Forestry resource remote sensing analysis method and system monitored by unmanned aerial vehicle

Abstract

The invention discloses a forestry resource remote sensing analysis method and system monitored by an unmanned aerial vehicle, and relates to the technical field of forestry resource analysis, and the method comprises the steps: starting the unmanned aerial vehicle to carry out the data collection and data screening of a target forestry region, and obtaining a plurality of remote sensing image data; feature extraction is carried out, clustering analysis is carried out on the multiple remote sensing image data, and multiple forestry resource types are determined; performing health assessment on the target forestry area, generating a health score, and constructing a forestry resource change trend chart; identifying the target forestry area, and determining a resource distribution map; and traversing the resource distribution map to perform management analysis, generating a management protection decision, and performing intelligent management monitoring. The technical problems that in the prior art, forestry resource monitoring data are not comprehensively acquired, and health conditions and dynamic changes cannot be scientifically evaluated are solved, and the technical effects that health evaluation, change prediction, distribution drawing and intelligent management of the target forestry area are achieved, and monitoring and management scientificity is improved are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of forestry resource analysis, and particularly to a remote sensing analysis method and system for forestry resources monitored by unmanned aerial vehicles (UAVs). Background Art

[0002] In traditional forestry resource monitoring, it mainly relies on methods such as manual field surveys and ground monitoring stations. Manual field surveys not only consume a large amount of manpower, material resources and time, but also are restricted by factors such as terrain and traffic, making it difficult to comprehensively and timely monitor large forestry areas, resulting in limitations and lags in the obtained data. Although ground monitoring stations can obtain data of local areas in real time to a certain extent, due to their limited distribution density, it is difficult to cover the entire forestry area and cannot meet the requirements for all-round and refined monitoring of forestry resources. From the perspective of data acquisition, traditional methods are difficult to obtain comprehensive forestry resource information. For example, in mountainous areas, remote areas or areas with complex terrain, manual field surveys are difficult and the data collection cost is high, resulting in serious data gaps in these areas. In terms of the assessment of the health status of forestry resources, there is a lack of scientific and effective assessment methods. In the past, the health of forests was mostly judged through simple indicators, such as observing the growth trend of trees and the surface signs of pest and disease occurrences. This method is highly subjective, lacking systematicness and scientificity, and cannot accurately assess the overall health status of the forest ecosystem, nor can it predict potential health risks in advance.

[0003] The prior art has technical problems such as incomplete acquisition of forestry resource monitoring data, inability to scientifically evaluate the health status, and predict dynamic changes. Summary of the Invention

[0004] The present application provides a remote sensing analysis method and system for forestry resources monitored by UAVs, which are used to solve the technical problems in the prior art of incomplete acquisition of forestry resource monitoring data, inability to scientifically evaluate the health status, and predict dynamic changes.

[0005] In view of the above problems, the present application provides a remote sensing analysis method and system for forestry resources monitored by UAVs.

[0006] In the first aspect of the present application, a remote sensing analysis method for forestry resources monitored by UAVs is provided. The method includes: Start the drone to collect data on the target forestry area, obtain a multi-source dataset of the target forestry area, filter the data based on the multi-source dataset to obtain multiple remote sensing image data; extract features based on the multiple remote sensing image data, perform clustering analysis on the multiple remote sensing image data according to multiple image features to determine multiple forestry resource types; perform a health assessment on the target forestry area according to the multiple forestry resource types with the multiple image features, generate a health score, predict the dynamic changes of forestry resources based on the health score, and construct a forestry resource change trend map; map the multiple remote sensing image data to the forestry resource change trend map to identify the target forestry area and determine the resource distribution map; traverse the resource distribution map for management analysis, generate management and protection decisions, and transmit the management and protection decisions to the drone for intelligent management and monitoring of the target forestry area.

[0007] In the second aspect of the present application, a remote sensing analysis system for forestry resources monitored by drones is provided. The system includes: A remote sensing image data acquisition module, which is used to start the drone to collect data on the target forestry area, obtain a multi-source dataset of the target forestry area, and filter the data based on the multi-source dataset to obtain multiple remote sensing image data; a forestry resource type determination module, which is used to extract features based on the multiple remote sensing image data, perform clustering analysis on the multiple remote sensing image data according to multiple image features to determine multiple forestry resource types; a forestry resource change trend map construction module, which is used to perform a health assessment on the target forestry area according to the multiple forestry resource types with the multiple image features, generate a health score, predict the dynamic changes of forestry resources based on the health score, and construct a forestry resource change trend map; a resource distribution map determination module, which is used to map the multiple remote sensing image data to the forestry resource change trend map to identify the target forestry area and determine the resource distribution map; an intelligent management and monitoring module, which is used to traverse the resource distribution map for management analysis, generate management and protection decisions, and transmit the management and protection decisions to the drone for intelligent management and monitoring of the target forestry area.

[0008] One or more technical solutions provided in the present application have at least the following technical effects or advantages: Start the drone to collect data on the target forestry area, obtain a multi-source dataset of the target forestry area, perform data screening based on the multi-source dataset to obtain multiple remote sensing image data; perform feature extraction on the multiple remote sensing image data, conduct clustering analysis on the multiple remote sensing image data to determine multiple forestry resource types; perform a health assessment on the target forestry area according to multiple image features according to multiple forestry resource types, generate a health score, predict the dynamic changes of forestry resources based on the health score, and construct a forestry resource change trend map; map the multiple remote sensing image data to the forestry resource change trend map to identify the target forestry area and determine the resource distribution map; traverse the resource distribution map for management analysis, generate management and protection decisions, and conduct intelligent management and monitoring. It achieves the technical effect of realizing the health assessment, change prediction, distribution mapping and intelligent management of the target forestry area, and improving the scientific nature of monitoring and management. Brief Description of the Drawings

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for description in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0010] Figure 1 It is a schematic flowchart of a remote sensing analysis method for forestry resources monitored by a drone provided by an embodiment of the present application; Figure 2 It is a schematic structural diagram of a remote sensing analysis system for forestry resources monitored by a drone provided by an embodiment of the present application.

[0011] Description of reference numerals: Remote sensing image data acquisition module 10, forestry resource type determination module 20, forestry resource change trend map construction module 30, resource distribution map determination module 40, intelligent management and monitoring module 50. Detailed Embodiments

[0012] The present application provides a remote sensing analysis method and system for forestry resources monitored by a drone, which are used to solve the technical problems of incomplete acquisition of forestry resource monitoring data, inability to scientifically evaluate the health status and predict dynamic changes in the prior art.

[0013] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0014] Embodiment 1, asFigure 1 As shown in the figure, the present application provides a remote sensing analysis method for forestry resources monitored by an unmanned aerial vehicle (UAV), and the method includes: Step S100: Start the UAV to collect data on the target forestry area, obtain a multi-source data set of the target forestry area, and perform data screening based on the multi-source data set to obtain multiple remote sensing image data.

[0015] Specifically, when using the UAV to monitor forestry resources, according to information such as the geographical location and scope of the target forestry area, the flight route, altitude, data collection frequency, and sensor parameters of the UAV are preset in advance. After preparation, the UAV is started to fly and operate over the target forestry area according to the established plan. During the flight, various sensors carried by the UAV, such as optical cameras and spectrometers, collect data in all directions. These data are widely sourced and include image data in different spectral bands, geographical location information, terrain data, etc., thus forming a multi-source data set of the target forestry area. In order to obtain key data suitable for subsequent analysis, it is necessary to screen the multi-source data set, eliminate data with low relevance to forestry resource analysis, and focus on selecting multiple remote sensing image data with qualified quality and meeting the requirements of forestry resource analysis. These remote sensing image data will become an important basis for accurately analyzing forestry resources subsequently.

[0016] Step S200: Perform feature extraction based on the multiple remote sensing image data, perform clustering analysis on the multiple remote sensing image data according to multiple image features, and determine multiple forestry resource types.

[0017] Specifically, after obtaining the remote sensing image data, perform feature extraction on the multiple remote sensing image data. Through geometric transformation, perform data registration to correct image deviations caused by differences in acquisition angles, positions, etc., so that different images have consistency in space; then divide the data according to spectral bands to obtain multiple spectral band image data. Then use a convolutional neural network to construct a convolutional kernel sliding window, traverse the spectral band image data, extract local image features such as texture, color, and shape, and then simplify the data through pooling dimensionality reduction operations to determine multiple image features for subsequent analysis. Based on these image features, perform clustering analysis on the multiple remote sensing image data. First, perform multi-dimensional calibration on the image data to reduce systematic errors, and then randomly select K initial clustering centers from the calibrated image features. Allocate data according to the similarity between each image and the clustering centers to determine the initial number of clusters. To improve the clustering accuracy, retrieve the historical forestry record log of the target forestry area, recompute the clustering data in combination with historical data, and update the K initial clustering centers to obtain more accurate P clustering centers. Finally, traverse the multiple remote sensing image data according to the P clustering centers, and classify images with similar features into one category, thereby determining multiple forestry resource types and providing a classification basis for subsequent forestry resource management.

[0018] Step S300: Conduct a health assessment on the target forestry area according to the multiple forestry resource types for the multiple image features, generate a health score, predict the dynamic changes of forestry resources based on the health score, and construct a forestry resource change trend chart.

[0019] Specifically, after determining the multiple image features and forestry resource types, perform fine instance segmentation on the target forestry area according to different forestry resource types, and divide it into multiple forestry sub-areas. For each sub-area, organically combine the corresponding multiple image features to construct a representative image feature vector to comprehensively reflect the feature information of the sub-area. Then introduce the data acquisition time interval information, and combine it with the historical forestry record log of the target forestry area to conduct in-depth health analysis on these image feature vectors, comprehensively considering factors such as vegetation growth cycle, occurrence law of pests and diseases, and climate impact, and construct multiple scientific and reasonable health assessment indicators. Based on these indicators, quantitatively evaluate the health status of each forestry sub-area to obtain multiple regional health scores. To accurately reflect the importance of each sub-area in the overall forestry resources, assign weights to the multiple forestry sub-areas to obtain multiple weight coefficients, and perform weighted calculation with the corresponding regional health scores to generate a health score that can represent the entire target forestry area. Based on this health score, combined with historical data and relevant model algorithms, predict the dynamic changes of forestry resources in a future period of time, such as the increase or decrease of vegetation coverage area, the spread trend of forest pests and diseases, etc. Finally, with time as the horizontal axis, visually present the predicted changes in forestry resources to construct a forestry resource change trend chart, providing an intuitive and important decision-making basis for the scientific management and protection of forestry resources.

[0020] Step S400: Map the multiple remote sensing image data to the forestry resource change trend chart to identify the target forestry area and determine the resource distribution map.

[0021] Specifically, after obtaining the forestry resource change trend diagram, multiple remote sensing image data are associated with the trend diagram to determine the resource distribution map. First, according to the data acquisition time interval information, the forestry resource change trend diagram is continuously tracked, and the change trajectory parameters are detailedly recorded and plotted. These parameters reflect the dynamic change characteristics of the trend diagram over time. Then, multiple remote sensing image data are synchronized to the forestry resource change trend diagram according to the change trajectory parameters. Through correlation analysis, multiple image-resource pairs are generated, and the corresponding relationship between the image data and the forestry resource change is established. Subsequently, within the scope of the target forestry area, these image-resource pairs are traversed, and matching identification is carried out according to certain matching rules to generate multiple image resource change labels, which clarify the forestry resource change situations in different regions. On this basis, three-dimensional analysis of the target forestry area is carried out based on multiple remote sensing image data, and an accurate geospatial model is constructed. According to the image resource change labels, multiple spatial pixel features are extracted from the remote sensing image data respectively, and multiple spatial change features are extracted from the forestry resource change trend diagram. Finally, using these spatial pixel features and spatial change features, multiple remote sensing image data and the forestry resource change trend diagram are accurately mapped into the geospatial of the target forestry area, and an intuitive and clear resource distribution map is integrated, visually presenting the distribution status and change trend of forestry resources in different regions.

[0022] Step S500: Traverse the resource distribution map for management analysis, generate management and protection decisions, and transmit the management and protection decisions to the unmanned aerial vehicle to conduct intelligent management and monitoring of the target forestry area.

[0023] Specifically, after completing the drawing of the resource distribution map, carefully traverse the resource distribution map, and deeply analyze information such as the distribution characteristics of forestry resources, the health status of resources in different regions, the change trends of various resources, and the mutual relationships between resources. Combining the overall situation of the forestry ecosystem, historical data, current problems, and future development needs, comprehensively applying forestry scientific knowledge, ecological protection theories, and data analysis techniques, a series of scientific and reasonable management and protection decisions are formulated. These decisions cover forest tending plans, specifying which areas need to be thinned and fertilized to promote the growth of forest trees; pest control strategies, determining the key monitoring and control areas and the control methods adopted; ecological protection measures, such as designating nature reserves and restricting human activity areas, etc. After the decision-making is completed, these management and protection decisions are transmitted to the unmanned aerial vehicle performing the task through wireless communication technology. After receiving the instructions, the unmanned aerial vehicle carries out intelligent management and monitoring work in the target forestry area according to the decision content. For example, conduct encrypted monitoring of key areas according to the set flight route, timely detect signs of pests or abnormal resource changes; perform precise operations on the areas that need to be tended to improve the efficiency and accuracy of forestry resource management, and achieve the effective protection and sustainable development of the target forestry area.

[0024] In a possible implementation manner, step S200 further includes: Step S210: Perform data registration on the multiple remote sensing image data through geometric transformation to obtain multiple image registration data.

[0025] Step S220: Divide the multiple image registration data according to spectral bands to determine multiple spectral band image data.

[0026] Step S230: Construct a convolutional kernel sliding window using a convolutional neural network, and traverse the multiple spectral band image data through the convolutional kernel sliding window for feature analysis to obtain multiple local image features.

[0027] Step S240: Perform pooling dimensionality reduction based on the multiple local image features to determine the multiple image features.

[0028] Specifically, when processing and analyzing multiple remote sensing image data, due to differences in the shooting angle, shooting time, shooting equipment, etc. of the acquired remote sensing image data, there may be inconsistencies in position, scale, rotation, etc. between the images. Therefore, it is necessary to perform data registration on the multiple remote sensing image data through geometric transformation. Using geometric transformation operations such as translation, rotation, and scaling, with a certain reference image as a reference, other images are adjusted to make them consistent in the geographical space, and finally multiple image registration data are obtained, providing a reliable basis for subsequent analysis.

[0029] After obtaining multiple image registration data, these data are divided according to spectral bands. Different spectral bands can reflect different characteristic information of forestry resources. For example, the near-infrared band is sensitive to the growth status of vegetation, while the visible light band can present the color and texture of vegetation, etc. By dividing the spectral bands, the image data are separated according to different spectral ranges, thereby determining multiple spectral band image data, which is convenient for more detailed analysis of forestry resource information under different spectral characteristics.

[0030] In order to extract useful features from the spectral band image data, a convolutional kernel sliding window is constructed using a convolutional neural network. The convolutional neural network has strong feature extraction capabilities, and its convolutional kernel can slide on the image data to extract local features in the data through convolutional operations. When constructing the convolutional kernel sliding window, parameters are set according to the characteristics of the spectral band image data. Then, this convolutional kernel sliding window is used to traverse multiple spectral band image data, perform convolutional operations on each pixel point and its neighborhood, and after a series of convolutional operations, multiple local image features are obtained. These local features contain rich detail information such as texture and shape.

[0031] After obtaining multiple local image features, the data dimension is often high, which not only increases the computational amount but also may lead to overfitting of the model. Therefore, a pooling dimensionality reduction operation is performed based on these local image features. The pooling operation can reduce the data dimension while retaining the main features. Common pooling methods include max pooling and average pooling, etc. By using these methods to process the local image features, the spatial resolution of the data is reduced, thereby determining multiple image features. These image features after dimensionality reduction not only retain the key information of the original image but also improve the efficiency of subsequent analysis and processing, providing strong support for subsequent tasks such as determining forestry resource types.

[0032] In a possible implementation manner, step S230 further includes: Step S231: Perform pixel analysis according to the multiple spectral band image data to generate pixel parameters.

[0033] Step S232: Determine whether the pixel parameters are greater than or equal to a preset pixel threshold. When the pixel parameters are greater than or equal to the preset pixel threshold, generate a cropping instruction, and perform image cutting on the multiple spectral band image data through the cropping instruction to determine multiple image blocks.

[0034] Step S233: Activate the convolutional kernel filter according to the convolutional neural network, and set the convolutional kernel sliding window according to the multiple image blocks.

[0035] Specifically, for the multiple spectral band image data, each pixel is analyzed in detail one by one, considering multiple dimensions such as the gray value, color information of the pixel, and the relationship with surrounding pixels, and then pixel parameters that can comprehensively reflect the pixel characteristics are generated. These pixel parameters provide a basis for subsequent judgment of the feature importance of the image data.

[0036] Compare the generated pixel parameters with a preset pixel threshold. This preset pixel threshold is determined based on a large amount of experimental data and research on the characteristics of forestry resource remote sensing images. It is the standard for judging which parts of the image data contain key features. When the pixel parameters are greater than or equal to the preset pixel threshold, it means that this part of the image data may contain important forestry resource feature information. At this time, a cropping instruction is generated. According to this instruction, perform image cutting on the multiple spectral band image data, separate the area containing the key features from the original image, and form multiple image blocks. These image blocks focus on the key information in the image, reduce the data volume for subsequent processing, and at the same time highlight the content that needs to be analyzed key.

[0037] After determining multiple image blocks, the convolutional kernel filter is activated by means of a convolutional neural network. The convolutional kernel filter is the core component for feature extraction in a convolutional neural network, and its parameters and structure determine the types of features that can be extracted. According to the size of the image block, a convolutional kernel sliding window that matches it is set so that the window size is the same as the image block size. In this way, when subsequently traversing the spectral band image data with the convolutional kernel sliding window, the key features in the image block can be extracted more pertinently, improving the accuracy and efficiency of feature extraction and laying a foundation for obtaining high-quality local features of the image.

[0038] In a possible implementation manner, step S200 further includes: Step S250: Perform multi-dimensional correction on the multiple remote sensing image data to generate an image correction result, and randomly select according to the multiple image features from the image correction result to determine K initial cluster centers.

[0039] Step S260: Traverse the multiple remote sensing image data according to the K initial cluster centers for data assignment to determine multiple cluster numbers, where K is a positive integer greater than 0.

[0040] Step S270: Retrieve the historical forestry record log of the target forestry area, recalculate the multiple cluster numbers based on the historical forestry record log, update the K initial cluster centers, and determine P cluster centers, where P is a positive integer greater than 0.

[0041] Step S280: Traverse the multiple remote sensing image data according to the P cluster centers for forestry classification to determine the multiple forest resource types.

[0042] Specifically, since multi-source remote sensing data may be affected by various factors during the acquisition process, there are spatial position deviations and radiation differences. Therefore, multi-dimensional correction needs to be performed on the multiple remote sensing image data, including geometric correction and radiation correction specifically. Geometric correction can eliminate spatial position errors caused by shooting angles, terrain undulations, etc., so that different images are accurately aligned in space; radiation correction can reduce the influence of factors such as lighting conditions and sensor differences on the radiation values of the images, ensuring the consistency of the images in terms of radiation. After the correction is completed, an image correction result is generated. Subsequently, based on the multiple image features extracted previously, K data points are randomly selected from the image correction result as the initial cluster centers. K is a positive integer greater than 0, and these initial cluster centers will be used as the starting reference points for subsequent cluster analysis.

[0043] Based on these K initial clustering centers, traverse multiple remote sensing image data comprehensively. By calculating the similarity between each data point and each initial clustering center (using a distance metric method), assign each data point to the category where the initial clustering center it is most similar to is located. After completing the assignment of all data points, count the number of data points in each category to determine multiple clustering numbers.

[0044] To make the clustering results more accurate and in line with the actual situation, retrieve the historical forestry record logs of the target forestry area. These logs contain the past forestry resource information of this area, such as the distribution and growth status of different types of forests. Based on the historical forestry record logs, recalculate the current multiple clustering numbers. Here, the calculation is specifically to find the mean of all data points currently assigned to this cluster. In this process, K clustering centers are initially set, but these initial clustering centers do not fully conform to the actual situation. For example, when K makes a rough division of the area, due to the large division scale, it cannot cover all surface forestry types. Some special vegetation communities or newly emerging forestry resource changes are not covered. At this time, it is necessary to further refine the clustering, and the number of clustering centers obtained after recalculation will increase, that is, P is greater than K. On the contrary, if K initially divides the large categories into too many small categories during the initial setting, resulting in insignificant differences between some clusters and over-segmentation, then after recalculation based on the historical forestry record logs, these small categories will be merged, and the finally determined number of clustering centers P will be less than K. After such recalculation and adjustment, finally determine P clustering centers.

[0045] Using the determined P clustering centers, traverse the multiple remote sensing image data again. By comparing the similarity between each data point and these P clustering centers, divide each data point into the category represented by the clustering center it is most similar to. After such classification operations, the forestry resource information reflected by the multiple remote sensing image data can be classified, thereby determining multiple forestry resource types, providing an important basis for subsequent forestry resource management and protection.

[0046] In a possible implementation manner, step S300 further includes: Step S310: Perform instance segmentation on the target forestry area according to the multiple forestry resource types to determine multiple forestry sub-areas.

[0047] Step S320: Combine the multiple image features according to the multiple forestry sub-areas to construct an image feature vector.

[0048] Step S330: Introduce the data acquisition time interval information and combine it with the historical forestry record logs to perform a health analysis on the image feature vector and construct multiple health assessment indicators.

[0049] Step S340: Based on the multiple health assessment indicators, conduct a health assessment on the multiple forestry sub-regions to obtain multiple regional health scores.

[0050] Step S350: Assign weights to the multiple forestry sub-regions to obtain multiple weight coefficients, and perform weighted calculation on the multiple regional health scores according to the multiple weight coefficients to generate the health score.

[0051] Specifically, based on the determined multiple forestry resource types, conduct detailed instance segmentation on the target forestry area. Different forestry resource types often have different distribution characteristics in the geographical space. Through advanced image segmentation algorithms, combined with information such as the spectrum and texture of remote sensing image data, the target forestry area is divided into multiple forestry sub-regions corresponding to different forestry resource types, laying a foundation for subsequent precise analysis of each sub-region.

[0052] Reasonably combine the previously extracted multiple image features according to each forestry sub-region. Each forestry sub-region has its unique resource characteristics. By integrating the related image features, an image feature vector that can comprehensively reflect the characteristics of the sub-region is constructed. These image feature vectors contain rich information, such as the growth status and coverage of vegetation, and are the key data for subsequent health analysis.

[0053] When constructing health assessment indicators, comprehensively consider the data collection time interval information and historical forestry record logs, and conduct a health analysis on the image feature vectors. The data collection time interval information can reflect the dynamic changes of forestry resources at different time periods, such as the growth trend of vegetation in different seasons and the seasonal outbreak patterns of pests and diseases. The historical forestry record logs cover various past information of the target forestry area, such as the previous occurrence of pests and diseases, the growth cycle records of trees, and the human intervention measures and effects. Combine these two to conduct a detailed analysis of the image feature vectors. For the information such as the growth status, coverage, and spectral features of vegetation reflected by the image feature vectors, combine the time interval and historical records to determine whether it is within the normal range. For example, if in a specific season, the image feature shows that the chlorophyll content of the vegetation is lower than the average level in the same period of history and shows a continuous downward trend within the data collection time interval, then this may indicate a problem with the vegetation health. Based on such analysis, starting from multiple dimensions such as vegetation vitality, pest and disease infection degree, and ecological stability, construct multiple health assessment indicators. For example, the change rate of the normalized difference vegetation index (NDVI) can be used to measure the change of vegetation vitality, and the occurrence frequency of pests and diseases can be used as an indicator to evaluate the pest and disease infection degree, so as to comprehensively and scientifically evaluate the health status of forestry resources.

[0054] Based on the multiple health assessment indicators, a comprehensive health assessment is conducted on each forestry sub-region. For each health assessment indicator, a corresponding scoring standard is formulated, and each sub-region is scored according to its performance on different indicators, and finally multiple regional health scores are obtained. These scores intuitively reflect the health status of each forestry sub-region.

[0055] Considering that the importance and influence of different forestry sub-regions in the entire target forestry area may vary, it is necessary to assign weights to multiple forestry sub-regions. The weight coefficient of each sub-region is determined by comprehensively considering factors such as the size of the sub-region, resource scarcity, and ecological function. Then, according to these weight coefficients, the health scores of multiple regions are weighted and calculated, and the health status of each sub-region is integrated according to its importance, and finally a health score that can represent the health status of the entire target forestry area is generated.

[0056] In a possible implementation, step S400 further includes: Step S410: continuously tracking the forestry resources change trend chart according to the data collection time interval information, and drawing the change trajectory parameters of the forestry resources change trend chart.

[0057] Step S420: Synchronize the multiple remote sensing image data to the forestry resource change trend chart according to the change trajectory parameters for correlation analysis to generate multiple image-resource pairs.

[0058] Step S430: traverse the target forestry area to match and identify the multiple image-resource pairs, and generate multiple image resource change tags.

[0059] Step S440: Mapping the multiple remote sensing image data and the forestry resource change trend map to the geographic space of the target forestry area based on the multiple image resource change tags to construct the resource distribution map.

[0060] Specifically, the forest resource change trend chart is continuously tracked based on the data collection time interval information. Since data collection is not completed at one time, but is carried out at different time points, these time intervals contain key information about the changes in forest resources over time. Through continuous monitoring of the trend chart, the changes in the status of forest resources at each time point are recorded in detail, and then the change trajectory parameters of the forest resource change trend chart are drawn. These parameters can accurately reflect the dynamic evolution of forest resources in the time dimension.

[0061] Synchronously associate multiple remote sensing image data with the forestry resource change trend chart. Each remote sensing image data corresponds to the forestry resource status at a specific time point. Through the change trajectory parameters, these image data can be accurately matched to the corresponding time and status on the trend chart. After correlation analysis, integrate the image data with the corresponding forestry resource information to generate multiple image-resource pairs, which establish a close connection between the images and resource changes.

[0062] Comprehensively traverse the target forestry area. For each location point, conduct a detailed matching identification of the generated multiple image-resource pairs. Combine the spatial information of the remote sensing images and the characteristic information of the forestry resources to judge the resource change situation at this location at different time points, and assign a clear identifier to each image-resource pair, thereby generating multiple image resource change labels. These labels clearly record the forestry resource change information at each location within the target forestry area, such as vegetation growth, forest deforestation, and the impact of pests and diseases.

[0063] Map multiple remote sensing image data and the forestry resource change trend chart to the geographical space of the target forestry area. Through Geographic Information System (GIS) technology, according to the spatial and time information contained in the image resource change labels, accurately place the image data and the trend chart at the corresponding positions in the geographical space. After integration processing, construct a resource distribution map that can intuitively display the distribution and change of forestry resources in the target forestry area, providing an important visual basis for the management and decision-making of forestry resources.

[0064] In a possible implementation manner, step S440 further includes: Step S441: Perform three-dimensional analysis on the target forestry area based on the multiple remote sensing image data to construct a geographical space.

[0065] Step S442: Extract multiple spatial pixel features of the multiple remote sensing image data according to the multiple image resource change labels.

[0066] Step S443: Extract multiple spatial change features of the forestry resource change trend chart according to the multiple image resource change labels.

[0067] Step S444: Map the multiple remote sensing image data to the geographical space based on the multiple spatial pixel features, and map the forestry resource change trend chart to the geographical space based on the multiple spatial change features, and construct the resource distribution map of the target forestry area.

[0068] Specifically, multiple remote sensing image data are utilized, and three-dimensional analysis technology is applied to conduct a comprehensive analysis of the target forestry area. By comprehensively processing remote sensing images obtained from different angles and at different times, and combining geographical positioning information, an accurate three-dimensional geographical space model is constructed. This model can visually and stereoscopically display basic geographical information such as the topography and vegetation distribution of the target forestry area, providing an accurate spatial framework for subsequent resource distribution analysis.

[0069] Based on the multiple generated image resource change tags, multiple spatial pixel features are accurately extracted from multiple remote sensing image data. Each image pixel point is associated with the corresponding geographical coordinates. By analyzing the image resource change tags, the image pixel points corresponding to each tag are screened out, and their feature information such as color, brightness, and texture is extracted. These spatial pixel features detailedly record the spatial attributes of the forestry resources represented by each pixel point, providing data support at the micro level for accurately depicting the distribution of forestry resources.

[0070] Similarly, according to the multiple image resource change tags, the forestry resource change trend chart is deeply mined to extract multiple spatial change features. Each change information in the change trend chart, such as the increase or decrease in vegetation coverage area and the spread range of forest pests and diseases, is accurately located at the corresponding geographical position through the image resource change tags, so as to obtain the specific position, range, and change direction and other features of each change information in the geographical space. These spatial change features reflect the spatial presentation of the dynamic changes of forestry resources in the time series.

[0071] Based on the multiple spatial pixel features and spatial change features extracted above, the multiple remote sensing image data and the forestry resource change trend chart are respectively mapped into the constructed geographical space. The spatial pixel features extracted from the remote sensing image data are matched with the corresponding positions in the geographical space, and the image data are accurately placed in the geographical space. At the same time, according to the spatial change features, the information in the forestry resource change trend chart is also accurately integrated into the geographical space, so that the current situation and change trend of forestry resources can be visually presented at each position in the geographical space. Finally, a complete and accurate resource distribution map of the target forestry area is constructed, providing an intuitive and valuable reference basis for forestry resource management and decision-making.

[0072] Embodiment 2, based on the same inventive concept as the forestry resource remote sensing analysis method monitored by the unmanned aerial vehicle in the foregoing embodiment, as Figure 2 shown, the present application provides a forestry resource remote sensing analysis system monitored by an unmanned aerial vehicle. The system in the embodiment of the present application and the method embodiment are based on the same inventive concept. Among them, the system includes: Remote sensing image data acquisition module 10, which is used to start the drone to collect data on the target forestry area, obtain a multi-source data set of the target forestry area, and perform data screening based on the multi-source data set to obtain multiple remote sensing image data.

[0073] Forestry resource type determination module 20, which is used to perform feature extraction based on the multiple remote sensing image data, perform clustering analysis on the multiple remote sensing image data according to multiple image features, and determine multiple forestry resource types.

[0074] Forestry resource change trend map construction module 30, which is used to perform a health assessment on the target forestry area according to the multiple forestry resource types based on the multiple image features, generate a health score, predict the dynamic changes of forestry resources according to the health score, and construct a forestry resource change trend map.

[0075] Resource distribution map determination module 40, which is used to map the multiple remote sensing image data to the forestry resource change trend map to identify the target forestry area and determine the resource distribution map.

[0076] Intelligent management and monitoring module 50, which is used to traverse the resource distribution map for management analysis, generate management and protection decisions, and transmit the management and protection decisions to the drone for intelligent management and monitoring of the target forestry area.

[0077] Furthermore, the forestry resource type determination module 20 further includes: Image registration data acquisition unit, which is used to perform data registration on the multiple remote sensing image data through geometric transformation to obtain multiple image registration data.

[0078] Spectral band image data determination unit, which is used to divide the multiple image registration data according to spectral bands to determine multiple spectral band image data.

[0079] Image local feature acquisition unit, which is used to construct a convolutional kernel sliding window by using a convolutional neural network, traverse the multiple spectral band image data through the convolutional kernel sliding window for feature analysis, and obtain multiple image local features.

[0080] Image feature determination unit, which is used to perform pooling dimensionality reduction based on the multiple image local features to determine the multiple image features.

[0081] Furthermore, the image local feature acquisition unit further includes: Pixel parameter generation unit, which is used to perform pixel analysis according to the multiple spectral band image data to generate pixel parameters.

[0082] An image block determination unit, configured to determine whether the pixel parameter is greater than or equal to a preset pixel threshold. When the pixel parameter is greater than or equal to the preset pixel threshold, a cropping instruction is generated, and the multi-spectral band image data is image-cropped through the cropping instruction to determine a plurality of image blocks.

[0083] A convolution kernel sliding window setting unit, configured to activate a convolution kernel filter according to a convolutional neural network, and the convolution kernel filter sets the convolution kernel sliding window according to the plurality of image blocks.

[0084] Further, the forestry resource type determination module 20 further includes: An initial clustering center determination unit, configured to perform multi-dimensional correction on the plurality of remote sensing image data to generate an image correction result, and randomly select according to the plurality of image features for the image correction result to determine K initial clustering centers.

[0085] A clustering number determination unit, configured to traverse the plurality of remote sensing image data according to the K initial clustering centers for data allocation to determine a plurality of clustering numbers, where K is a positive integer greater than 0.

[0086] A clustering center determination unit, configured to retrieve the historical forestry record log of the target forestry area, recalculate the plurality of clustering numbers based on the historical forestry record log, update the K initial clustering centers, and determine P clustering centers, where P is a positive integer greater than 0.

[0087] A forestry classification unit, configured to traverse the plurality of remote sensing image data according to the P clustering centers for forestry classification to determine the plurality of forestry resource types.

[0088] Further, the forestry resource change trend graph construction module 30 further includes: A forestry sub-region determination unit, configured to perform instance segmentation on the target forestry area according to the plurality of forestry resource types to determine a plurality of forestry sub-regions.

[0089] An image feature vector construction unit, configured to combine the plurality of image features according to the plurality of forestry sub-regions to construct an image feature vector.

[0090] A health assessment index construction unit, configured to introduce data acquisition time interval information and combine the historical forestry record log to perform health analysis on the image feature vector to construct a plurality of health assessment indexes.

[0091] A regional health score acquisition unit, configured to perform health assessment on the plurality of forestry sub-regions based on the plurality of health assessment indexes to obtain a plurality of regional health scores.

[0092] A weighted calculation unit is configured to assign weights to the multiple forestry sub-regions to obtain multiple weight coefficients, and perform weighted calculation on the multiple regional health scores according to the multiple weight coefficients to generate the health score.

[0093] Further, the resource distribution map determination module 40 further includes: A change trajectory parameter drawing unit is configured to continuously track the forestry resource change trend map according to the data acquisition time interval information, and draw the change trajectory parameters of the forestry resource change trend map.

[0094] An image-resource pair generation unit is configured to synchronize the multiple remote sensing image data to the forestry resource change trend map according to the change trajectory parameters for correlation analysis, and generate multiple image-resource pairs.

[0095] An image resource change label generation unit is configured to traverse the target forestry area to match and identify the multiple image-resource pairs, and generate multiple image resource change labels.

[0096] A forestry resource change trend map mapping unit is configured to map the multiple remote sensing image data and the forestry resource change trend map to the geographical space of the target forestry area based on the multiple image resource change labels, and construct the resource distribution map.

[0097] Further, the forestry resource change trend map mapping unit further includes: A geographical space construction unit is configured to perform three-dimensional analysis on the target forestry area based on the multiple remote sensing image data, and construct a geographical space.

[0098] A spatial pixel feature extraction unit is configured to extract multiple spatial pixel features of the multiple remote sensing image data according to the multiple image resource change labels.

[0099] A spatial change feature extraction unit is configured to extract multiple spatial change features of the forestry resource change trend map according to the multiple image resource change labels.

[0100] A resource distribution map construction unit is configured to map the multiple remote sensing image data to the geographical space based on the multiple spatial pixel features, and map the forestry resource change trend map to the geographical space based on the multiple spatial change features, and construct the resource distribution map of the target forestry area.

[0101] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. In addition, the specific embodiments of this specification have been described. Further, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0102] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0103] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A remote sensing analysis method for forestry resources monitored by unmanned aerial vehicles, characterized in that: The method comprises: Launching a drone to collect data from a target forestry area to obtain a multi-source data set of the target forestry area, and performing data screening based on the multi-source data set to obtain a plurality of remote sensing image data; Extracting features based on the multiple remote sensing image data, performing cluster analysis on the multiple remote sensing image data according to the multiple image features, and determining multiple forestry resource types; Using the multiple image features to perform health assessment on the target forestry area according to the multiple forestry resource types, generating a health score, predicting the dynamic changes of forestry resources based on the health score, and constructing a forestry resource change trend map; Mapping the plurality of remote sensing image data to the forestry resource change trend map to identify the target forestry area and determine a resource distribution map; The resource distribution map is traversed to perform management analysis, generate management and protection decisions, and transmit the management and protection decisions to drones to perform intelligent management and monitoring of the target forestry area.

2. The forestry resources remote sensing analysis method monitored by an unmanned aerial vehicle according to claim 1, characterized in that: Feature extraction is performed based on the multiple remote sensing image data, and the method includes: Performing data registration on the plurality of remote sensing image data through geometric transformation to obtain a plurality of image registration data; Dividing the plurality of image registration data according to spectral bands to determine a plurality of spectral band image data; A convolutional neural network is used to construct a convolution kernel sliding window, and the convolution kernel sliding window is used to traverse the image data of the multiple spectral bands to perform feature analysis to obtain multiple local features of the image; Pooling dimensionality reduction is performed based on the multiple image local features to determine the multiple image features.

3. The forestry resources remote sensing analysis method monitored by an unmanned aerial vehicle as claimed in claim 2, characterized in that: The convolutional neural network is used to construct a convolution kernel sliding window. The method includes: Perform pixel analysis according to the plurality of spectral band image data to generate pixel parameters; Determining whether the pixel parameter is greater than or equal to a preset pixel threshold, and when the pixel parameter is greater than or equal to the preset pixel threshold, generating a cropping instruction, and performing image segmentation on the plurality of spectral band image data through the cropping instruction to determine a plurality of image blocks; A convolution kernel filter is activated according to the convolution neural network, and the convolution kernel filter sets the convolution kernel sliding window according to the multiple image blocks.

4. The forestry resources remote sensing analysis method monitored by an unmanned aerial vehicle according to claim 1, characterized in that: Cluster analysis is performed on the plurality of remote sensing image data according to a plurality of image features to determine a plurality of forestry resource types, the method comprising: Performing multi-dimensional correction on the multiple remote sensing image data to generate image correction results, randomly selecting the image correction results according to the multiple image features, and determining K initial clustering centers; Traversing the plurality of remote sensing image data according to the K initial cluster centers to perform data distribution and determine a plurality of cluster numbers, where K is a positive integer greater than 0; Retrieving historical forestry record logs of the target forestry area, recalculating the plurality of cluster numbers based on the historical forestry record logs, updating the K initial cluster centers, and determining P cluster centers, where P is a positive integer greater than 0; The plurality of remote sensing image data are traversed according to the P cluster centers to perform forestry classification and determine the plurality of forestry resource types.

5. The forestry resources remote sensing analysis method monitored by an unmanned aerial vehicle as claimed in claim 4, characterized in that: Using the multiple image features to perform a health assessment on the target forestry area according to the multiple forestry resource types to generate a health score, the method comprising: Perform instance segmentation on the target forestry area according to the multiple forestry resource types to determine multiple forestry sub-areas; Combining the plurality of image features according to the plurality of forestry sub-regions to construct an image feature vector; Introducing data collection time interval information and combining it with the historical forestry record log to perform health analysis on the image feature vector and construct multiple health assessment indicators; Performing health assessment on the plurality of forestry sub-regions based on the plurality of health assessment indicators to obtain a plurality of regional health scores; Weights are assigned to the multiple forestry sub-regions to obtain multiple weight coefficients, and weighted calculations are performed on the health scores of the multiple regions according to the multiple weight coefficients to generate the health scores.

6. The forestry resources remote sensing analysis method monitored by an unmanned aerial vehicle as claimed in claim 5, characterized in that: Mapping the plurality of remote sensing image data to the forestry resource change trend map to identify the target forestry area and determine the resource distribution map, the method comprising: Continuously tracking the forestry resources change trend graph according to the data collection time interval information, and drawing the change trajectory parameters of the forestry resources change trend graph; Synchronizing the plurality of remote sensing image data to the forestry resource change trend map according to the change trajectory parameters for correlation analysis to generate a plurality of image-resource pairs; Traversing the target forestry area to match and identify the multiple image-resource pairs, and generating multiple image resource change labels; Based on the multiple image resource change tags, the multiple remote sensing image data and the forestry resource change trend map are mapped to the geographic space of the target forestry area to construct the resource distribution map.

7. The forestry resources remote sensing analysis method monitored by an unmanned aerial vehicle according to claim 6, characterized in that: Based on the multiple image resource change labels, the multiple remote sensing image data and the forestry resource change trend map are mapped to the geographic space of the target forestry area to construct the resource distribution map, the method comprising: Performing three-dimensional analysis on the target forestry area based on the multiple remote sensing image data to construct a geographic space; Extracting a plurality of spatial pixel features of the plurality of remote sensing image data according to the plurality of image resource change labels; Extracting multiple spatial change features of the forestry resource change trend map according to the multiple image resource change tags; Based on the multiple spatial pixel features, the multiple remote sensing image data are mapped to the geographic space, and based on the multiple spatial change characteristics, the forestry resource change trend map is mapped to the geographic space to construct the resource distribution map of the target forestry area.

8. The forestry resources remote sensing analysis system monitored by unmanned aerial vehicles is characterized by: The system is used to implement the remote sensing analysis method for forestry resources monitored by an unmanned aerial vehicle according to any one of claims 1 to 7, and the system comprises: A remote sensing image data acquisition module is used to start the UAV to collect data from the target forestry area, obtain a multi-source data set of the target forestry area, perform data screening based on the multi-source data set, and obtain multiple remote sensing image data; A forestry resource type determination module, configured to extract features based on the plurality of remote sensing image data, perform cluster analysis on the plurality of remote sensing image data according to the plurality of image features, and determine a plurality of forestry resource types; A forestry resource change trend map construction module is used to perform a health assessment on a target forestry area according to the multiple forestry resource types using the multiple image features, generate a health score, predict the dynamic changes of forestry resources based on the health score, and construct a forestry resource change trend map; A resource distribution map determination module is used to map the plurality of remote sensing image data to the forestry resource change trend map to identify the target forestry area and determine the resource distribution map; The intelligent management and monitoring module is used to traverse the resource distribution map to perform management analysis, generate management and protection decisions, and transmit the management and protection decisions to the drone to perform intelligent management and monitoring of the target forestry area.

Images