Method and platform for dynamically monitoring cultivated land loss by fusing remote sensing and AI analysis

Abstract

The invention provides a remote sensing and AI analysis-fused farmland loss dynamic monitoring method and platform, and relates to the technical field of farmland monitoring, and the method comprises the steps: carrying out the coverage traversal of a target region according to a data collection frequency through a remote sensing technology, obtaining a remote sensing data set, carrying out the spatial analysis after semantic segmentation, determining a plurality of spatial features, and carrying out the prediction of farmland loss. Performing time sequence analysis on the target area to capture area space time sequence characteristics, performing time-space change analysis to generate a time-space change trend, performing correlation analysis on the cultivated land loss of the target area and determining a correlation coefficient, performing identification and classification in combination with the time-space change trend, and drawing a cultivated land loss space distribution map. And mapping the remote sensing data set to the cultivated land loss spatial distribution diagram to monitor cultivated land loss, and making a cultivated land loss management scheme, thereby solving the problems of inaccurate loss risk assessment and lack of effective management decision support caused by incapability of realizing real-time dynamic monitoring of cultivated land loss in cultivated land loss monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of cultivated land monitoring, and particularly to a dynamic monitoring method and platform for cultivated land loss integrating remote sensing and AI analysis. Background Art

[0002] Cultivated land loss is a major environmental problem globally. Due to factors such as the acceleration of urbanization, frequent agricultural activities, and natural disasters, the area of cultivated land has been decreasing year by year. To effectively prevent and control cultivated land loss, relevant departments have been committed to carrying out monitoring work on cultivated land loss. Traditional cultivated land loss monitoring methods mostly rely on manual inspections, ground surveys, and regular sampling data, and usually conduct data analysis based on manual interpretation of remote sensing images. Remote sensing technology, especially satellite remote sensing and aerial remote sensing, as an effective means of obtaining large-scale and long-term change information, has been widely used in fields such as land use, crop growth, and environmental monitoring. Currently, the application of remote sensing technology mainly focuses on the static acquisition of ground information and conducts spatial analysis by combining ground samples or historical data. Although it can provide wide-area remote sensing image data, its deficiencies are becoming increasingly apparent. For example, in the prior art, remote sensing data analysis usually lacks comprehensive processing of the time dimension, and most monitoring methods only rely on image data at a single moment, making it difficult to capture the dynamic changes during the process of cultivated land loss. With the increasing complexity and expansion of the spatial scale of cultivated land loss, traditional remote sensing monitoring methods are no longer able to meet the requirements of accurate, real-time, and dynamic monitoring of cultivated land loss. In addition, although modern computer vision and deep learning technologies have made important progress in the automatic recognition and processing of remote sensing images, in the field of cultivated land loss monitoring, these technologies mostly stay at the level of image classification in a single space, lacking comprehensive capture of spatio-temporal features and effective analysis of the spatio-temporal dynamic changes of cultivated land loss. In summary, existing cultivated land loss monitoring often has technical problems such as the inability to achieve real-time dynamic monitoring of cultivated land loss, resulting in inaccurate loss risk assessment and lack of effective management decision support. Summary of the Invention

[0003] This application provides a dynamic monitoring method and platform for cultivated land loss integrating remote sensing and AI analysis, aiming to solve the technical problems existing in existing cultivated land loss monitoring, such as the inability to achieve real-time dynamic monitoring of cultivated land loss, resulting in inaccurate loss risk assessment and lack of effective management decision support.

[0004] In view of the above problems, this application provides a dynamic monitoring method and platform for cultivated land loss integrating remote sensing and AI analysis.

[0005] In a first aspect, the present application provides a dynamic monitoring method for cultivated land loss by integrating remote sensing and AI analysis. The method includes: covering and traversing a target area according to a data acquisition frequency through remote sensing technology to obtain a remote sensing data set; performing semantic segmentation based on the remote sensing data set to obtain a plurality of segmentation data, performing spatial analysis on the plurality of segmentation data to determine a plurality of spatial features; performing temporal analysis on the target area according to the plurality of spatial features to capture regional spatial temporal features, performing spatio-temporal change analysis on the regional spatial temporal features and the plurality of spatial features to generate a spatio-temporal change trend; performing correlation analysis on the cultivated land loss in the target area according to the spatio-temporal change trend to determine a spatio-temporal correlation coefficient, performing identification and classification based on the spatio-temporal correlation coefficient combined with the spatio-temporal change trend, and drawing a spatial distribution map of cultivated land loss according to the classification result; mapping the remote sensing data set to the spatial distribution map of cultivated land loss for monitoring cultivated land loss, generating a monitoring report, and performing dynamic analysis of cultivated land loss according to the monitoring report to formulate a management plan for cultivated land loss.

[0006] In a second aspect, the present application provides a dynamic monitoring platform for cultivated land loss by integrating remote sensing and AI analysis. The platform includes: a data acquisition module for covering and traversing a target area according to a data acquisition frequency through remote sensing technology to obtain a remote sensing data set; a semantic segmentation module for performing semantic segmentation based on the remote sensing data set to obtain a plurality of segmentation data, performing spatial analysis on the plurality of segmentation data to determine a plurality of spatial features; a temporal analysis module for performing temporal analysis on the target area according to the plurality of spatial features to capture regional spatial temporal features, performing spatio-temporal change analysis on the regional spatial temporal features and the plurality of spatial features to generate a spatio-temporal change trend; a correlation analysis module for performing correlation analysis on the cultivated land loss in the target area according to the spatio-temporal change trend to determine a spatio-temporal correlation coefficient, performing identification and classification based on the spatio-temporal correlation coefficient combined with the spatio-temporal change trend, and drawing a spatial distribution map of cultivated land loss according to the classification result; a dynamic analysis module for mapping the remote sensing data set to the spatial distribution map of cultivated land loss for monitoring cultivated land loss, generating a monitoring report, and performing dynamic analysis of cultivated land loss according to the monitoring report to formulate a management plan for cultivated land loss.

[0007] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0008] The dynamic monitoring method for cultivated land loss that integrates remote sensing and AI analysis provided by this application traverses the target area according to the data acquisition frequency through remote sensing technology to obtain a remote sensing data set; performs semantic segmentation based on the remote sensing data set to obtain multiple segmentation data, conducts spatial analysis on the multiple segmentation data to determine multiple spatial features; performs temporal analysis on the target area according to the multiple spatial features to capture the regional spatial temporal features, conducts spatio-temporal change analysis on the regional spatial temporal features and the multiple spatial features to generate a spatio-temporal change trend; conducts correlation analysis on the cultivated land loss in the target area according to the spatio-temporal change trend to determine the spatio-temporal correlation coefficient, performs identification and classification based on the spatio-temporal correlation coefficient combined with the spatio-temporal change trend, and draws a spatial distribution map of cultivated land loss according to the classification result; maps the remote sensing data set to the spatial distribution map of cultivated land loss for monitoring cultivated land loss, generates a monitoring report, conducts dynamic analysis of cultivated land loss according to the monitoring report, and formulates a management plan for cultivated land loss, solving the technical problems existing in the existing cultivated land loss monitoring, such as the inability to achieve real-time dynamic monitoring of cultivated land loss, resulting in inaccurate loss risk assessment and lack of effective management decision support. By integrating remote sensing and AI analysis, it realizes the efficient and real-time dynamic monitoring of cultivated land loss, accurately captures the spatio-temporal change trend, provides accurate loss area classification and risk assessment, thereby supporting the formulation of a scientific management plan for cultivated land loss. Description of the Drawings

[0009] Figure 1 FIG. is a schematic flow chart of the dynamic monitoring method for cultivated land loss that integrates remote sensing and AI analysis provided by this application.

[0010] Figure 2 FIG. is a schematic structural diagram of the dynamic monitoring platform for cultivated land loss that integrates remote sensing and AI analysis provided by this application.

[0011] Description of the reference numerals: data acquisition module 11, semantic segmentation module 12, temporal analysis module 13, correlation analysis module 14, dynamic analysis module 15. Detailed Embodiments

[0012] In order to make the objectives, technical solutions and advantages of this application clearer, the following further elaborates on this application in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0013] Example 1, as Figure 1 shown, the dynamic monitoring method for cultivated land loss that integrates remote sensing and AI analysis provided by this application includes:

[0014] Traverse the target area according to the data acquisition frequency through remote sensing technology to obtain a remote sensing data set.

[0015] Specifically, the target area is comprehensively covered and traversed by remote sensing technology to ensure that data can be collected from each geographical unit within the target area. The target area is the area where cultivated land loss monitoring is to be carried out. Remote sensing technology mainly collects information about the earth's surface through sensors carried by satellites, aerial platforms or drones. These sensors can collect multi-dimensional ground information to form a remote sensing data set, including but not limited to data such as surface type, vegetation index, soil moisture, temperature, etc. Among them, the surface type refers to the category of ground objects divided according to different spectral reflection characteristics in remote sensing images, such as cultivated land, grassland, water area, urban area, etc.; by identifying different surface types, the land use status of the target area can be comprehensively understood. The vegetation index provided by remote sensing technology (such as NDVI, Normalized Difference Vegetation Index) can reflect the coverage and growth status of ground vegetation and is often used to evaluate the growth health of cultivated land and crops. In addition, soil moisture refers to the water content in the soil. Remote sensing technology can measure the change in reflectivity through specific sensors, thereby inferring the moisture level of the soil. The temperature is obtained through thermal infrared imaging technology and can be used to identify the heat distribution of the land, and then evaluate the temperature change and potential evaporation loss in the cultivated land area. By controlling the data collection frequency, the timeliness and coverage of remote sensing data can be ensured. For example, the collection frequency can be set according to seasonal changes or specific climate conditions to regularly obtain data to capture the dynamic changes of cultivated land loss. Usually, the data collection frequency can be set according to actual needs and the characteristics of the target area. For example, during the critical growth period of certain crops, the remote sensing data collection frequency can be increased to ensure real-time monitoring. Through the above methods, the collected remote sensing data set includes multiple important geographical feature information, providing basic data support for subsequent spatial analysis and temporal analysis. The information in the data set will provide a reliable data source and basis for the subsequent monitoring, analysis and prediction of cultivated land loss.

[0016] Based on the remote sensing data set, semantic segmentation is performed to obtain multiple segmentation data, and spatial analysis is performed on the multiple segmentation data to determine multiple spatial features.

[0017] Optionally, each image data in the remote sensing dataset will undergo semantic segmentation processing. Semantic segmentation refers to classifying each pixel in a remote sensing image into different categories or labels, and classifying them through the pixel values of the image to identify different types of ground object information. In this step, semantic segmentation is not just simply cutting the image into different regions, but rather classifying the objects in the image with high precision through deep learning algorithms (such as convolutional neural network CNN). These classification results include, but are not limited to, various ground object types such as cultivated land, water area, forest, urban buildings, etc., which lays the foundation for subsequent spatial analysis. After semantic segmentation, multiple segmentation data are obtained. These segmentation data are based on different ground object categories of the remote sensing image, and each category is marked as an independent region. These regions can be represented by the spatial positions and labels of the image pixels, and then multiple regions with different spatial characteristics are obtained. For example, due to different spectral reflection characteristics, the cultivated land area data and urban area data formed after segmentation have different characteristic labels. Next, spatial analysis is performed on the multiple segmentation data. Spatial analysis refers to analyzing the relative positions and spatial distribution in the target area according to the spatial positions, shapes, and distribution characteristics of different segmentation regions. This analysis usually uses spatial statistical methods, such as using spatial autocorrelation analysis to determine the spatial relationships between different ground object categories. Through spatial analysis, spatial characteristics such as the distances, boundary shapes, areas, distribution densities, etc. between different categories can be determined. For example, in cultivated land loss monitoring, special attention needs to be paid to the boundary changes between the cultivated land area and the urban expansion area, or the distance distribution between cultivated land and water area, which can all be obtained through spatial analysis. The multiple spatial characteristics obtained through spatial analysis refer to the characteristics of each ground object category in the target area in the spatial dimension. Common spatial characteristics include area, shape complexity, boundary length, distribution density, connectivity, etc. For example, the area and distribution density of the cultivated land area can reflect the spatial proportion and loss situation of cultivated land; while the connectivity between cultivated land and urban areas can reflect the impact of urban expansion on cultivated land loss. Through the extraction of these spatial characteristics, the spatial distribution and change laws of different ground object categories can be understood more accurately, providing a reliable basis for subsequent spatio-temporal analysis and cultivated land loss prediction.

[0018] Perform temporal analysis on the target area according to the multiple spatial characteristics, capture the regional spatial temporal characteristics, and perform spatio-temporal change analysis on the regional spatial temporal characteristics and the multiple spatial characteristics to generate a spatio-temporal change trend.

[0019] Exemplarily, temporal analysis is performed on the target area using the obtained multiple spatial features. Temporal analysis refers to analyzing the change trends of various ground features (such as cultivated land, water areas, urban construction, etc.) within the target area over time based on the spatial feature data at different time nodes through the changes in the time dimension. During this process, the remote sensing data sets corresponding to each time node form corresponding spatial feature data sets through spatial feature extraction, and a clear corresponding relationship is formed between these data sets and time. For example, the remote sensing data of each year will be processed as a time node, corresponding to the spatial features of different ground feature types within the recorded area. In the process of temporal analysis, capturing the regional spatial temporal features is a key step. Regional spatial temporal features refer to the change trends and spatio-temporal patterns of different ground feature categories within the target area at multiple time nodes. For example, the expansion or contraction of cultivated land areas, the increase or decrease in water area, the acceleration of the urbanization process, etc. These temporal features help identify the change rules and amplitudes of different ground feature categories within the target area, thus providing dynamic change benchmark data for subsequent analysis. By calculating the change amounts between different time nodes, the long-term change trends and short-term fluctuation characteristics existing within the area can be clearly identified. Next, spatio-temporal change analysis is performed on the regional spatial temporal features and the multiple spatial features. Spatio-temporal change analysis refers to combining spatial features with time changes for multi-dimensional comprehensive analysis in order to reveal the spatio-temporal characteristics of cultivated land loss or other land use changes. In spatio-temporal change analysis, the rules of change of different ground feature categories in time and space are mainly analyzed by fusing and comparing the data in the time dimension and the spatial dimension. For example, analyze the expansion speed of cultivated land areas in different periods and combine it with their spatial distribution characteristics to evaluate the degree and trend of cultivated land loss in this area. In addition, through correlation analysis with external factors such as climate and urban expansion, the impacts of different factors on cultivated land loss can also be revealed. Through spatio-temporal change analysis, a spatio-temporal change trend can finally be generated. The spatio-temporal change trend refers to the change rules and development trends of different ground feature types within the target area within a certain time range, and this trend reveals the future development directions and potential risks of phenomena such as cultivated land loss and land use changes. For example, the spatio-temporal change trend can help predict the speed of cultivated land loss in a certain area in the next few years, or the impact of the urbanization process on the surrounding cultivated land. The generated spatio-temporal change trend provides a scientific basis for subsequent decision-making support, early warning of cultivated land loss, and risk management.

[0020] Perform correlation analysis on the cultivated land loss in the target area according to the spatio-temporal change trend, determine the spatio-temporal correlation coefficient, perform identification and classification based on the spatio-temporal correlation coefficient combined with the spatio-temporal change trend, and draw a spatial distribution map of cultivated land loss according to the classification results.

[0021] Further, first, perform a correlation analysis on the cultivated land loss in the target area based on the obtained spatio-temporal change trend. Correlation analysis refers to studying the relationship between the spatio-temporal change trend and cultivated land loss through statistical methods, and exploring the correlation between cultivated land loss and spatio-temporal characteristics. Specifically, analyze the loss change of the cultivated land area at different time nodes, and use the spatio-temporal change trend to reveal the driving factors of its loss, such as urban expansion, climate change, etc. Through this analysis, the close connection between cultivated land loss and time change and spatial distribution can be identified. Next, the calculation of the spatio-temporal correlation coefficient is the core step in the correlation analysis. The spatio-temporal correlation coefficient is a numerical indicator used to quantify the relationship between spatio-temporal characteristics and cultivated land loss, usually calculated through regression analysis or correlation analysis. This coefficient can reflect the degree of correlation between the cultivated land loss area and spatio-temporal characteristics. For example, if the cultivated land loss in a certain area is highly correlated with the acceleration of the urbanization process, the spatio-temporal correlation coefficient of this area will be relatively high. By analyzing the cultivated land loss data of different spatial units and time periods in the target area, the spatio-temporal correlation coefficients of each area are generated. Then, based on the spatio-temporal correlation coefficient, combine it with the spatio-temporal change trend for identification and classification. This step divides the cultivated land loss situations in different areas into different categories, such as high-loss areas, low-loss areas, fluctuating areas, etc., by setting thresholds for the spatio-temporal correlation coefficient. Through identification and classification, different types and risk levels of cultivated land loss in the target area can be systematically distinguished. For example, some areas with a relatively high rate of cultivated land loss and a close relationship with the spatio-temporal change trend may be classified as "high-loss areas"; while some stable areas may be classified as "low-loss areas". This classification helps decision-makers clearly identify the severity and potential risks of cultivated land loss. Finally, draw a spatial distribution map of cultivated land loss according to the results of the identification and classification. The spatial distribution map of cultivated land loss is to mark different types of cultivated land loss areas on the map based on the aforementioned spatio-temporal correlation coefficient and classification results, and represent the loss degree of each type of area through different colors or symbols. This map can visually show the spatial distribution of cultivated land loss in the target area, providing a visual decision-making basis for policymakers. For example, through different color markings, high-risk loss areas, medium-risk loss areas, and low-risk loss areas can be clearly shown, facilitating the precise management and intervention of cultivated land loss. Generating the spatial distribution map of cultivated land loss provides a systematic analysis tool and visual display for the precise identification, risk assessment, and management decision-making of cultivated land loss.

[0022] Map the remote sensing data set to the spatial distribution map of cultivated land loss for monitoring cultivated land loss, generate a monitoring report, and conduct dynamic analysis of cultivated land loss based on the monitoring report to formulate a cultivated land loss management plan.

[0023] Specifically, mapping the remote sensing data set to the spatial distribution map of cultivated land loss for the monitoring of cultivated land loss. This process generates the spatial distribution map of cultivated land loss by combining the information of different land cover types (such as cultivated land, urban construction areas, water areas, etc.) in the remote sensing data set with the spatio-temporal characteristics of cultivated land loss. In this process, each land cover type (such as cultivated land area, urban expansion, vegetation index, etc.) in the remote sensing data set will be matched with the loss data through spatial registration technology to ensure the consistency and accuracy of the data, thereby generating a visual and accurate spatial distribution map reflecting the situation of cultivated land loss. In this way, the spatial distribution map of cultivated land loss can clearly show the areas of cultivated land loss, the degree of loss, and the trend of loss, providing basic data for subsequent dynamic monitoring and analysis. This map shows the situation of cultivated land loss in different regions and identifies areas with different degrees of loss (such as high-loss areas, medium-loss areas, low-loss areas) through colors, symbols, etc. For example, high-loss areas may be represented by red, indicating that the cultivated land in this area has a fast loss rate and a large area; while low-loss areas are represented by green, indicating that the cultivated land loss in this area is less or more stable. Next, based on the spatial distribution map of cultivated land loss, a monitoring report is generated. The monitoring report includes a comprehensive analysis of the cultivated land loss areas, specifically including key information such as risk areas, change speed, and degree of loss. Risk areas refer to high-risk areas determined according to the severity and occurrence probability of cultivated land loss. The cultivated land in these areas has a relatively fast loss rate and may be strongly affected by factors such as urbanization and climate change. The change speed refers to the temporal change rate of cultivated land loss, that is, the increase or decrease speed of the cultivated land loss area per unit time. For example, the cultivated land loss rate in some areas is relatively fast, which may be caused by excessive agricultural activities, land salinization, etc. The degree of loss measures the severity of cultivated land loss, usually evaluated by the reduction amount of cultivated land area, the deterioration degree of soil quality, and the changes in the ecological environment. These data will be detailed in the monitoring report, providing a quantitative basis for subsequent dynamic analysis and management plans. Based on the information in the monitoring report, dynamic analysis of cultivated land loss is carried out. Dynamic analysis refers to predicting the possible situation of cultivated land loss in the future based on historical data and current monitoring results, combined with the change trend of cultivated land loss. By comprehensively analyzing the trend, change speed, risk areas, etc. of cultivated land loss, potential high-risk loss areas and future hotspots of cultivated land loss can be revealed. The purpose of dynamic analysis is to predict the situation of cultivated land loss in the next few years in advance and help relevant departments prepare early warnings and response measures. Finally, a cultivated land loss management plan is formulated according to the results of the dynamic analysis of cultivated land loss. The cultivated land loss management plan is a scientific management measure proposed based on a comprehensive assessment of the current situation and future trend of cultivated land loss. According to the information in the monitoring report, targeted management measures can be formulated for different risk areas and areas with different degrees of loss.For example, for areas with a high risk of soil loss, measures such as land protection and vegetation restoration may be required; for areas with a relatively fast rate of soil loss, more stringent agricultural policies may need to be implemented to restrict behaviors such as over-tilling. In addition, the management plan can also formulate corresponding compensation policies or ecological restoration measures according to the actual conditions of different regions to ensure the sustainable use of cultivated land resources.

[0024] Furthermore, performing temporal analysis on the target area according to the multiple spatial features to capture the regional spatial temporal features includes: arranging the remote sensing data set according to the data acquisition time sequence to generate a remote sensing time sequence, where the remote sensing time sequence has multiple time nodes, and the multiple time nodes have a corresponding relationship with the remote sensing data set; associating and matching the multiple spatial features with the remote sensing data set according to the multiple time nodes in the remote sensing time sequence to construct a spatial temporal data set; and dynamically capturing the target area based on the spatial temporal data set to obtain the regional spatial temporal features.

[0025] In a specific embodiment, the remote sensing data set is arranged according to the data acquisition time sequence. The data acquisition time sequence refers to sorting the remote sensing image data at different time points in the order of data acquisition time, ensuring that the remote sensing data at each moment can be recorded and used at the correct time node. For example, the remote sensing data may be acquired annually, seasonally, or monthly, and each acquisition moment corresponds to specific ground object change information. Therefore, the generated remote sensing time sequence will be a data set of remote sensing images arranged in chronological order, containing multiple time nodes, and these time nodes correspond to the remote sensing image data acquired at different time points. Then, a clear corresponding relationship is established between each time node included in the remote sensing time sequence and the image data in the corresponding remote sensing data set. The corresponding relationship between each time node and the image in the data set makes the subsequent spatio-temporal analysis systematic and consistent, facilitating the dynamic tracking and analysis of ground object changes. Then, the multiple spatial features are associated and matched with the remote sensing data set according to the multiple time nodes in the remote sensing time sequence to construct a spatio-temporal data set. In this step, the spatial features refer to the spatial attributes of different ground object types extracted from the remote sensing images, such as cultivated land area, vegetation coverage, ground object boundary changes, etc. The change process of these spatial features at different time nodes will be matched with the remote sensing data at each time node, thus forming a multi-dimensional spatio-temporal data set. Specifically, the spatial features (such as cultivated land area, boundary changes, vegetation index, etc.) included in the remote sensing data at each time node will be matched with the remote sensing image at that time node and combined with the spatial feature data at other time nodes. In this way, the data in the time dimension and the space dimension are effectively integrated to form a time sequence data set containing multiple time nodes and spatial features. Next, based on the spatio-temporal data set, dynamic capture is performed on the target area to obtain the spatio-temporal features of the area. Dynamic capture refers to extracting the dynamic features of the spatial features changing with time in the target area through the analysis of the spatio-temporal data set. For example, the change in cultivated land area, the seasonal fluctuation of vegetation coverage, the speed of urban expansion, etc. These change features can not only reveal the spatial distribution of ground objects at a certain moment but also capture the change trends between different time points, helping to analyze the dynamic evolution process of the target area. For example, the change in cultivated land area can reflect the change in land use, and the change in vegetation index can show the impact of agricultural production activities or natural disasters on land cover. By dynamically capturing these change trends, the change features of the target area can be accurately identified, and then scientific predictions can be made for phenomena such as cultivated land loss and urban expansion. In summary, by arranging the remote sensing data set in chronological order, a remote sensing time sequence with multiple time nodes is generated, and the spatial features at each time node are associated and matched with the remote sensing data to construct a spatio-temporal data set.Then, based on this dataset, dynamic capture of the target area is carried out to obtain the dynamic information of the spatial features within the area changing over time. Through this process, the spatio-temporal change characteristics of the target area can be comprehensively understood, providing important data support for subsequent cultivated land loss analysis, change prediction, and dynamic monitoring.

[0026] Furthermore, spatio-temporal change analysis is performed on the regional spatial temporal features and the multiple spatial features to generate spatio-temporal change trends, including: fusing the regional spatial temporal features and the multiple spatial features according to the time dimension to generate a time feature dataset; fusing the regional spatial temporal features and the multiple spatial features according to the spatial dimension to generate a spatial feature dataset; performing correlation analysis based on the time feature dataset and the spatial feature dataset to generate multiple correlation coefficients; performing multi-scale spatio-temporal analysis on the regional spatial temporal features and the multiple spatial features according to the multiple correlation coefficients to obtain multiple spatio-temporal change data; performing clustering analysis and identification based on the multiple spatio-temporal change data to determine multiple types of spatio-temporal change data, and constructing the spatio-temporal change trend according to the multiple types of spatio-temporal change data.

[0027] Specifically, the regional spatial temporal features and multiple spatial features will be fused according to the time dimension to generate a time feature dataset. The fusion in the time dimension means combining the spatial features of the target region at different time nodes with the corresponding time information to form a feature dataset that comprehensively reflects the temporal changes. Specifically, the time feature dataset will contain the spatial change information of various ground features (such as cultivated land, vegetation, urban expansion, etc.) at each time node of the target region. For example, the area change of cultivated land, the fluctuation of vegetation index, etc. will be reflected in this dataset, providing crucial support in the time dimension for spatio-temporal change analysis. Next, the regional spatial temporal features and multiple spatial features will be fused according to the spatial dimension to generate a spatial feature dataset. The fusion in the spatial dimension refers to matching the spatial features of the target region at each time node, such as the distribution of ground features, boundary changes, coverage, etc., with the spatial position at each moment to generate a dataset covering spatial distribution information. Specifically, the spatial feature dataset will contain the feature information at each time point on different spatial units (such as different grid cells, regions, etc.) of the target region, such as the boundary changes of cultivated land, the expansion of urban construction areas, etc. This process provides the basic data in the spatial dimension for subsequent spatio-temporal change analysis. After these two feature datasets (time feature dataset and spatial feature dataset) are generated, a correlation analysis is carried out. The purpose of the correlation analysis is to reveal the relationship between the time features and the spatial features, generating multiple correlation coefficients, which reflect the mutual relationship between the temporal changes and the spatial distribution. Specifically, the correlation coefficients can be calculated by statistical methods such as Pearson correlation coefficient or regression analysis to measure the closeness of the changes in spatial features such as cultivated land and vegetation at different time nodes to the temporal changes. For example, if the loss of cultivated land in a certain region is closely related to the urbanization process, then the correlation coefficient between the time features and the spatial features of this region will be very high. Next, based on multiple correlation coefficients, multi-scale spatio-temporal analysis is performed on the regional spatial temporal features and multiple spatial features to obtain multiple spatio-temporal change data. Multi-scale spatio-temporal analysis is to process remote sensing data at different resolutions by using multi-scale analysis methods, such as multi-resolution analysis or Scale-Invariant Feature Transform (SIFT), etc., so as to reveal the different scale features of ground feature changes in the region. Multi-scale analysis can consider the changes at different spatial and temporal scales simultaneously and identify the loss features at different scales. For example, at a larger scale (such as regional level), the pattern of cultivated land loss may show a slow and steady trend; while at a smaller scale (such as grid level), it may show local rapid loss. Through this analysis, change features such as cultivated land loss and urban expansion at different scales can be obtained. After obtaining multiple spatio-temporal change data, cluster analysis and identification are carried out to determine multiple types of spatio-temporal change data. Cluster analysis groups the spatio-temporal data according to similarity to identify different types of spatio-temporal change patterns.Each type of spatio-temporal change data represents the change characteristics of a specific region within a specific time period. For example, different types of regions such as "rapid cultivated land loss areas", "stable areas", and "rapid urbanization areas" can be identified. These different categories of spatio-temporal data will provide precise regional division for the management of cultivated land loss. Finally, based on the described multiple types of spatio-temporal change data, a spatio-temporal change trend is constructed. The spatio-temporal change trend is a comprehensive summary of the change laws of the target region in the time and space dimensions, revealing the change trends of various land cover types (such as cultivated land, water areas, cities, etc.). For example, the cultivated land loss may intensify in some regions in recent years, while other regions may remain stable or recover. By constructing the spatio-temporal change trend, it is possible to clarify the possible changes in different regions in the future, providing a scientific basis for the prediction and management of cultivated land loss.

[0028] Furthermore, perform a correlation analysis on the cultivated land loss in the target region according to the spatio-temporal change trend to determine the spatio-temporal correlation coefficient. Based on the spatio-temporal correlation coefficient and combined with the spatio-temporal change trend, perform identification and classification. According to the classification results, draw a spatial distribution map of cultivated land loss, including: perform a spatial autocorrelation analysis based on the spatio-temporal change trend to obtain spatial aggregation related parameters; perform a spatio-temporal regression analysis based on the spatio-temporal change trend and combined with the spatial aggregation related parameters to generate the spatio-temporal correlation coefficient; perform a cluster analysis and identification on the target region according to the spatio-temporal correlation coefficient to obtain multiple types of regions; divide the cultivated land loss level of the target region by combining the multiple types of regions with the historical cultivated land loss record data of the target region to determine the high-loss region information, low-loss region information, and fluctuating region information; divide the cultivated land loss risk of the target region according to the spatio-temporal change trend and combined with the historical cultivated land loss record data of the target region to determine the high-risk loss region information, medium-risk loss region information, and low-risk loss region information; match the high-loss region information, the low-loss region information, the fluctuating region information with the high-risk loss region information, the medium-risk loss region information, and the low-risk loss region information according to the spatio-temporal correlation coefficient, perform hierarchical classification according to the matching results to generate multiple classification layers; perform an associated integration on the multiple classification layers to construct the spatial distribution map of cultivated land loss.

[0029] Optionally, spatial autocorrelation analysis is performed based on the spatio-temporal change trend to obtain spatial aggregation-related parameters. Spatial autocorrelation analysis is used to measure the spatial relationship between the cultivated land loss areas and the surrounding areas within the target region. By calculating the similarity between regions, it reveals the spatial distribution pattern of cultivated land loss. Specifically, the spatial aggregation-related parameters reflect the degree of spatial aggregation of cultivated land loss. If the cultivated land loss in a certain region is highly correlated with the loss in neighboring regions, the spatial aggregation-related parameters of this region are relatively high, indicating a strong loss correlation in this region and its surrounding areas. Then, spatio-temporal regression analysis is performed based on the spatio-temporal change trend and the spatial aggregation-related parameters to generate spatio-temporal correlation coefficients. Spatio-temporal regression analysis analyzes the correlation of cultivated land loss areas at different time and spatial scales through a regression model, revealing the relationship between the temporal change law of cultivated land loss and its spatial distribution. This analysis can identify which regions' cultivated land loss is affected by both temporal and spatial factors. Through the generated spatio-temporal correlation coefficients, the correlation strength between cultivated land loss in different regions and spatio-temporal factors can be quantitatively measured. Next, cluster analysis and identification are performed on the target region based on the spatio-temporal correlation coefficients to obtain multi-type regions. Cluster analysis will divide the target region into multiple types according to the spatio-temporal characteristics and correlation coefficients of cultivated land loss. These types include high-loss areas, low-loss areas, and fluctuating areas. High-loss areas refer to those regions where cultivated land loss is closely related to the spatio-temporal change trend and the loss degree is severe; low-loss areas are those where the change in cultivated land loss is not obvious and remains relatively stable; while fluctuating areas refer to those regions where the change in cultivated land loss is relatively large and may be affected by various factors, showing a relatively high change amplitude. On this basis, the cultivated land loss level of the target region is divided by combining the historical cultivated land loss record data of the target region. The cultivated land loss level division is based on the degree of cultivated land loss and is usually divided into high-loss areas, low-loss areas, and fluctuating areas. High-loss areas usually need to be treated preferentially; the management strategy for low-loss areas can be moderate; and the information of fluctuating regions may require regular monitoring and adjustment of management measures. According to the spatio-temporal change trend and historical cultivated land loss records, the cultivated land loss risk of the target region is further divided to determine the information of high-risk loss regions, medium-risk loss regions, and low-risk loss regions. The cultivated land loss risk division is an assessment based on the cultivated land loss rate. The higher the rate, the greater the risk, which is usually measured by the change speed of the loss area and the loss trend. For example, if the cultivated land loss speed in a certain region is relatively fast and continues to intensify over time, this region will be classified as a high-risk loss region; regions with slower loss changes will be classified as low-risk loss regions. Next, the information of high-loss regions, low-loss regions, fluctuating regions is matched with the information of high-risk loss regions, medium-risk loss regions, and low-risk loss regions according to the spatio-temporal correlation coefficients, and hierarchical classification is performed based on the matching results to generate multiple classification layers.This step classifies and stratifies by matching the information on the loss level and risk level of the area with the spatio-temporal correlation coefficient, ensuring that the spatial distribution of cultivated land loss can fully reflect the regional characteristics of different loss levels and risk levels. Through hierarchical classification, different types of cultivated land loss areas can be clearly displayed on the map, helping managers identify high-risk and low-risk areas and take corresponding measures. Finally, the multiple classification layers are associated and integrated to construct the spatial distribution map of cultivated land loss. This final layer can intuitively display the spatial distribution of cultivated land loss in the target area, and various loss areas of different categories and risk levels are clearly marked, providing an effective visualization tool and decision-making basis for the monitoring, management and decision-making of cultivated land loss. Through the integration of these layers, detailed and clear geographical information support can be provided for the dynamic management and policy formulation of cultivated land loss.

[0030] Furthermore, map the remote sensing data set to the spatial distribution map of cultivated land loss for the monitoring of cultivated land loss, and generate a monitoring report, including: constructing a three-dimensional spatial geographic coordinate system based on the target area, mapping the remote sensing data set to the three-dimensional spatial geographic coordinate system for geometric correction to generate a remotely sensed corrected data set; performing matching slicing on the remotely sensed corrected data set according to the multiple classification layers to generate multiple remotely sensed corrected data sets; associatively storing the multiple remotely sensed corrected data sets in the multiple classification layers for synchronous monitoring to generate multiple types of monitoring data; performing differential analysis on the cultivated land loss in the target area based on the multiple types of monitoring data combined with the multiple remotely sensed corrected data sets to generate multiple differential data, and performing evaluation and identification according to the multiple differential data combined with the multiple classification layers to determine multiple loss classification layers; adding the multiple loss classification layers to the monitoring report.

[0031] Further, a three-dimensional spatial geographic coordinate system is constructed based on the geographic coordinate information of the target area. The three-dimensional spatial geographic coordinate system refers to a three-dimensional geographic coordinate system constructed by introducing the longitude, latitude, and elevation information of each point on the earth's surface, ensuring that remote sensing data can be accurately spatially located within a standard spatial framework. This coordinate system can provide a comprehensive geographic reference framework to ensure a high degree of consistency between the remote sensing image data and the actual geographical location of the target area. Next, the remote sensing data set is mapped to the three-dimensional spatial geographic coordinate system for geometric correction to ensure the consistency between the remote sensing image and the geographic coordinate system. Geometric correction means adjusting the pixel positions in the remote sensing image to the actual positions in the corresponding geographic coordinate system to eliminate geometric distortions caused by factors such as sensor angle and terrain undulation. Through geometric correction, the remote sensing image can be accurately aligned with the actual geographical location on the map, thereby improving the accuracy of subsequent analysis. During the geometric correction process, the pixel coordinates of the remote sensing image are matched with the longitude, latitude, and elevation data in the geographic coordinate system to generate a new remote sensing correction data set, which has been corrected in the geographic coordinate system and can be used for subsequent analysis and monitoring. Further, the remote sensing correction data set is sliced according to the multiple classification layers to generate multiple remote sensing correction data sets. Slicing means dividing different regions in the remote sensing image according to different land cover classification layers and pairing them with the corrected data set. Each classification layer represents different types of land cover information (such as cultivated land, water area, urban buildings, etc.). By slicing, these different types of data are extracted and processed independently. For example, the cultivated land area and urban area can be analyzed separately to ensure that the monitoring information of different land cover types can be accurately extracted. The remote sensing image data of each type of land cover will be processed separately and converted into multiple correction data sets, which are convenient for subsequent classification and analysis. Then, the multiple remote sensing correction data sets are associated and stored in the multiple classification layers to achieve synchronous monitoring and generate multiple types of monitoring data. Associated storage means storing each correction data set according to its corresponding classification layer information to ensure that the data of each land cover type can be updated synchronously and jointly analyzed. Synchronous monitoring refers to the real-time tracking and analysis of data of different land cover types within the same time period. For example, the remote sensing data of the cultivated land area and the urban area can be processed synchronously to achieve simultaneous monitoring of cultivated land loss and urban expansion. Furthermore, based on the multiple types of monitoring data and combined with the multiple remote sensing correction data sets, a differential analysis of cultivated land loss in the target area is carried out to generate multiple differential data. Differential analysis means calculating the differences between remote sensing images at different times. By comparing remote sensing data at different times, the areas of cultivated land loss, change, or expansion are identified. Differential analysis can reveal the specific changes in cultivated land loss in the target area, helping decision-makers understand the speed and trend of cultivated land loss between different time points.For example, by comparing remote sensing data from 2010 and 2020, it is possible to detect which areas have experienced a significant reduction in cultivated land area, thereby assessing the progress of cultivated land loss. The generated difference data will serve as the basis for cultivated land loss monitoring, guiding subsequent assessment and identification work. Then, based on the difference data, combined with the multiple classification layers, assessment and identification are carried out to determine multiple loss classification layers. In this process, through further analysis and comparison of the difference data, different types of cultivated land loss areas can be identified. For example, based on the difference data, it can be evaluated which areas have significant cultivated land loss and which areas have relatively small changes. According to these analysis results, different loss classification layers can be divided, such as high-loss areas, low-loss areas, fluctuating areas, etc. Each classification layer represents a different degree of cultivated land loss, helping to conduct precise loss assessment and management for different regions. Finally, the multiple loss classification layers are added to the monitoring report. The monitoring report is a comprehensive document generated based on the aforementioned difference analysis and classification results. This report provides managers with detailed information on cultivated land loss and relevant data support, which can be used for decision-making analysis and management optimization.

[0032] Furthermore, the multiple remote sensing correction data sets are associated and stored with the multiple classification layers for synchronous monitoring to generate multiple types of monitoring data, including: using a support vector machine to perform supervised classification on the multiple remote sensing correction data sets to generate multiple data classes; associating and matching the multiple data classes with the multiple classification layers to generate layer matching results; mapping and storing the multiple remote sensing correction data sets to the multiple classification layers according to the multiple data classes in accordance with the layer matching results for updating to generate multiple classification update layers; and performing multi-temporal monitoring on the target area based on the multiple classification update layers to generate the multiple types of monitoring data.

[0033] Exemplarily, the multiple remotely sensed correction data sets are subjected to supervised classification by a support vector machine (SVM). A support vector machine is a machine learning algorithm commonly used for classification, especially suitable for classification problems of large-scale data. Through a training sample set, the support vector machine can learn the characteristics of different land cover types (such as cultivated land, forest, water area, urban buildings, etc.) and classify the remotely sensed image data based on these characteristics. Through this process, the remotely sensed data set can be divided into multiple data classes, and each data class corresponds to a different land cover category. Next, each data class will be matched with the corresponding classification layer according to its distribution in the geospatial space. The classification layer usually contains the division information of different land cover categories at different spatial positions, and the layer matching result is to correspond each classified data class to the corresponding geographical area. This matching result ensures the effective docking of the classified data with the land cover spatial information, thus providing accurate spatial positioning for subsequent data update and monitoring. Then, according to the layer matching result, the multiple remotely sensed correction data sets are mapped and stored into the multiple classification layers for update according to the multiple data classes, generating multiple classification update layers. Through this process, each classification layer will be updated according to the new data class information, and the new remotely sensed data will be mapped to the corresponding spatial positions according to the categories of the data classes, thus forming classification update layers. These update layers contain the latest information of each category in its respective area and provide accurate basic data for subsequent monitoring and analysis. For example, changes in cultivated land areas, expansion of new urban construction areas, reduction of forest cover, etc., will be accurately identified through the classification update layers. Furthermore, based on the multiple classification update layers, multi-temporal monitoring of the target area is carried out to generate the multi-class monitoring data. Multi-temporal monitoring refers to the superposition of classification results at multiple time points to form time series data. By analyzing the classification update layers at different times, the change situation of each land cover category (such as cultivated land, forest, water area, etc.) at different time periods can be formed. For example, the cultivated land layers in 2010, 2015, and 2020 can be superimposed to analyze the speed and spatial distribution changes of cultivated land loss. In this way, it is possible to clearly identify which areas have a relatively fast speed of cultivated land loss and which areas remain stable, thus helping to judge the trend and dynamic changes of the loss. These multi-class monitoring data will provide detailed data support for cultivated land loss monitoring and land use change monitoring. In cultivated land loss monitoring, by analyzing the changes in the cultivated land layer, the areas and speeds of cultivated land loss can be identified, and further the hotspots of cultivated land loss can be revealed, facilitating the management agency to take targeted protection measures. In land use change monitoring, by analyzing the spatial changes of each land cover category, the hotspots of land use change can be identified. For example, the increase in urban expansion areas, the transformation of agricultural areas into urban construction areas, etc., can all be effectively identified through this monitoring data, thus providing a scientific basis for land use planning and policy making.

[0034] Furthermore, perform dynamic analysis of cultivated land loss based on the monitoring report, and formulate a management plan for cultivated land loss, including: traversing the monitoring report to analyze hotspots of cultivated land loss in the target area to determine cultivated land loss aggregation areas; conducting source tracing analysis based on the cultivated land loss aggregation areas to determine multiple loss factors; dynamically evaluating the loss impact on the target area according to the multiple loss factors to obtain multiple loss impact factors; setting multiple management objectives according to the multiple loss impact factors, and setting multiple areas to be protected according to the cultivated land loss aggregation areas; associating and integrating the multiple management objectives with the multiple areas to be protected to formulate the management plan for cultivated land loss.

[0035] Specifically, hotspots refer to those areas with high loss rates, drastic changes, and significant risks, which are usually characterized by prominent large-scale reduction or loss of cultivated land. Through the analysis of hotspots of cultivated land loss, the concentrated areas of cultivated land loss can be accurately determined. The concentrated areas of cultivated land loss refer to the areas where the phenomenon of cultivated land loss is highly concentrated within a certain time range. The degree of loss in these areas is severe and is usually triggered or exacerbated by certain specific factors (such as urbanization process, climate change, etc.). By identifying these areas, key areas can be provided for subsequent management and intervention measures, thereby improving the management efficiency and effectiveness. Next, based on the concentrated areas of cultivated land loss, a loss source tracing analysis is carried out to determine multiple loss factors. Loss source tracing analysis refers to exploring the specific reasons for cultivated land loss by reviewing and analyzing historical data of cultivated land loss. This process helps to clarify which factors play a dominant role in the loss of cultivated land in a specific area. For example, urban expansion, agricultural development, climate change (such as droughts, floods) may all be the main driving factors of loss. Furthermore, based on the multiple loss factors, a dynamic assessment of the loss impact on the target area is carried out and multiple loss impact factors are obtained. The dynamic assessment of loss impact is to evaluate the relative importance of each factor by analyzing the temporal changes, spatial distribution, and impact degree of different factors on cultivated land loss. By establishing an impact assessment model, the specific impact values of each factor on cultivated land loss can be obtained, and the contributions of different factors to cultivated land loss can be quantified. For example, the urbanization process may have a significant impact on loss, while climate factors may play an important role in exacerbating loss in certain years. Through the assessment, multiple loss impact factors are finally obtained, and these factors reflect the impact intensity of different factors on cultivated land loss, providing a basis for formulating management objectives and protection measures. Next, multiple management objectives are set based on the multiple loss impact factors, and multiple areas to be protected are set according to the concentrated areas of cultivated land loss. In this step, the setting of management objectives should be adjusted according to different cultivated land loss factors and different loss impact factors. Management objectives refer to the cultivated land protection objectives expected to be achieved by taking specific measures within a certain time, such as reducing the loss area of certain high-risk areas, increasing the cultivated land recovery rate, etc. Combining the identification results of the concentrated areas of cultivated land loss, some areas to be protected can be set. These areas are usually the areas where cultivated land loss is the most serious or the areas with the greatest potential for recovery. The selection of areas to be protected is based on the analysis of the current situation of cultivated land loss and loss factors, focusing on protecting those areas with severe loss or great recovery potential to ensure that limited resources can be used most effectively. Finally, the multiple management objectives and the multiple areas to be protected are associated and integrated to formulate the cultivated land loss management plan. Association and integration refer to matching different management objectives and areas to be protected to ensure the reasonable allocation of resources and the effectiveness of management measures. For example, areas with high loss rates may require key loss mitigation objectives, while relatively stable areas can adopt monitoring and early warning measures.On this basis, the formulated farmland loss management plan will include specific goals, strategies and measures to achieve farmland protection, restoration and sustainable use. This plan may include land restoration, improvement of irrigation systems, restrictions on urban expansion, etc., and corresponding measures will be taken according to the characteristics and needs of different regions.

[0036] Example 2. Based on the same inventive concept as the method for dynamically monitoring farmland loss by integrating remote sensing and AI analysis in the foregoing example, as Figure 2 shown, the present application provides a platform for dynamically monitoring farmland loss by integrating remote sensing and AI analysis, and the platform includes:

[0037] A data acquisition module 11, configured to traverse and cover a target area according to a data acquisition frequency through remote sensing technology to obtain a remote sensing data set.

[0038] A semantic segmentation module 12, configured to perform semantic segmentation based on the remote sensing data set to obtain a plurality of segmentation data, perform spatial analysis on the plurality of segmentation data, and determine a plurality of spatial features.

[0039] A time series analysis module 13, configured to perform time series analysis on the target area according to the plurality of spatial features, capture regional spatial time series features, perform spatio-temporal change analysis on the regional spatial time series features and the plurality of spatial features, and generate a spatio-temporal change trend.

[0040] An association analysis module 14, configured to perform association analysis on the farmland loss of the target area according to the spatio-temporal change trend to determine a spatio-temporal association coefficient, perform identification and classification based on the spatio-temporal association coefficient in combination with the spatio-temporal change trend, and draw a spatial distribution map of farmland loss according to the classification result.

[0041] A dynamic analysis module 15, configured to map the remote sensing data set to the spatial distribution map of farmland loss for monitoring farmland loss, generate a monitoring report, perform dynamic analysis of farmland loss according to the monitoring report, and formulate a farmland loss management plan.

[0042] Furthermore, the time series analysis module 13 is further configured to perform the following steps: arrange the remote sensing data set according to the data acquisition time series to generate a remote sensing time series, the remote sensing time series has a plurality of time nodes, and the plurality of time nodes have a corresponding relationship with the remote sensing data set; associate and match the plurality of spatial features with the remote sensing data set according to the plurality of time nodes in the remote sensing time series to construct a spatial time series data set; perform dynamic capture on the target area based on the spatial time series data set to obtain the regional spatial time series features.

[0043] Furthermore, the timing analysis module 13 is further configured to perform the following steps: fuse the regional spatial timing features and the multiple spatial features in the time dimension to generate a time feature dataset; fuse the regional spatial timing features and the multiple spatial features in the spatial dimension to generate a spatial feature dataset; perform correlation analysis on the time feature dataset and the spatial feature dataset to generate multiple correlation coefficients; perform multi-scale spatio-temporal analysis on the regional spatial timing features and the multiple spatial features according to the multiple correlation coefficients to obtain multiple spatio-temporal change data; perform clustering analysis and identification based on the multiple spatio-temporal change data to determine multiple types of spatio-temporal change data, and construct the spatio-temporal change trend according to the multiple types of spatio-temporal change data.

[0044] Furthermore, the association analysis module 14 is further configured to perform the following steps: perform spatial autocorrelation analysis based on the spatio-temporal change trend to obtain spatial aggregation related parameters; perform spatio-temporal regression analysis based on the spatio-temporal change trend in combination with the spatial aggregation related parameters to generate spatio-temporal correlation coefficients; perform clustering analysis and identification on the target area according to the spatio-temporal correlation coefficients to obtain multiple types of areas; classify the target area according to the multiple types of areas in combination with the historical cultivated land loss record data of the target area to determine high-loss area information, low-loss area information, and fluctuating area information; divide the cultivated land loss risk of the target area according to the spatio-temporal change trend in combination with the historical cultivated land loss record data of the target area to determine high-risk loss area information, medium-risk loss area information, and low-risk loss area information; match the high-loss area information, the low-loss area information, the fluctuating area information with the high-risk loss area information, the medium-risk loss area information, and the low-risk loss area information according to the spatio-temporal correlation coefficients, and perform hierarchical classification according to the matching results to generate multiple classification layers; associate and integrate the multiple classification layers to construct the cultivated land loss spatial distribution map.

[0045] Furthermore, the dynamic analysis module 15 is further configured to perform the following steps: construct a three-dimensional spatial geographic coordinate system based on the target area, map the remote sensing dataset to the three-dimensional spatial geographic coordinate system for geometric correction to generate a remotely sensed corrected dataset; perform matching slicing on the remotely sensed corrected dataset according to the multiple classification layers to generate multiple remotely sensed corrected datasets; associate and store the multiple remotely sensed corrected datasets to the multiple classification layers for synchronous monitoring to generate multiple types of monitoring data; perform difference analysis on the cultivated land loss of the target area based on the multiple types of monitoring data in combination with the multiple remotely sensed corrected datasets to generate multiple difference data, and perform evaluation and identification according to the multiple difference data in combination with the multiple classification layers to determine multiple loss classification layers; add the multiple loss classification layers to the monitoring report.

[0046] Furthermore, the dynamic analysis module 15 is further configured to perform the following steps: perform supervised classification on the multiple remotely sensed correction data sets by using a support vector machine to generate multiple data classes; associate and match the multiple data classes with the multiple classification layers to generate a layer matching result; map and store the multiple remotely sensed correction data sets into the multiple classification layers according to the layer matching result for updating to generate multiple classification update layers; perform multi-temporal monitoring on the target area based on the multiple classification update layers to generate the multiple types of monitoring data.

[0047] Furthermore, the dynamic analysis module 15 is further configured to perform the following steps: traverse the monitoring report to analyze the hot spots of cultivated land loss in the target area to determine the cultivated land loss aggregation areas; perform loss source tracing analysis based on the cultivated land loss aggregation areas to determine multiple loss factors; perform dynamic assessment of the loss impact on the target area according to the multiple loss factors to obtain multiple loss impact factors; set multiple management objectives according to the multiple loss impact factors, and set multiple areas to be protected according to the cultivated land loss aggregation areas; associate and integrate the multiple management objectives with the multiple areas to be protected to formulate the cultivated land loss management plan.

[0048] Through the foregoing detailed description of the method for dynamically monitoring cultivated land loss by integrating remote sensing and AI analysis in this specification, those skilled in the art can clearly know the platform for dynamically monitoring cultivated land loss by integrating remote sensing and AI analysis in this embodiment. For the platform disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0049] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A dynamic monitoring method for cultivated land loss integrating remote sensing and AI analysis, characterized in that: The method comprises: Through remote sensing technology, the target area is covered and traversed according to the data collection frequency to obtain remote sensing data sets; Performing semantic segmentation based on the remote sensing data set to obtain a plurality of segmented data, performing spatial analysis on the plurality of segmented data, and determining a plurality of spatial features; Performing a time series analysis on the target area according to the multiple spatial features, capturing the regional spatial time series features, performing a time-space change analysis on the regional spatial time series features and the multiple spatial features, and generating a time-space change trend; According to the temporal and spatial variation trend, a correlation analysis is performed on the cultivated land loss in the target area to determine the temporal and spatial correlation coefficient, identification and classification are performed based on the temporal and spatial correlation coefficient combined with the temporal and spatial variation trend, and a spatial distribution map of cultivated land loss is drawn according to the classification result; The remote sensing data set is mapped to the spatial distribution map of cultivated land loss to monitor cultivated land loss, generate a monitoring report, perform a dynamic analysis of cultivated land loss based on the monitoring report, and formulate a cultivated land loss management plan.

2. The method for dynamic monitoring of cultivated land loss integrating remote sensing and AI analysis as claimed in claim 1, characterized in that: Performing time series analysis on the target area according to the multiple spatial features to capture the regional spatial time series features includes: Arranging the remote sensing data set according to a data acquisition time sequence to generate a remote sensing time sequence, wherein the remote sensing time sequence has a plurality of time nodes, and the plurality of time nodes correspond to the remote sensing data set; Associating and assembling the multiple spatial features with the remote sensing data set according to the multiple time nodes in the remote sensing time series to construct a spatial time series data set; The target area is dynamically captured based on the spatial time series data set to obtain the spatial time series characteristics of the area.

3. The method for dynamic monitoring of cultivated land loss integrating remote sensing and AI analysis as claimed in claim 1, characterized in that: Performing a spatiotemporal change analysis on the regional spatial time series features and the multiple spatial features to generate a spatiotemporal change trend includes: Based on the fusion of the regional spatial temporal features and the multiple spatial features according to the time dimension, a time feature data set is generated; Generate a spatial feature data set based on fusing the regional spatial temporal feature with the multiple spatial features according to the spatial dimension; Performing correlation analysis based on the temporal feature data set and the spatial feature data set to generate a plurality of correlation coefficients; Performing a multi-scale spatiotemporal analysis on the regional spatial time series characteristics and the multiple spatial characteristics according to the multiple correlation coefficients to obtain multiple spatiotemporal change data; Cluster analysis and identification are performed based on the multiple spatiotemporal change data to determine multiple categories of spatiotemporal change data, and the spatiotemporal change trend is constructed according to the multiple categories of spatiotemporal change data.

4. The method for dynamic monitoring of cultivated land loss integrating remote sensing and AI analysis as claimed in claim 1, characterized in that: According to the temporal and spatial variation trend, a correlation analysis is performed on the cultivated land loss in the target area to determine the temporal and spatial correlation coefficient, identification and classification are performed based on the temporal and spatial correlation coefficient combined with the temporal and spatial variation trend, and a spatial distribution map of cultivated land loss is drawn according to the classification result, including: Perform spatial autocorrelation analysis based on the temporal and spatial variation trends to obtain spatial aggregation related parameters; Based on the spatiotemporal variation trend and the spatial aggregation related parameters, a spatiotemporal regression analysis is performed to generate a spatiotemporal correlation coefficient; Performing cluster analysis and identification on the target area according to the spatiotemporal correlation coefficient to obtain multiple types of areas; The target area is classified into the level of cultivated land loss by combining the multi-type areas with the historical cultivated land loss record data of the target area, and the information of high-loss area, low-loss area and fluctuating area is determined; According to the temporal and spatial variation trend and the historical cultivated land loss record data of the target area, the cultivated land loss risk of the target area is divided to determine the high-risk loss area information, the medium-risk loss area information, and the low-risk loss area information; Matching the high churn area information, the low churn area information, the fluctuating area information with the high-risk churn area information, the medium-risk churn area information, and the low-risk churn area information according to the spatiotemporal correlation coefficient, performing hierarchical classification according to the matching results, and generating multiple classification layers; The multiple classification layers are associated and integrated to construct the spatial distribution map of cultivated land loss.

5. The method for dynamic monitoring of cultivated land loss integrating remote sensing and AI analysis as claimed in claim 4, characterized in that: Mapping the remote sensing data set to the spatial distribution map of cultivated land loss to monitor cultivated land loss and generate a monitoring report, including: Constructing a three-dimensional geographic coordinate system based on the target area, mapping the remote sensing data set to the three-dimensional geographic coordinate system for geometric correction, and generating a remote sensing correction data set; Matching and slicing the remote sensing correction data set according to the multiple classification layers to generate multiple remote sensing correction data sets; Associating and storing the multiple remote sensing correction data sets to the multiple classification layers for synchronous monitoring to generate multiple types of monitoring data; Based on the multiple types of monitoring data combined with the multiple remote sensing correction data sets, a difference analysis is performed on the cultivated land loss in the target area to generate multiple difference data, and an evaluation and identification is performed based on the multiple difference data combined with the multiple classification layers to determine multiple loss classification layers; The plurality of loss classification layers are added to the monitoring report.

6. The method for dynamic monitoring of cultivated land loss integrating remote sensing and AI analysis as claimed in claim 5, characterized in that: The multiple remote sensing correction data sets are associated and stored in the multiple classification layers for synchronous monitoring to generate multiple types of monitoring data, including: Performing supervised classification on the multiple remote sensing correction data sets using a support vector machine to generate multiple data classes; Associating and matching the multiple data classes with the multiple classification layers to generate a layer matching result; According to the layer matching results, the multiple remote sensing correction data sets are stored in the multiple classification layers according to the multiple data class mappings for updating, and multiple classification update layers are generated; Based on the multiple classification update layers, multi-temporal monitoring is performed on the target area to generate the multiple types of monitoring data.

7. The method for dynamic monitoring of cultivated land loss integrating remote sensing and AI analysis as claimed in claim 1, characterized in that: According to the monitoring report, dynamic analysis of cultivated land loss is conducted and management plan for cultivated land loss is formulated, including: Traversing the monitoring report, analyzing the hot spots of farmland loss in the target area, and determining the concentrated areas of farmland loss; Conduct loss source analysis based on the cultivated land loss concentration areas and determine multiple loss factors; Performing a dynamic evaluation of the impact of churn on the target area according to the multiple churn factors to obtain multiple churn impact factors; Setting a plurality of management targets according to the plurality of loss influencing factors, and setting a plurality of areas to be protected according to the cultivated land loss concentration areas; The multiple management objectives are associated and integrated with the multiple areas to be protected to formulate the farmland loss management plan.

8. A dynamic monitoring platform for cultivated land loss that integrates remote sensing and AI analysis, characterized by: The platform is used to implement the method for dynamic monitoring of cultivated land loss by integrating remote sensing and AI analysis as described in any one of claims 1 to 7, and comprises: The data acquisition module is used to cover and traverse the target area according to the data collection frequency through remote sensing technology to obtain remote sensing data sets; A semantic segmentation module, used for performing semantic segmentation based on the remote sensing data set, obtaining a plurality of segmentation data, performing spatial analysis on the plurality of segmentation data, and determining a plurality of spatial features; A time series analysis module, used to perform time series analysis on the target area according to the multiple spatial features, capture the regional spatial time series features, perform spatiotemporal change analysis on the regional spatial time series features and the multiple spatial features, and generate a spatiotemporal change trend; An association analysis module is used to perform association analysis on the cultivated land loss in the target area according to the temporal and spatial variation trend, determine the temporal and spatial correlation coefficient, perform identification and classification based on the temporal and spatial correlation coefficient combined with the temporal and spatial variation trend, and draw a spatial distribution map of cultivated land loss according to the classification result; The dynamic analysis module is used to map the remote sensing data set to the spatial distribution map of cultivated land loss to monitor cultivated land loss, generate a monitoring report, perform a dynamic analysis of cultivated land loss based on the monitoring report, and formulate a cultivated land loss management plan.

Images