Method and system for recording illegal pattern spot change property of defense film

Abstract

The invention discloses a method and a system for recording illegal pattern spot change properties of a defense film. The method comprises the following steps of: preparing and preprocessing data; carrying out satellite image classification based on an optimized random forest method; extracting a change reference region; land class pattern spot boundary and attribute correction; identifying a change type; through spatial position overlay analysis, comparing the change pattern spots with land utilization historical data, and judging the change properties of the pattern spots through a set judgment rule; and in the change type identification and change property analysis process, automatically recording the change type of the pattern spot and a change property analysis result, and displaying the change type and the change property analysis result in real time. Based on the latest database, the satellite image and other auxiliary data, change information is extracted through man-machine interaction, and the change range, the change type and the change property of the pattern spot are analyzed in real time, so that the detection efficiency and the accuracy of illegal pattern spot change are improved.

Description

Technical Field

[0001] The present invention relates to the field of analysis of satellite image law enforcement data, and particularly to a method and system for recording the change nature of illegal map patches in satellite images. Background Art

[0002] With the continuous development of remote sensing technology, the application of satellite images and remote sensing data has been widely used in land use monitoring, environmental management, etc. In the field of satellite image law enforcement, accurately and efficiently recording the change nature of illegal map patches is the key to ensuring the compliant use of land resources, timely discovering and handling illegal acts, which involves fields such as remote sensing technology, land use change monitoring, geographic information system (GIS) technology, and data analysis.

[0003] In the prior art, satellite image data is mainly collected manually and illegal map patches are identified and classified through post - analysis, so as to realize the identification of changed map patches in satellite images and the analysis of change nature. This process requires a large amount of human and time costs, and there are problems such as low efficiency, poor real - time performance, and insufficient accuracy. At the same time, due to the complexity of manual analysis, it is often difficult to timely reflect the nature of map patch changes, such as changes in types like non - agriculturalization and non - grainification of cultivated land, resulting in a lag in the feedback of map patch change information and reducing the efficiency and accuracy of monitoring. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for recording the change nature of illegal map patches in satellite images, which can at least improve the detection efficiency and accuracy of changes in illegal map patches.

[0005] According to the first aspect of the present invention, there is provided a method for recording the change nature of illegal map patches in satellite images, which includes: Data preparation and pre - processing: Prepare satellite image data, perform radiometric correction and atmospheric correction; load the latest land use data, add a map patch number field to the land use data, assign values to the field according to the numbering rule, and obtain a map patch number for each map patch; collect regional crop planting attribute vector data and on - site photo data; Satellite image classification: Perform satellite image classification based on an optimized random forest method to obtain the satellite image classification result; Change reference area extraction: Based on the satellite image classification result and land use historical data analysis, obtain the reference area for the change of land use categories in satellite images; based on the regional crop planting attribute vector data and land use historical data analysis, obtain the reference area for the change of cultivated land planting attributes in satellite images; Boundary and attribute correction of land use map patches: Correct the patch boundaries of the land use type changes in the current land use data to accurately obtain the land use patch boundaries; correct the patch boundaries of the cultivated land planting attribute changes in the current land use data to accurately obtain the land use patch boundaries; During the patch boundary correction process, perform spatial position overlay analysis on the patches, the satellite image classification result data, and the regional crop planting attribute vector data, and modify the patch attribute values; Change type identification: Identify newly added patches and changed patches based on the attribute information and graphic information of the patches; Change nature analysis: Through spatial position overlay analysis, compare the changed patches with the historical land use data, and judge the change nature of the patches according to the set judgment rules; Among them, during the change type identification and change nature analysis processes, automatically record the change types and change nature analysis results of the patches, and display them in real time.

[0006] According to the method for recording the change nature of the satellite image illegal patches, in the data preparation and preprocessing steps, The land use data shall include the basic patch information, and ensure that the spatial coordinate system of the land use data is consistent with the spatial coordinate systems of the satellite image and the regional crop planting attribute vector data; The adopted numbering rule is the administrative region code + sequential code, ensuring that the patch numbers of each patch are unique and cannot be changed; Among them, the basic patch information includes the element number, land use type name, and planting attribute.

[0007] According to the method for recording the change nature of the satellite image illegal patches, in the satellite image classification step, perform satellite image classification based on the optimized random forest method, and the obtained satellite image classification results include: Multi-source feature extraction: Extract spectral features, vegetation indices, water body indices, enhanced vegetation indices, texture features, shape features, and band ratio features based on satellite remote sensing images; Training set and validation set construction: Select multiple regions from the existing remote sensing images, and label the land cover types of each region for use as the training set; select regions different from the training set, label the land cover types, and use them as the validation set; Feature screening: Use the mutual information theory to calculate the correlation between each feature and the land cover type samples, screen out the feature subset that contributes the most to the classification, and remove redundant and noisy features through the screening process; Adjustment of the random forest model construction method: Including weighted random sampling and enhancement of the randomness and diversity of the feature subsets; Model Training, Feature Importance Evaluation, and Iterative Optimization: The random forest model is used. By randomly extracting multiple subsets from the training data, multiple decision trees are trained. When each tree splits a node, a feature is randomly selected from a part of the features for splitting, reducing the risk of overfitting of the model. By adjusting the number of trees, the maximum tree depth, and the number of features selected during each split, the performance of the model is optimized. During the construction of the random forest model, by evaluating the importance of each feature in the decision tree node split, the feature subset is iteratively adjusted until the optimal classification effect is achieved; Classification and Prediction: The trained random forest model is used to classify remote sensing images. Each pixel will be assigned to a class, and the classification result is determined by the voting mechanism of multiple decision trees to decide the final class; Result Evaluation: The classification result is evaluated through confusion matrix, overall accuracy, and Kappa system metrics; Post - processing and Accuracy Improvement: This includes performing spatial filtering on the classification result to remove small - area noise and smooth the boundaries; improving the classification accuracy by optimizing the model parameters of the random forest and improving the quality of the selected training samples; Classification Result Arrangement: The image classification result data is converted into vector data. After conversion, each polygon is associated with the corresponding ground object class attribute. The value of the class in the raster is used as the ground class name attribute field value of the vector data. The irregular or serrated parts of the polygon boundary are smoothed, and adjacent patches with the same class are merged.

[0008] According to the method for recording the change nature of illegal land use map patches in satellite images, in the step of extracting the change reference area, Obtaining the reference area of land use class change in satellite images based on satellite image classification results and land use history data analysis includes: performing spatial position overlay analysis on satellite image classification results and land use history data, determining the area where the land use class has changed by comparing and analyzing the land use class name attributes, and thus forming the reference area for obtaining the land use class change in satellite images; Obtaining the reference area of cultivated land planting attribute change in satellite images based on regional crop planting attribute vector data and land use history data analysis includes: performing spatial position overlay analysis on regional crop planting attribute vector data and land use history data, obtaining the area where the planting type has changed by comparing and analyzing the planting attributes, and forming the reference area for obtaining the cultivated land planting attribute change in satellite images.

[0009] According to the method for recording the change nature of illegal land use map patches in satellite images, in the step of correcting the boundary and attributes of land use map patches, When performing spatial location overlay analysis on the patch and the classified result data of the satellite image and the vector data of the regional crop planting attributes, if the current patch intersects with multiple patches in the classified result layer or the vector layer of the regional crop planting attributes, the attribute value corresponding to the patch with the largest intersecting vector area is taken to modify the patch attribute value.

[0010] According to the method for recording the change nature of the satellite image illegal patches described above, in the change type identification step, identifying new patches and changed patches based on the attribute information and graphic information of the patches includes: Identification of new patches: Compare the attribute information of the currently edited patch data with the land use historical data. If the patch number of the current patch does not exist in the land use historical data, it is marked as a new patch; Identification of changed patches: Compare the graphic and attribute information of the currently edited patch data with the land use historical data. If the patch number of the current patch exists in the land use historical data, but the patch has changed in shape or attributes, it is marked as a changed patch.

[0011] According to the method for recording the change nature of the satellite image illegal patches described above, when identifying changed patches, compare the historical patches with the same patch number as the current patch. The change situation of the attribute information can be quickly obtained through field value comparison. If the attribute information has changed, it is marked as a changed patch, and the graphic information is no longer compared; if the attribute information has not changed, further compare whether the indicators of the graphic information have changed. If so, it is identified as a changed patch; the indicators of the graphic information include the graphic center point coordinates, area, perimeter, and the sequence of graphic boundary inflection point coordinates.

[0012] According to the method for recording the change nature of the satellite image illegal patches described above, in the change nature analysis step, the set judgment rules include: The situation where the attributes have changed but the form has not changed: When the boundary of the patch has not changed and only the attributes have changed, judge the change nature according to the change situation of the attributes: if the original patch land type name is cultivated land and the current patch land type name is non-cultivated land, then the patch change nature is the conversion of cultivated land to non-cultivated land; if the original patch land type name is cultivated land and the current patch land type name is still cultivated land, but the planting attribute has changed from the original main grain crop to a non-main grain crop, then the change nature of the current patch is the non-grainization of cultivated land; if the current patch and the original patch have changed in attributes but not in form, and do not conform to the above conversion of cultivated land to non-cultivated land or non-grainization of cultivated land, then the change nature of the current patch is other; The situation where the form has changed but the attributes have not changed: Since there is no change in the patch attributes, the change nature of the current patch is other; The situation where attribute changes are accompanied by form changes: When the current patch is formed by merging multiple original patches, it is judged and analyzed based on the first sub-rule; when the current patch intersects with multiple original patches, it is judged and analyzed based on the second sub-rule; when the current patch is split from the original patch, it is judged and analyzed based on the third sub-rule.

[0013] According to the method for recording the change nature of the illegal patches in the satellite images, the first sub-rule includes: Merging without arable land: If the land type names of the multiple original patches before merging do not include arable land, then the change nature of the current patch is other; Merging with arable land: When the land type names of the multiple original patches before merging include arable land: if the land type name of the current patch is not arable land, then the change nature is non-agriculturalization of arable land; if the land type name of the current patch is still arable land, and the planting attribute of the original arable patch includes staple food crops and the planting attribute of the current patch becomes non-staple food crops, then the change nature of the current patch is non-grainification of arable land; if the land type name of the current patch is still arable land, and the planting attribute of the patch does not change from staple food crops to non-staple food crops, then the change nature of the current patch is other; The second sub-rule includes: Extract the overlapping area between the current patch and the multiple original patches; When the overlapping area does not include arable land, then the change nature of the current patch is other; When the overlapping area includes arable land: if the current patch is not arable land, then the change nature of the patch is non-agriculturalization of arable land; if the current patch is still arable land, and the planting attribute of the original arable patch in the overlapping area includes staple food crops and the planting attribute of the current patch becomes non-staple food crops, then the change nature of the current patch is non-grainification of arable land; if the current patch is still arable land, and the planting attribute of the patch does not change from staple food crops to non-staple food crops, then the change nature of the current patch is other; The third sub-rule includes: When the land type name of the original patch is not arable land, then the change nature of the current patch is other; When the land type name of the original patch is arable land: if the current patch is not arable land, then the change nature is non-agriculturalization of arable land; if the current patch is still arable land, and the planting attribute of the arable land of the original patch is staple food crops and the planting attribute of the arable land of the current patch is non-staple food crops, then the change nature of the current patch is non-grainification of arable land; if the current patch is still arable land, and the planting attribute of the patch does not change from staple food crops to non-staple food crops, then the change nature of the current patch is other.

[0014] According to the second aspect of the present invention, a system is provided, which includes a processor and a memory, and the memory stores multiple instructions; the processor loads the instructions from the memory to execute the method for recording the change nature of the illegal patches in the satellite images as described above.

[0015] Beneficial effects: Combination of manual collection and automatic analysis: Different from the prior art that completely relies on manual or automated collection, this technical solution uses a human-computer interaction method to correct the boundaries and attributes of changed patches, giving full play to the accuracy of manual collection and the efficiency of automatic analysis, and avoiding the limitations of relying solely on a certain technical means; Real-time and high efficiency: Traditional methods usually perform change analysis after patch collection, while this technical solution can analyze the types and natures of changes in real time during the patch editing process, provide immediate feedback, and improve the response speed of land illegal behavior monitoring; Improved accuracy: By automatically analyzing the types and natures of changed patches, subjective errors in manual analysis are reduced, and the recognition accuracy of illegal patches is improved. Automatic recording and classification helps improve the quality and consistency of data; Simplify subsequent analysis work: By automatically generating change records, the workload of subsequent data processing and analysis in traditional methods is reduced, making land monitoring simpler and more efficient.

[0016] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be understood through the practice of the present invention. Brief description of the drawings

[0017] The present invention will be further described below in conjunction with the drawings and embodiments: Figure 1 It shows the overall process of automatic classification of satellite images to extraction of changed patches, and automatic analysis and recording of the types and natures of patch changes; Figure 2 Then it details the specific process of machine learning automatic classification of satellite images, automatic extraction of change reference areas, human-computer interaction to correct the boundaries and attributes of changed patches, and automatic determination of the types and natures of changed patches using preset change type analysis rules and change nature analysis rules; Figure 3 It presents an operation scenario during the satellite image patch collection process: analyzing, recording, and real-time displaying the collected patches through preset change recognition rules. Detailed implementation manners

[0018] This part will describe in detail the specific embodiments of the present invention. The preferred embodiments of the present invention are shown in the drawings. The drawings are used to supplement the description in the text part of the specification, enabling people to intuitively and vividly understand each technical feature and the overall technical solution of the present invention, but it cannot be understood as a limitation on the protection scope of the present invention.

[0019] Refer to Figures 1 - 3, the method for recording the change nature of illegal map patches in satellite images according to the embodiments of the present invention includes: Data preparation and preprocessing: Prepare satellite image data, perform radiometric correction and atmospheric correction; load the latest land use data, add a map patch number field to the land use data, assign values to the field according to the numbering rule, and obtain a map patch number for each map patch; collect regional crop planting attribute vector data and on-site photo data; Perform satellite image classification: Perform satellite image classification based on the optimized random forest method to obtain the satellite image classification results; Extract the change reference area: Obtain the satellite image land type category change reference area based on the satellite image classification results and land use historical data analysis; obtain the satellite image cultivated land planting attribute change reference area based on the regional crop planting attribute vector data and land use historical data analysis; Modify the land type map patch boundary and attributes: Modify the map patch boundary of the land type category change in the current land use data to accurately obtain the land type map patch boundary; modify the map patch boundary of the cultivated land planting attribute change in the current land use data to accurately obtain the land type map patch boundary; During the map patch boundary modification process, perform spatial position overlay analysis on the map patch with the satellite image classification result data and the regional crop planting attribute vector data, and modify the map patch attribute value; Identify the change type: Identify new map patches and changed map patches based on the attribute information and graphic information of the map patches; Analyze the change nature: Through spatial position overlay analysis, compare the changed map patches with the land use historical data, and judge the change nature of the map patches through the set judgment rules; Among them, during the change type identification and change nature analysis process, automatically record the change type and change nature analysis results of the map patches, and display them in real time.

[0020] Specifically, in the data preparation and preprocessing step, the satellite image data is the main basis for manually collecting map patches, and its clarity and timeliness are crucial. The regional crop planting attribute vector data, on-site photo data, etc. are used to assist in confirming the planting attributes of crops.

[0021] The land use data shall include detailed basic information of map patches, such as feature numbers, land use types, planting attributes, etc., and ensure that the spatial coordinate system of the land use data is consistent with that of data such as satellite image, vector of regional crop planting attributes, etc. The consistency of the coordinate system ensures the spatial alignment of the map patch data and the image data, so as to avoid errors caused by coordinate misalignment and ensure the accuracy of the extraction of changed map patches. Add a map patch number field to the land use data, and assign values to the field according to a custom numbering rule (such as generating numbers in the way of administrative region code + sequence code), ensuring that the map patch numbers of each map patch are unique and unchangeable, providing a basis for subsequent comparative analysis based on the map patch numbers. In addition, back up a copy of the land use data as historical data to provide a basis for subsequent comparative analysis.

[0022] In the steps of satellite image classification, satellite image classification is carried out based on the optimized random forest method to obtain the satellite image classification results, which include: multi-source feature extraction, construction of training set and validation set, feature screening, adjustment of the construction method of random forest model, model training, feature importance evaluation and iterative optimization, classification and prediction, result evaluation, post-processing and accuracy improvement, and classification result arrangement. In order to automatically and accurately analyze and determine the land use type attributes of map patches and the reference range of map patch changes in the subsequent stages of map patch collection, this technical solution adopts a satellite image land use type classification method based on the improved random forest method, and improves the accuracy and efficiency of satellite image land use type classification by optimizing feature screening and improving the decision tree generation strategy. The core of the random forest algorithm lies in enhancing the prediction ability of the entire model by integrating multiple weak learners (i.e., decision trees), rather than relying solely on a single decision tree.

[0023] Among them, in the sub-steps of multi-source feature extraction, spectral features, vegetation indices, water body indices, enhanced vegetation indices, texture features, shape features, and band ratio features are extracted based on satellite remote sensing images. In satellite remote sensing image classification, different ground object categories have different spectral and spatial characteristics. The purpose of feature extraction is to convert these characteristics into data that can be processed by the classification model. This technical solution mainly constructs a rich feature set by extracting the following multiple features to provide a rich information basis for subsequent classification.

[0024] Spectral feature extraction: Spectral features are the basis for remote sensing image classification. The reflection characteristics of different ground objects are significantly different in different bands. Remote sensing images usually include multiple bands such as visible light, near-infrared (NIR), and short-wave infrared (SWIR), and visible light includes red light (Red), green light (Green), and blue light (Blue). The difference in reflectance of these bands can help distinguish different ground object categories.

[0025] Vegetation index (NDVI): The normalized difference vegetation index (NDVI) is the most commonly used remote sensing image vegetation monitoring index, and its calculation formula is: NDVI = (NIR - Red) / (NIR + Red) High NDVI values (close to 1) indicate strong vegetation cover and are applicable to categories such as forest land, orchard land, and grassland; low NDVI values (close to 0 or negative values) help distinguish non-vegetation areas such as water bodies, buildings, and construction sites.

[0026] Modified Normalized Difference Water Index (MNDWI): The Modified Normalized Difference Water Index (MNDWI) is used to extract water bodies, and the calculation formula is: MNDWI = (Green - SWIR) / (Green + SWIR) Regions with higher MNDWI values represent water bodies and are particularly effective for water body extraction. Other features (such as buildings, construction sites, cultivated land, etc.) usually have lower MNDWI values.

[0027] Enhanced Vegetation Index (EVI): The Enhanced Vegetation Index (EVI) is an improved version of the vegetation index and is applicable to areas with high-density vegetation. The calculation formula is: Among them, G, C1, C2, and L are constants. EVI can well distinguish high-density forest land, orchard land, and grassland. Especially in dense vegetation areas, EVI can effectively reduce the atmospheric influence and extract vegetation information.

[0028] Texture features: Texture features are very effective for distinguishing feature types with similar spectral characteristics. The gray-level co-occurrence matrix (GLCM) method is used to extract the texture features of the image, and texture features are very effective for distinguishing features with similar spectral characteristics but different morphologies.

[0029] Shape features: Shape features are very effective for distinguishing areas with regular shapes such as buildings and construction sites. For example, among land types, buildings usually present regular shapes such as rectangles and polygons, construction site areas may show irregular shapes, and water bodies usually show regular circles or irregular edges.

[0030] Band ratio features: Band ratios (such as red / near-infrared ratio, red / green ratio) help distinguish different feature types. The red / near-infrared ratio is applicable to distinguish cultivated land, forest land, and grassland because they have different reflection patterns in the red and near-infrared bands. The red / green ratio is used to distinguish buildings from orchard land, and buildings usually exhibit lower reflectance.

[0031] In the sub - steps of constructing the training set and validation set, select multiple representative regions from existing remote - sensing images, and label the land - cover classes of each region (including cultivated land, buildings, construction sites, gardens, forests, grasslands, water bodies, others, etc.) for use as the training set. These regions should cover various land - cover types and preferably include different features of the images. Select regions different from the training set, label the land - cover classes, and use them as the validation set for subsequent model evaluation.

[0032] In the sub - steps of feature selection, use the mutual - information theory to calculate the correlation between each feature and the land - cover samples, and select the feature subset that makes the greatest contribution to classification. Through the selection process, redundant and noisy features can be effectively removed, improving the performance of the classifier.

[0033] In the sub - steps of adjusting the construction method of the random - forest model, when constructing the random - forest model, mainly make the following two adjustments: 1) Weighted random sampling Adopt a category - based weighted random - sampling method to balance the number of land - cover samples of different classes, reduce classification bias, and improve the model's recognition ability for different land - cover classes; 2) Enhancement of feature - subset randomness and diversity At each node split, randomly select a part of the features from the selected feature subset for optimal splitting, and at the same time ensure the diversity of the feature subsets between different decision trees, enhancing the generalization ability and robustness of the model.

[0034] In the sub - steps of model training, feature - importance evaluation, and iterative optimization: Model training includes: Use the adjusted random - forest model for training. The random forest randomly selects multiple subsets (bootstrap sampling) from the training data to train multiple decision trees. When each tree splits nodes, only randomly select one feature from a part of the features for splitting, thereby reducing the risk of model overfitting. Optimize the model's performance by adjusting parameters such as the number of trees (n_estimators), the maximum depth of the tree (max_depth), and the number of features selected at each split (max_features); Feature - importance evaluation and iterative optimization include: During the construction process of the random - forest model, by evaluating the importance of each feature in the decision - tree node split, iteratively adjust the feature subset until the optimal classification effect is achieved, ensuring that the selected features are not only highly correlated with the land - cover classes but also play a key role in model construction.

[0035] In the sub-step of classification and prediction, the trained random forest model is used to classify remote sensing images. Each pixel will be assigned to a class, and the classification result is determined by the voting mechanism (majority voting principle) of multiple decision trees. When outputting the classification result, the output is the classification label of each pixel. For example, a certain pixel may be classified as cultivated land or forest land.

[0036] In the sub-step of result evaluation, the classification result is evaluated through indicators such as the confusion matrix, overall accuracy, and Kappa coefficient. Among them, the confusion matrix is used to evaluate the accuracy of the classification result and count the confusion between different classes; the overall accuracy calculates the proportion of correctly classified pixels in the total number of pixels; the Kappa coefficient measures the difference between the classification result and the random classification result, reflecting the reliability of the classification.

[0037] In the sub-step of post-processing and accuracy improvement, it includes filtering and accuracy improvement. The filtering here is to perform spatial filtering on the classification result to remove small-area noise and smooth the boundaries. The accuracy improvement is achieved by optimizing the model parameters of the random forest and improving the quality of the selected training samples to enhance the classification accuracy.

[0038] In the sub-step of classification result arrangement, the image classification result data is converted into vector data, that is, the pixel value of each class is converted into a closed polygon (area) object. After conversion, each polygon is associated with the corresponding ground object class, and the value of the class in the raster is used as the attribute field value of the land class name of the vector data. The irregular or jagged parts existing in the polygon boundary are smoothed, and adjacent patches with the same class also need to be merged.

[0039] In the step of extracting the change reference area, based on the classification results of satellite images and the analysis of land use historical data, the change reference area of the land class of satellite images is obtained. Specifically, it includes: performing a spatial position overlay analysis on the classification results of satellite images and land use historical data, and determining the area where the land class has changed by comparing and analyzing the land class name attributes, and thus forming the change reference area of the land class of satellite images to provide a spatial reference for subsequent boundary correction of land class patches; Based on the vector data of regional crop planting attributes and the analysis of land use historical data, the change reference area of the cultivated land planting attributes of satellite images is obtained. Specifically, it includes: performing a spatial position overlay analysis on the vector data of regional crop planting attributes and land use historical data, and obtaining the area where the planting type has changed by comparing and analyzing the planting attributes, and forming the change reference area of the cultivated land planting attributes of satellite images to provide a spatial reference for the precise acquisition of the boundary of the planting attribute change area.

[0040] In the steps of correcting the boundaries and attributes of land type patches, referring to the data of the reference areas of land type changes in the satellite images obtained in the step of extracting the reference areas of changes, based on the satellite image data, manually correct the boundaries of the patches with changed land type categories in the current land use data, and accurately edit to obtain the boundaries of land type patches. Referring to the data of the reference areas of changes in the cultivated land planting attributes in the satellite images obtained in the step of extracting the reference areas of changes, based on the satellite image data and on-site photo materials, manually correct the boundaries of the patches with changed cultivated land planting attributes in the current land use data, and accurately edit to obtain the boundaries of land type patches.

[0041] In the process of correcting the patch boundaries, the patches are automatically subjected to spatial position overlay analysis with the satellite image classification result data obtained in the above steps and the vector data of the regional crop planting attributes collected, and the attribute values such as the land type name and planting attributes of the patches are automatically modified. In the process of overlay analysis, if the current patch intersects with multiple patches in the classification result layer or the vector layer of regional crop planting attributes, then take the attribute value corresponding to the patch with the largest intersecting vector area, and the taken attribute value is used for modifying the patch attribute value.

[0042] The attribute information such as the land type name and planting attributes of the land type patches is mainly obtained through automatic machine analysis, and the collection personnel manually intervene to confirm the patch attribute information in combination with data materials such as satellite images and on-site photos. In particular, the planting types of crops are difficult to directly analyze through image information and need to be supplemented or modified manually based on more materials. During the editing process, the newly generated patches will automatically obtain unique numbers according to the numbering rules mentioned in the data preparation and preprocessing steps.

[0043] In the change type recognition step, new patches and changed patches are recognized based on the attribute information and graphic information of the patches, which includes new patch recognition and changed patch recognition. Among them, the judgment rule for new patch recognition is: compare the attribute information of the currently edited patch data with the land use historical data. If the patch number of the current patch does not exist in the land use historical data, it is automatically marked as a new patch. The judgment rule for changed patch recognition is: compare the graphic and attribute information of the currently edited patch data with the land use historical data. If the patch number of the current patch exists in the land use historical data, but the patch has changed in shape (by comparing indicators such as the central point coordinates, area, perimeter, and boundary point coordinate sequence of the patch) or attributes, it is automatically marked as a changed patch.

[0044] Specifically, during the identification process of changed patches, only the historical patches with the same patch number as the current patch need to be compared. The change situation of attribute information can be quickly obtained through field value comparison. If the attribute information changes, it is marked as a changed patch, and the graphic information is no longer compared. If the attribute information does not change, the graphic information is further compared. Since the change situation of the graphics is relatively complex, some are obtained by splitting historical patches, some are obtained by merging multiple historical patches, and some are obtained by modifying the boundaries of historical patches. It is necessary to comprehensively judge whether the graphics have changed according to multiple indicators. This technical solution makes a comparison and judgment by combining indicators such as the central point coordinates, area, perimeter, and graphic boundary inflection point coordinate sequence of the graphics. If any of these indicators changes, it is automatically identified as a changed patch.

[0045] In the step of analyzing the change nature, through spatial position overlay analysis, the changed patches are compared with the historical land use data, and the change nature of the patches is automatically judged according to the preset judgment rules. This process mainly focuses on the attribute changes of the patches, especially the changes in the attributes related to cultivated land. The judgment results (the change nature of the patches) mainly include three situations: non-agriculturalization of cultivated land, non-grainification of cultivated land, and others. The set judgment rules specifically include: The situation where the attribute changes but the form does not change: When the boundary of the patch does not change and only the attribute changes, the change nature is judged according to different attribute change situations: if the original patch land use type name is cultivated land and the current patch land use type name is non-cultivated land, then the change nature of the patch is non-agriculturalization of cultivated land; if the original patch land use type name is cultivated land and the current patch land use type name is still cultivated land, but the planting attribute changes from the original main grain crop to a non-main grain crop, then the change nature of the current patch is non-grainification of cultivated land; if the current patch and the original patch have changed attributes but the form has not changed, then the change nature of the current patch is other; The situation where the form changes but the attribute does not change: Since there is no change in the patch attribute, the change nature of the current patch is other; The situation where the attribute change is accompanied by the form change: When the patch form changes, consider the spatial relationship between the current patch and the original relevant patches, such as merging, intersecting, or splitting, etc., and make an automatic judgment based on the land use type name and planting attribute information before and after the change. The following is the judgment according to the spatial relationship situation between the current patch and the historical patch.

[0046] (1) When the current patch is merged from multiple patches in the historical data, judge and analyze based on the first sub-rule: No cultivated land merger: If the land use type names of the original multiple patches before the merger do not contain cultivated land, then the change nature of the current patch is other; Including arable land consolidation: If the land type names of the original multiple patches before consolidation contain arable land, it is further processed in the following situations; A. The current patch is not arable land: If the land type name of the current patch is not arable land, the nature of the change is the non - agriculturalization of arable land; B. The current patch is arable land: If the land type name of the current patch is still arable land, it is necessary to further judge the planting attribute. For example, if the planting attribute of the original arable land patch contains staple food crops and the planting attribute of the current patch changes to non - staple food crops, the nature of the change of the current patch is the non - grainification of arable land; if the planting attribute of the patch does not change from staple food crops to non - staple food crops, the nature of the change of the current patch is other.

[0047] (2) When the current patch intersects with the original multiple patches in the historical data, mainly analyze the attribute changes in the overlapping area rather than the whole patch, and judge and analyze based on the second sub - rule: Extract the overlapping area between the current patch and the original multiple patches; The original patches corresponding to the overlapping area have no arable land: If the original multiple intersecting patches do not contain arable land in the overlapping area, the nature of the change of the current patch is other; The original patches corresponding to the overlapping area contain arable land: If the original multiple intersecting patches contain arable land in the overlapping area, it is further processed in the following situations; A. The current patch is not arable land: The nature of the change of the current patch is the non - agriculturalization of arable land; B. The current patch is still arable land, and it is necessary to further judge the planting attribute. For example, if the planting attribute of the original arable land patch in the overlapping area contains staple food crops and the planting attribute of the current patch changes to non - staple food crops, the nature of the change of the current patch is the non - grainification of arable land; if the planting attribute of the patch does not change from staple food crops to non - staple food crops, the nature of the change of the current patch is other.

[0048] Among them, when the current patch intersects with one original patch in the historical data, it belongs to the situation where the original patch has changed to a new patch. Essentially, they are still the same patch and have the same patch number.

[0049] (3) When the current patch is split from the original patch in the historical data, judge and analyze based on the third sub - rule: The original patch is non - arable land: If the land type name of the original patch is not arable land, the nature of the change of the current patch is other; The original patch is arable land: When the land type name of the original patch is arable land, it is further judged in the following situations; A. The current patch is not arable land, then the nature of the change is the non - agriculturalization of arable land; B. The current patch is still arable land, and its planting attribute needs to be further determined. For example, if the planting attribute of the original patch of arable land is for staple food crops and the planting attribute of the current patch of arable land is for non-staple food crops, then the nature of the change of the current patch is the non-grainification of arable land; if the planting attribute of the patch has not changed from staple food crops to non-staple food crops, then the nature of the change of the current patch is other.

[0050] If it is found that the result of the intelligent analysis of the nature of the change is unreasonable, the attribute can be manually corrected on the basis of the steps of the boundary and attribute correction of the land type patch to ensure the accuracy of the recorded result of the nature of the patch change.

[0051] During the process of change type recognition and change nature analysis, the change type and the result of change nature analysis of the patch are automatically recorded and displayed to the user in real time on the page (see Figure 3 ). For example, it is shown that the patch is newly added or changed, and the nature of the change (such as the non-agriculturalization of arable land, etc.) is given. The editor can immediately understand the change situation of the patch and confirm or modify the attribute information of the patch according to the real-time feedback information, so as to timely discover land use change problems, avoid lags and errors in the manual analysis process, and improve the efficiency and accuracy of land monitoring. This function not only ensures the accuracy of the change information, but also improves the work efficiency and response speed.

[0052] The present invention also provides a system, which includes a processor and a memory, and the memory stores multiple instructions; the processor loads instructions from the memory to execute the method for recording the nature of the change of the illegal patch of the satellite image as described above.

[0053] Although the methods described above have been illustrated and described as a series of acts for simplicity of explanation, it should be understood and appreciated that the methods are not limited by the order of the acts, since in accordance with one or more embodiments, some acts may occur in a different order and / or concurrently with other acts that are illustrated and described herein or that are not illustrated and described herein but would be understood by one of ordinary skill in the art. Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, the various illustrative components, blocks, modules, circuits, and steps have been described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention. The various illustrative logical blocks, modules, and circuits described in connection with the embodiments disclosed herein may be implemented using a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration. The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of both. The software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such that the processor can read from, and write to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In the alternative, the processor and the storage medium may reside as discrete components in a user terminal. In one or more exemplary embodiments, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functions may be stored on or transmitted via a computer readable medium as one or more instructions or code.Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. Storage media can be any available media that can be accessed by a computer. By way of example, and not limitation, such computer-readable media can include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Any connection is properly termed a computer-readable media. For example, if software is transmitted from a web site, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of the medium. As used herein, the terms "disk" and "disc" include compact discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy disks, and Blu-ray discs where disks typically reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0054] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those of ordinary skill in the art.

Claims

1. A method for recording the nature of changes in illegal spots in satellite images, characterized in that: include: Data preparation and preprocessing: Prepare satellite image data and perform radiation correction and atmospheric correction; Load the latest land use data, add a patch number field to the land use data, assign values ​​to the field using numbering rules, and obtain patch numbers for each patch; collect regional crop planting attribute vector data and field photo data; Classification of satellite images: Classify satellite images based on the optimized random forest method to obtain satellite image classification results; Variation reference area extraction: Based on the classification results of satellite images and the analysis of historical land use data, the reference areas of land category changes in satellite images are obtained; based on the analysis of regional crop planting attribute vector data and historical land use data, the reference areas of cultivated land planting attribute changes in satellite images are obtained; Correction of land class patch boundaries and attributes: Correct the boundaries of the spots where the land type changes in the current land use data, and accurately obtain the boundaries of the land type spots; Correct the boundaries of the spots where the cultivated land planting attributes change in the current land use data, and accurately obtain the boundaries of the land type spots; During the patch boundary correction process, the patch is spatially overlaid with the satellite image classification results data and the regional crop planting attribute vector data to modify the patch attribute value; Change type identification: Identify new and changed spots based on the spot's attribute information and graphic information; Analysis of the nature of changes: Through spatial position overlay analysis, the change spots are compared with the historical land use data, and the change nature of the spots is determined by the set judgment rules; Among them, in the process of change type identification and change nature analysis, the change type and change nature analysis results of the map are automatically recorded and displayed in real time.

2. The method for recording the changing nature of illegal spots in satellite images according to claim 1, characterized in that: In the data preparation and preprocessing steps, Land use data must include basic information about the spots and ensure that the spatial coordinate system of the land use data is consistent with the spatial coordinate system of satellite images and regional crop planting attribute vector data; The numbering rule adopted is administrative area code + sequence code, ensuring that the number of each map spot is unique and cannot be changed; Among them, the basic information of the map includes feature number, land type name, and planting attributes.

3. The method for recording the changing nature of illegal spots in satellite images according to claim 1, characterized in that: In the step of satellite image classification, satellite image classification is performed based on the optimized random forest method, and the satellite image classification results include: Multi-source feature extraction: Extract spectral features, vegetation index, water index, enhanced vegetation index, texture features, shape features and band ratio features based on satellite remote sensing images; Training set and validation set construction: From the existing remote sensing images, select multiple areas and mark the object category of each area to use as the training set; select areas different from the training set, mark the object category, and use them as the verification set; Feature screening: The mutual information theory is used to calculate the correlation between each feature and the land sample, and the feature subset that contributes most to the classification is screened out. Through the screening process, redundant and noise features are removed; Adjustments to the way the random forest model is built: Including weighted random sampling and feature subset randomness and diversity enhancement; Model training, feature importance evaluation and iterative optimization: The random forest model randomly extracts multiple subsets from the training data to train multiple decision trees. When each tree splits a node, it randomly selects one from some features to split, thereby reducing the risk of overfitting of the model. The performance of the model is optimized by adjusting the number of trees, the maximum tree depth, and the number of features selected at each split. In the process of building the random forest model, the importance of each feature in the splitting of the decision tree node is evaluated, and the feature subset is iteratively adjusted until the optimal classification effect is achieved. Classification and prediction: Use the trained random forest model to classify remote sensing images. Each pixel will be assigned to a category, and the classification results will be determined by the voting mechanism of multiple decision trees. Result evaluation: Evaluate the classification results through confusion matrix, overall accuracy and Kappa system indicators; Post-processing and accuracy improvement: This includes spatial filtering of the classification results, removing small-area noise, and smoothing boundaries; optimizing the model parameters of the random forest and improving the quality of the training sample selection to improve the classification accuracy; Classification results: The image classification result data is converted into vector data. After the conversion, each polygon is associated with the corresponding land feature category. The category value in the raster is used as the land class name attribute field value of the vector data. The irregular or jagged parts of the polygon boundary are smoothed, and adjacent patches of the same category are merged.

4. The method for recording the changing nature of illegal spots in satellite images according to claim 1, characterized in that: In the step of extracting the variation reference region, Obtaining a reference area for satellite image land category change based on satellite image classification results and land use historical data analysis includes: performing spatial position superposition analysis on satellite image classification results and land use historical data, determining the area where the land category has changed by comparing and analyzing the land category name attributes, and thereby forming a reference area for obtaining satellite image land category change; Obtaining a reference area for changes in cultivated land planting attributes in satellite images based on analysis of regional crop planting attribute vector data and land use historical data includes: performing spatial position overlay analysis on regional crop planting attribute vector data and land use historical data, obtaining areas where planting types have changed by comparative analysis of planting attributes, and forming a reference area for changes in cultivated land planting attributes in satellite images.

5. The method for recording the changing nature of illegal spots in satellite images according to claim 1, characterized in that: In the step of modifying the boundaries and attributes of land class patches, When performing spatial position overlay analysis on the map spots, satellite image classification results data, and regional crop planting attribute vector data, if the current map spot and multiple spots in the classification results layer or regional crop planting attribute vector layer intersect, the attribute value corresponding to the map spot with the largest intersection vector area is taken to modify the map spot attribute value.

6. The method for recording the changing nature of illegal spots in satellite images according to claim 2, characterized in that: In the change type identification step, the identification of new spots and changed spots based on the attribute information and graphic information of the spots includes: Newly added spot recognition: Compare the attribute information of the currently edited patch data with the historical land use data. If the patch number of the current patch does not exist in the historical land use data, it is marked as a newly added patch. Change pattern recognition: Compare the graphics and attribute information of the currently edited patch data with the historical land use data. If the patch number of the current patch exists in the historical land use data, but the patch has changed in shape or attribute, it will be marked as a changed patch.

7. The method for recording the changing nature of illegal spots in satellite images according to claim 6, characterized in that: When identifying the changed spots, compare the historical spots with the same spot number as the current spot, and quickly get the result of the change of attribute information by comparing the field values. If the attribute information changes, it is marked as a changed spot and the graphic information is no longer compared. If the attribute information has not changed, the graphic information indicators are further compared to see if they have changed. If so, they are identified as changed spots. The graphic information indicators include the coordinates of the center point of the graphic, the area, the perimeter, and the coordinate sequence of the inflection point of the graphic boundary.

8. The method for recording the changing nature of illegal spots in satellite images according to claim 1, characterized in that: In the change property analysis step, the judgment rules set include: When the properties change but the form does not: When the boundary of the patch has not changed, but only the attribute has changed, the nature of the change is judged according to the attribute change: if the original patch is called cultivated land and the current patch is called non-cultivated land, then the nature of the patch change is the conversion of cultivated land to non-agricultural land; if the original patch is called cultivated land and the current patch is still called cultivated land, but the planting attribute has changed from the original staple food crop to non-staple food crop, then the nature of the current patch change is the conversion of cultivated land to non-grain crops; if the current patch has changed in attribute but not in form compared to the original patch, and does not meet the above conditions of the conversion of cultivated land to non-agricultural land or the conversion of cultivated land to non-grain crops, then the nature of the current patch change is other; The situation where the form changes but the properties do not change: Since it does not involve changes in the attributes of the spots, the nature of the current changes in the spots is other; When attribute changes are accompanied by morphological changes: When the current spot is formed by merging multiple original spots, the first sub-rule is used for judgment and analysis; when the current spot intersects with multiple original spots, the second sub-rule is used for judgment and analysis; when the current spot is formed by splitting the original spots, the third sub-rule is used for judgment and analysis.

9. The method for recording the changing nature of illegal spots in satellite images according to claim 8, characterized in that: The first sub-rule includes: No arable land merger: If the land type names of the original multiple map blocks before the merger do not include cultivated land, the changed nature of the current map block is other; Including arable land merger: When the land class names of the original multiple map blocks before the merger contain cultivated land: if the land class name of the current map block is not cultivated land, the nature of the change is the conversion of cultivated land to non-agricultural land; if the land class name of the current map block is still cultivated land, and the planting attribute of the original cultivated map block contains staple crops and the planting attribute of the current map block has changed to non-staple crops, then the nature of the change of the current map block is the conversion of cultivated land to non-staple crops; if the land class name of the current map block is still cultivated land, and the planting attribute of the map block has not changed from staple crops to non-staple crops, then the nature of the change of the current map block is other; The second sub-rule includes: Extract the overlapping area between the current image spot and the original multiple image spots; When the overlapping area does not contain cultivated land, the change nature of the current map patch is other; When the overlapping area contains cultivated land: if the current map spot is not cultivated land, the nature of the map spot change is cultivated land non-agriculturalization; if the current map spot is still cultivated land, and the planting attributes of the original cultivated map spot in the overlapping area include staple crops and the planting attributes of the current map spot are changed to non-staple crops, then the nature of the change of the current map spot is cultivated land non-grainization; if the current map spot is still cultivated land, and the planting attributes of the map spot have not changed from staple crops to non-staple crops, then the nature of the change of the current map spot is other; The third sub-rule includes: If the land type name of the original map patch is not cultivated land, the changed nature of the current map patch is other; When the land type name of the original map spot is cultivated land: if the current map spot is not cultivated land, the nature of the change is the non-agriculturalization of cultivated land; if the current map spot is still cultivated land, and the planting attribute of the original map spot cultivated land is staple food crops, and the planting attribute of the current map spot cultivated land is non-staple food crops, then the nature of the change of the current map spot is the non-grainization of cultivated land; if the current map spot is still cultivated land, and the planting attribute of the map spot has not changed from staple food crops to non-staple food crops, then the nature of the change of the current map spot is other.

10. A system, characterized in that It comprises a processor and a memory, wherein the memory stores a plurality of instructions; the processor loads instructions from the memory to execute the method for recording the changing properties of illegal spots in satellite images as described in any one of claims 1 to 9.

Images