be-scanned area, the total health coefficient of the second-level clustering label is recalculated;

[0028] Through the streaming computing engine, the total health coefficient is compared with the health coefficient threshold, and the second-level clustering label and its label mark are updated.

[0029] Further, the planning of the second scanning of the target forest area by the unmanned aerial vehicle according to the health risk level specifically includes:

[0030] Based on the health index heat map and the health coefficient threshold, the target forest area is divided into a dangerous area, a potential dangerous area, and a low-dangerous area;

[0031] If the area proportion of the dangerous area in the feature region does not reach the set proportion of the dangerous area coverage rate, the dangerous area not covered is scanned again by the unmanned aerial vehicle, a certain proportion of overlapping scanning bands are set, the non-dangerous area is sampled and scanned, and hierarchical sensor configuration is adopted according to the regional danger level;

[0032] The dangerous area is tilted and photographed to generate a 3D disease canopy model, and the airborne artificial intelligence algorithm is combined to identify the disease and pests in real time and adjust the flight parameters.

[0033] Further, a forestry intelligent surveying and mapping system based on regional feature feedback is proposed, which is used to implement the forestry intelligent surveying and mapping method based on regional feature feedback according to any one of the above, comprising:

[0034] A multi-source remote sensing feature extraction and hierarchical module is used to extract the spectral and texture features of the target forest area image through satellite remote sensing data, realize the identification and hierarchical division of the feature region by using an intelligent segmentation algorithm, and generate a priority monitoring grid containing a quantitative evaluation index;

[0035] An air-ground collaborative tree species identification module is used to schedule the unmanned aerial vehicle to perform the first scanning on the priority monitoring grid, integrate the image data of the unmanned aerial vehicle and the satellite, fuse the spectral features and three-dimensional structure parameters to construct a tree species identification model, and establish a category system with tree species nouns as labels;

[0036] A health dynamic evaluation module is used to quantify the deviation degree of the health state of each region based on a health tree species image feature threshold library, and generate a health index heat map of the target forest area through spatial statistical analysis;

[0037] An adaptive response control module is used to establish a dynamic mapping relationship between the health index and the risk level, dynamically update the category system based on the scanning result of the unmanned aerial vehicle, and plan the second scanning of the target forest area by the unmanned aerial vehicle according to the health risk level.

[0038] Compared with the prior art, the present application has the following advantages:

[0039] The spectral and texture features of the target forest area are extracted through satellite remote sensing data, the intelligent identification and hierarchical division of the feature region are realized based on a semantic segmentation algorithm, and a non-uniform priority monitoring grid containing a feature density index is generated. The unmanned aerial vehicle is scheduled to perform the first scanning on the high-priority grid, the satellite multispectral data, the high-resolution image of the unmanned aerial vehicle and the LiDAR point cloud data are fused, the tree species are identified, and a health index heat map is generated, so that the high-risk area can be quickly identified, the scanning range of the unmanned aerial vehicle can be significantly compressed, and the monitoring efficiency can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 A flowchart of a forestry intelligent surveying and mapping method based on regional feature feedback is proposed for the present application;

[0041] Figure 2 A method flow chart for tree species identification classification in the present application;

[0042] Figure 3 A flow chart for obtaining a health index heat map in the present application;

[0043] Figure 4 A method flow chart for secondary scanning of the unmanned aerial vehicle in the present application;

[0044] Figure 5 A structure block diagram of a forestry intelligent surveying and mapping system based on regional feature feedback proposed in the present application. DETAILED DESCRIPTION

[0045] The following description is used to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be thought of by those skilled in the art.

[0046] REFERENCE Figures 1-5 As shown in the figure, a forestry intelligent surveying and mapping method based on regional feature feedback, the method comprises:

[0047] Step 1: Extract the spectral and texture features of the target forest area image through satellite remote sensing data, realize the identification and hierarchical division of the feature area by using intelligent segmentation algorithm, and generate a priority monitoring grid containing quantitative evaluation indicators;

[0048] Step 2: Schedule the unmanned aerial vehicle to perform the first scanning on the priority monitoring grid, integrate the image data of the unmanned aerial vehicle and the satellite, fuse the spectral features and three-dimensional structure parameters to construct a tree species identification model, and establish a category system with tree species nouns as labels;

[0049] Step 3: Quantify the deviation degree of each region health state based on the health tree species image feature threshold library, and generate a health index heat map of the target forest area through spatial statistical analysis;

[0050] Step 4: Establish a dynamic mapping relationship between the health index and the risk level, dynamically update the category system based on the scanning results of the unmanned aerial vehicle, and plan the secondary scanning of the target forest area by the unmanned aerial vehicle according to the health risk level.

[0051] Specifically, the extraction of the spectral and texture features of the target forest area image through satellite remote sensing data, the realization of the identification and hierarchical division of the feature area by using intelligent segmentation algorithm, and the generation of a priority monitoring grid containing quantitative evaluation indicators, specifically includes:

[0052] Obtain the multispectral satellite remote sensing image of the target forest area, extract the spectral reflectance after preprocessing through radiation calibration and atmospheric correction, and calculate the texture feature parameters based on the gray level co-occurrence matrix to generate the preliminary feature area;

[0053] The normalized spectral reflectance and the texture feature parameter value are spliced into a feature vector, the preliminary feature region is clustered based on the vector, and a first-level cluster label is assigned;

[0054] The satellite remote sensing image and the cluster label are input into a semantic segmentation model, and after optimization by a cross-entropy loss function, a probability distribution map of each pixel belonging to different feature categories is output, and a non-uniform grid is divided based on the probability distribution density using an adaptive method, denoted as a monitoring grid;

[0055] The ratio of the sum of the pixel category probabilities in each grid to the grid area is calculated as a feature density index, and a priority area is divided according to a preset density threshold;

[0056] When necessary, the priority area is topologically checked to ensure the spatial continuity of adjacent grids; for example, the preset density threshold can be 0.85, which is obtained by querying the 2024 forest remote sensing monitoring technical specification.

[0057] Referring to Figure 2 The scheduling unmanned aerial vehicle performs a first scan on the priority monitoring grid, integrates the image data of the unmanned aerial vehicle and the satellite, fuses the spectral features and three-dimensional structure parameters to construct a tree species identification model, and establishes a category system with tree species nouns as labels, specifically including:

[0058] High-spectral imaging and LiDAR point cloud data of the area to be scanned are collected, the Euclidean distance of each pixel to the standard tree species spectral library is calculated as a spectral fingerprint difference, and DEM and DSM are generated based on the point cloud, and the leaf layer density gradient and the crown layer concave-convex degree are extracted as vertical structure parameters;

[0059] The spectral fingerprint difference and the vertical structure parameters are fused by a random forest model to dynamically distinguish tree species, and the proportion of pixels of each tree species in the grid is calculated as its distribution area;

[0060] The unmanned aerial vehicle spectral fingerprint and the satellite remote sensing image are combined for mixed pixel decomposition, the proportion of each tree species in the satellite remote sensing image pixel is calculated, the tree species composition of the preliminary feature region is analyzed, and the first-level cluster label is refined into a second-level cluster label according to the tree species name;

[0061] The tree species in the non-preliminary feature region are clustered, and when the distribution area of a certain type of tree species accumulates to a cluster threshold, a second-level cluster label is assigned;

[0062] For example, the cluster threshold is dynamically calculated according to the total grid area of the target forest area, and the cluster threshold = total grid area x 0.15%.

[0063] Referring to Figure 3As shown, the health tree species image feature threshold library quantifies the deviation degree of each region health status, and generates a health index heat map of the target forest area through spatial statistical analysis, specifically including:

[0064] For each tree species in the secondary clustering label, the spectral reflectance threshold and crown concave-convex degree threshold under the health status are obtained, and each threshold is adjusted according to the seasonal dynamics, and the seasonal dynamic adjustment rule is that the threshold is increased by 15% in summer and decreased by 10% in winter;

[0065] The spectral reflectance difference and crown concave-convex degree difference between the target tree species and the health threshold are calculated pixel by pixel, and the 3 The abnormal difference data is removed according to the principle; 3 The data exceeding the mean value ±2.5 times the standard deviation is specifically removed;

[0066] The health coefficient of the tree species is generated by weighted summation with the distribution area of the tree species as the weight coefficient;

[0067] The data missing area is compensated by the Kriging spatial interpolation method, the health coefficients of each tree species are summarized and weighted averaged, and the health index heat map of the target forest area is output.

[0068] Specifically, the dynamic mapping relationship between the health index and the risk level is established, and the category system is dynamically updated based on the scanning results of the unmanned aerial vehicle, specifically including:

[0069] The health coefficient threshold interval is set, the secondary clustering label with a total health coefficient exceeding the upper limit of the threshold interval is marked as a dangerous label, the label in the interval is marked as a potential dangerous label, and the label below the lower limit of the threshold interval is marked as a low dangerous label;

[0070] For example, the health coefficient threshold interval can be set to [0.35, 0.65], 0.35 corresponds to the disease outbreak threshold defined in the forest health evaluation technical regulation, and 0.65 is the 25th percentile of the healthy tree species sample library;

[0071] When the unmanned aerial vehicle scans a new to-be-scanned area, the total health coefficient of the secondary clustering label is recalculated;

[0072] Through the stream computing engine, the total health coefficient is compared with the health coefficient threshold, and the secondary clustering label and its label mark are updated.

[0073] Referring to Figure 4 As shown, the secondary scanning of the target forest area by the unmanned aerial vehicle according to the health risk level specifically includes:

[0074] Based on the health index heat map and the health coefficient threshold, the target forest area is divided into a dangerous area, a potential dangerous area, and a low dangerous area;

[0075] If the area proportion of the dangerous area in the feature area does not reach the set proportion of the dangerous area coverage rate, the dangerous area not covered is scanned again by the unmanned aerial vehicle, a certain proportion of overlapping scanning bands are set, the non-dangerous area is sampled and scanned, and hierarchical sensor configuration is adopted according to the regional danger level;

[0076] Exemplarily, the dangerous area is scanned by 50m low-altitude multispectral LiDAR, the potential dangerous area is monitored by 100m high-altitude visible light + thermal infrared, and the low-dangerous area is sampled and scanned by visible light, so that the monitoring efficiency of the unmanned aerial vehicle is comprehensively improved;

[0077] Exemplarily, the set proportion of the dangerous area coverage rate can be 80%, and the width of the overlapping scanning band can be 12±2%. The 80% coverage rate ensures the integrity of the key area monitoring, and the 12% overlapping band can meet the image splicing accuracy requirement.

[0078] The dangerous area is tilt-photographed to generate a 3D disease canopy model, and a real-time disease and pest identification is performed in combination with an airborne artificial intelligence algorithm to adjust flight parameters.

[0079] Referring to Figure 5 The scheme proposes a forestry intelligent surveying and mapping system based on regional feature feedback, which is used to implement the above-mentioned forestry intelligent surveying and mapping method based on regional feature feedback, and includes:

[0080] A multi-source remote sensing feature extraction and hierarchical module is configured to extract spectral and texture features of a target forest area image through satellite remote sensing data, to realize identification and hierarchical division of a feature area by using an intelligent segmentation algorithm, and to generate a priority monitoring grid containing quantitative evaluation indexes;

[0081] An air-ground collaborative tree species identification module is configured to schedule an unmanned aerial vehicle to perform first scanning on the priority monitoring grid, to integrate image data of the unmanned aerial vehicle and the satellite, to construct a tree species identification model by fusing spectral features and three-dimensional structure parameters, and to establish a category system with tree species nouns as labels;

[0082] A health dynamic evaluation module is configured to quantify a health state deviation degree of each area based on a health tree species image feature threshold library, and to generate a health index heat map of the target forest area through spatial statistical analysis;

[0083] An adaptive response control module is configured to establish a dynamic mapping relationship between the health index and the risk level, to dynamically update the category system based on scanning results of the unmanned aerial vehicle, and to plan a second scanning of the target forest area by the unmanned aerial vehicle according to the health risk level.

[0084] The air-ground collaborative tree species identification module specifically includes:

[0085] A multi-source data fusion and feature extraction unit is configured to realize collaborative processing and feature quantization of hyperspectral and LiDAR data;

[0086] A dynamic tree species classification decision unit is configured to fuse multi-dimensional features to realize accurate tree species classification and area statistics.

[0087] A label system optimization unit is configured to complete dynamic construction and update of the label system.

[0088] The health dynamic assessment module specifically comprises:

[0089] A health feature threshold dynamic management unit is configured to realize seasonal self-adaptive adjustment and verification of the health threshold of multiple tree species.

[0090] A health anomaly detection and coefficient calculation sub-unit is configured to accurately quantify the health degree of a single tree species.

[0091] A space modeling and heat map generation unit is configured to construct a global health spatial distribution model and visualize the model.

[0092] The adaptive response control module specifically comprises:

[0093] A dynamic risk grading and label management unit is configured to realize dynamic quantification of health risks and real-time update of labels.

[0094] An intelligent scanning scheduling unit is configured to realize adaptive unmanned aerial vehicle task planning based on risk levels.

[0095] A real-time decision and 3D modeling unit is configured to realize accurate identification of diseases and dynamic optimization of flight parameters.

[0096] The above shows and describes the basic principles, main features and advantages of the present application. It should be understood by those skilled in the art that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A forestry intelligent mapping method based on regional feature feedback, characterized in that, The method includes: Step 1: Extract spectral and texture features of the target forest area image from satellite remote sensing data, use intelligent segmentation algorithm to identify and classify feature regions, and generate priority monitoring grids for quantitative assessment indicators; Step 2: Schedule the UAV to perform the first scan of the priority monitoring grid, integrate the image data of the UAV and the satellite, fuse spectral features and three-dimensional structural parameters to build a tree species identification model, and establish a category system with tree species names as labels; Step 3: Based on the healthy tree species image feature threshold library, quantify the degree of deviation of the health status of each region, and generate a health index heat map of the target forest area through spatial statistical analysis; Step 4: Establish a dynamic mapping relationship between health index and risk level, dynamically update the category system based on the scanning results of UAVs, and plan secondary scanning of the target forest area by UAVs according to the health risk level.

2. The method according to claim 1, characterized in that, The process involves extracting spectral and texture features from target forest area images using satellite remote sensing data, employing intelligent segmentation algorithms to identify and classify feature regions, and generating a priority monitoring grid for quantitative assessment indicators. Specifically, this includes: Multispectral satellite remote sensing images of the target forest area are acquired. After radiometric calibration and atmospheric correction preprocessing, spectral reflectance is extracted, and texture feature parameters are calculated based on the gray-level co-occurrence matrix to generate preliminary feature regions. The normalized spectral reflectance and texture feature parameter values ​​are concatenated into a feature vector. Based on this vector, the preliminary feature regions are clustered and assigned first-level cluster labels. Satellite remote sensing images and cluster labels are input into the semantic segmentation model. After optimization by the cross-entropy loss function, the probability distribution map of each pixel belonging to different feature categories is output. Based on the probability distribution density, an adaptive method is used to divide the non-uniform grid, which is denoted as the monitoring grid. The ratio of the sum of pixel category probabilities in each grid to the grid area is used as the feature density index, and priority regions are divided according to a preset density threshold.

3. The method according to claim 1, characterized in that, The dispatching drone performs an initial scan of the priority monitoring grid, integrates image data from the drone and satellite, fuses spectral features and three-dimensional structural parameters to construct a tree species identification model, and establishes a category system labeled with tree species names, specifically including: Hyperspectral imaging and LiDAR point cloud data of the area to be scanned are collected. The Euclidean distance between each pixel and the standard tree species spectral library is calculated as the spectral fingerprint difference. At the same time, DEM and DSM are generated based on the point cloud, and the leaf layer density gradient and canopy concavity and convexity are extracted as vertical structure parameters. By fusing spectral fingerprint differences and vertical structure parameters through a random forest model, tree species are dynamically distinguished, and the proportion of pixels of each tree species in the grid is counted as its distribution area. By combining UAV spectral fingerprints with satellite remote sensing images for hybrid pixel decomposition, the proportion of each tree species in the satellite remote sensing image pixels is calculated, the tree species composition of the preliminary feature region is analyzed, and the first-level cluster label is refined into a second-level cluster label based on the tree species name. Cluster the tree species in the non-initial feature regions. When the cumulative distribution area of ​​a certain tree species reaches the clustering threshold, assign a secondary clustering label.

4. The method according to claim 3, characterized in that, The method, based on a healthy tree species image feature threshold library, quantifies the degree of deviation in health status of each region and generates a health index heatmap of the target forest area through spatial statistical analysis, specifically including: For each tree species in the secondary cluster labels, obtain its spectral reflectance threshold and canopy concavity threshold under healthy conditions, and dynamically adjust each threshold according to the season; The difference in spectral reflectance and canopy concavity / convexity between the target tree species and the health threshold are calculated pixel-by-pixel, using 3D modeling. Data with outlier values ​​should be removed as a rule. The health coefficient of a tree species is generated by weighting and summing the distribution area of ​​the tree species as a weighting coefficient. Kriging space interpolation was used to compensate for missing data areas, the health coefficients of each tree species were summarized and weighted averaged, and a heat map of the health index of the target forest area was output.

5. The method according to claim 4, characterized in that, The establishment of a dynamic mapping relationship between health indices and risk levels, and the dynamic updating of the category system based on the scanning results of drones, specifically includes: Set a health coefficient threshold range, and mark the secondary cluster labels that exceed the upper limit of the threshold range as dangerous labels, those within the range as potentially dangerous labels, and those below the lower limit of the threshold range as low-risk labels; When the drone scans a new area to be scanned, the total health coefficient of the secondary cluster label is recalculated; The total health coefficient is compared with the health coefficient threshold using a streaming computing engine, and the secondary cluster labels and their tags are updated accordingly.

6. The method according to claim 5, characterized in that, The specific details of planning the secondary scanning of the target forest area by drones based on the health risk level include: Based on the health index heat map and health coefficient threshold, the target forest area is divided into dangerous areas, potentially dangerous areas and low-risk areas; If the proportion of dangerous area in the feature area does not reach the set proportion of dangerous area coverage, then the uncovered dangerous area will be scanned a second time by drone, and a certain proportion of overlapping scan bands will be set to sample the non-dangerous area. Graded sensor configuration will be adopted according to the regional hazard level. Oblique photography of the danger zone is used to generate a 3D disease canopy model, which is then combined with airborne artificial intelligence algorithms to identify pests and diseases in real time and adjust flight parameters accordingly.

7. A forestry intelligent surveying and mapping system based on regional feature feedback, characterized in that, A method for implementing a forestry intelligent mapping method based on regional feature feedback as described in any one of claims 1-6 includes: The multi-source remote sensing feature extraction and grading module is used to extract the spectral and texture features of target forest area images from satellite remote sensing data, use intelligent segmentation algorithms to identify and grade feature regions, and generate priority monitoring grids for quantitative evaluation indicators. The air-ground collaborative tree species identification module is used to schedule UAVs to perform the first scan of the priority monitoring grid, integrate UAV and satellite image data, fuse spectral features and three-dimensional structural parameters to construct a tree species identification model, and establish a category system with tree species names as labels; The health dynamic assessment module is used to quantify the degree of deviation of the health status of each region based on the image feature threshold library of healthy tree species, and generate a health index heat map of the target forest area through spatial statistical analysis. The adaptive response control module is used to establish a dynamic mapping relationship between health index and risk level, dynamically update the category system based on the scanning results of UAV, and plan the secondary scanning of the target forest area by UAV according to the health risk level.

8. A forestry intelligent surveying and mapping system based on regional feature feedback according to claim 7, characterized in that, The air-ground collaborative tree species identification module specifically includes: A multi-source data fusion and feature extraction unit is used to achieve collaborative processing and feature quantization of hyperspectral and LiDAR data; A dynamic tree species classification decision unit is used to integrate multi-dimensional features to achieve accurate tree species classification and area statistics. The tag system optimization unit is used to dynamically build and update the tag system.

9. A forestry intelligent surveying and mapping system based on regional feature feedback according to claim 7, characterized in that, The health dynamic assessment module specifically includes: A dynamic management unit for health feature thresholds is used to achieve seasonal adaptive adjustment and verification of health thresholds for multiple tree species; The health anomaly detection and coefficient calculation subunit is used to accurately quantify the health status of a single tree species; The spatial modeling and heatmap generation unit is used to construct and visualize a global health spatial distribution model.

10. A forestry intelligent surveying and mapping system based on regional feature feedback according to claim 7, characterized in that, The adaptive response control module specifically includes: The dynamic risk classification and label management unit is used to realize the dynamic quantification of health risks and the real-time updating of labels; Intelligent scanning and scheduling unit for adaptive UAV mission planning based on risk level; The real-time decision-making and 3D modeling unit is used to achieve accurate disease identification and dynamic optimization of flight parameters.

Images