A system and method for collecting and processing national land space planning data

Abstract

The present invention discloses a system and method for collecting and processing data for national land space planning, and specifically relates to the technical field of data processing. The system improves the detail accuracy of low-resolution satellite images in national land space planning through a data processing and analysis method for multi-resolution remote sensing images, obtains image data of different resolutions from different satellites according to needs, and evaluates the detail fidelity and captures the fluctuation state of low-resolution images through image segmentation and feature ratio analysis. After comprehensive analysis, the images are divided into high, medium and low accuracy levels, and abnormal prediction, data correction and optimization decision-making are performed for medium-accuracy images. This can not only effectively reduce the cumulative errors caused by low-resolution images, but also enhance the accuracy and long-term feasibility of planning, thereby ensuring the scientificity and practicality of national land space planning.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a system and method for collecting and processing land space planning data. Background Art

[0002] The data collection and processing of national land space planning refers to the process of acquiring, sorting and analyzing data such as geographic information, ecological resources, and infrastructure related to national land space through technical means and tools. Its purpose is to provide a scientific basis for national land space planning to ensure the reasonable layout of land use, ecological protection, urban development, etc. The means of data collection include remote sensing, satellite images, drone photography, field surveys, etc., while the processing stage involves data classification, cleaning, storage and analysis to generate information with decision-making reference value.

[0003] In specific national land space planning, the collection and processing process also needs to be connected with other information systems. In existing technologies, various spatial data can be integrated to build a complete national land space information database by using geographic information systems (GIS) and remote sensing technology. However, the resolution of remote sensing data directly affects the accuracy and practicality of the data. And lower-resolution satellite images may not accurately reflect surface details, especially in urban planning or ecological assessments that require high precision. For example, some satellite images may only be able to distinguish objects over 10 meters, resulting in small buildings, roads, vegetation, etc. cannot be accurately identified. At the same time, if lower-resolution data is used continuously, more errors will accumulate in long-term planning. Decisions at each stage may be based on data from the previous stage, and the lack of accuracy of these data will cause errors to gradually amplify. This cumulative error may cause national land space planning to be completely out of touch with reality after many years. Long-term accumulated errors will not only increase the cost of later corrections, but may also lead to systematic errors in the decision-making chain, seriously weakening the scientific nature and enforceability of the entire national land space planning system. Summary of the invention

[0004] The purpose of the present invention is to provide a national land space planning data acquisition and processing system and method to overcome the deficiencies in the background technology.

[0005] In order to achieve the above object, the present invention provides the following technical solution: a method for collecting and processing land space planning data, comprising the following steps:

[0006] S1: According to the application requirements of national land space planning, image data of different resolutions are obtained from different satellites, and the image data of different resolutions are divided into high-resolution images, medium-resolution images and low-resolution images according to the range of the surface area represented by each pixel in the image data;

[0007] S2: In images of different resolutions, the boundaries of objects are identified through image segmentation technology, the feature ratios in low-resolution images that match those in high-resolution images are calculated, and after analyzing the deviation changes in feature ratios, the detail fidelity of low-resolution images at different scales is evaluated;

[0008] S3: Compare the multi-resolution images of high detail fidelity and low detail fidelity areas, analyze the performance differences of low-resolution images in high-fidelity areas and low-fidelity areas, and determine the fluctuation state of the low-resolution image's ability to capture details when the detail fidelity changes;

[0009] S4: Comprehensively analyze the detail fidelity of low-resolution images at different scales and the fluctuation of the detail capture ability of low-resolution images, and determine the accuracy of low-resolution satellite images in reflecting surface details based on the analysis results;

[0010] S5: classify the accuracy of low-resolution satellite images in reflecting surface details into different levels, and classify the low-resolution satellite images into high-accuracy satellite images, medium-accuracy satellite images and low-accuracy satellite images, and perform corresponding processing;

[0011] S6: For medium-accuracy satellite images, predict the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period, and perform image data correction and decision optimization based on the prediction results.

[0012] Preferably, in S2, a ground feature proportion deviation anomaly index is generated according to the characteristic proportion deviation of different ground features, and the ground feature proportion deviation anomaly index is obtained by:

[0013] Determine the window size W, which represents the number of feature ratio deviation data points contained in the time period, set the moving step S, which represents the interval at which the window moves in the time series, and calculate the feature ratio difference between the low-resolution image and the high-resolution image at time t. , the expression is: ;in, Indicates the area of ​​the object in the low-resolution image. Represents the area of ​​the object in the high-resolution image. For each window, the mean of the feature ratio difference within the window is calculated. and standard deviation ; Set the standard fluctuation range of characteristic ratio difference under normal conditions according to historical data: ,in, ; ; In the formula, k is a constant. By calculating the absolute change of the deviation within the window, the abnormal index of the proportional deviation of the feature is calculated. The expression is: ; In the formula, It is the abnormal index of the proportion deviation of land features.

[0014] Preferably, in S2, the acquired ground feature proportion deviation anomaly index is compared with a preset ground feature proportion deviation anomaly index reference threshold; if the ground feature proportion deviation anomaly index is greater than or equal to the ground feature proportion deviation anomaly index reference threshold, it indicates that the ground feature proportion deviation anomaly degree is high, that is, the detail fidelity of the low-resolution image at different scales is low, and the corresponding area is divided into a low detail fidelity area; if the ground feature proportion deviation anomaly index is less than the ground feature proportion deviation anomaly index reference threshold, it indicates that the ground feature proportion deviation anomaly degree is low, that is, the detail fidelity of the low-resolution image at different scales is high, and the corresponding area is divided into a high detail fidelity area.

[0015] Preferably, in S3, after analyzing the detail capture capabilities of the high detail fidelity region and the low detail fidelity region at different times, a boundary complexity fluctuation index is generated, and the method for obtaining the boundary complexity fluctuation index is:

[0016] Perform image segmentation on images at different times to identify the boundaries of objects, extract the two-dimensional coordinate sequence of the boundaries of objects, and represent the boundary contour points as a set of coordinate points in two-dimensional space. ; Divide the high detail fidelity area and the low detail fidelity area into A square grid, for each scale , calculate the minimum number of boxes that can cover the boundary contour , gradually reducing the side length of the box , record the number of boxes at each scale By calculating different The value corresponds to The slope is obtained by regression fitting, which is the fractal dimension. , on a logarithmic scale, the fractal dimension Defined as: ; For each time point s, calculate the corresponding fractal dimension , and obtain the fractal dimension sequence at different times ; Where T is the length of the time series, which represents the fractal dimension at different times; the boundary complex fluctuation index is calculated based on the fractal dimension sequence at different times, and the expression is: ; Where BC is the boundary complex fluctuation index.

[0017] Preferably, in S4, the detail fidelity of the low-resolution image at different scales and the fluctuation state of the detail capturing ability of the low-resolution image are comprehensively analyzed, and the accuracy of the low-resolution satellite image in reflecting the surface details is determined according to the analysis results, specifically:

[0018] The feature proportion deviation anomaly index and the boundary complex fluctuation index are converted into the first eigenvector, and the first eigenvector is used as the input of the machine learning model. The machine learning model predicts the accuracy value label of the surface details reflected by the low-resolution satellite image for each group of first eigenvectors as the prediction target, and minimizes the sum of the prediction errors of the accuracy value labels of the surface details reflected by all low-resolution satellite images as the training target. The machine learning model is trained until the sum of the prediction errors converges and the model training is stopped. The accuracy value of the surface details reflected by the low-resolution satellite image is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0019] Preferably, in S5, the accuracy of the low-resolution satellite images in reflecting the surface details is divided into different levels, and the low-resolution satellite images are divided into high-accuracy satellite images, medium-accuracy satellite images and low-accuracy satellite images, and corresponding processing is performed, specifically:

[0020] Comparing the accuracy value of the acquired low-resolution satellite image reflecting the surface details with the gradient standard threshold, the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold, and comparing the accuracy value of the low-resolution satellite image reflecting the surface details with the first standard threshold and the second standard threshold respectively;

[0021] If the accuracy value of the low-resolution satellite image in reflecting the surface details is greater than the second standard threshold, it means that the low-resolution satellite image has high accuracy in reflecting the surface details. In this case, a high-accuracy signal is generated, and the low-resolution satellite image is classified as a high-accuracy satellite image. The high-accuracy satellite image is directly used for high-precision land space planning and fine regional analysis.

[0022] If the accuracy value of the low-resolution satellite image in reflecting the surface details is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it means that the accuracy of the low-resolution satellite image in reflecting the surface details is average. In this case, a medium-accuracy signal is generated, and the low-resolution satellite image is classified as a medium-accuracy satellite image. Further analysis is required on the accuracy of the medium-accuracy satellite image in reflecting the surface details.

[0023] If the accuracy value of the low-resolution satellite image in reflecting the surface details is less than the first standard threshold, it means that the low-resolution satellite image has low accuracy in reflecting the surface details. At this time, a low-accuracy signal is generated, and the low-resolution satellite image is classified as a low-accuracy satellite image. The low-accuracy satellite image is used for rough global analysis.

[0024] Preferably, in S6, for medium-accuracy satellite images, the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period is predicted, specifically:

[0025] For medium-accuracy satellite images, the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period is predicted, and the expression is: ; In the formula, Indicates at time The detail accuracy value predicted at time h, A(h) is the detail accuracy value at the current time h, is the detail accuracy value of the first p time steps, is the autoregressive coefficient, which indicates the weight of historical data on the current prediction. As the error term, the anomaly index of the accuracy of surface details reflected by low-resolution satellite images within a fixed time period is calculated, and the expression is: ; In the formula, For time The abnormal index at .

[0026] Preferably, the acquired anomaly index is compared with an anomaly index reference threshold. If the anomaly index is greater than or equal to the anomaly index reference threshold, it means that the satellite image reflects a high degree of anomaly in the accuracy of surface details, and a warning signal is generated at this time; if the anomaly index is less than the anomaly index reference threshold, it means that the satellite image reflects a low degree of anomaly in the accuracy of surface details, and no warning signal is generated at this time.

[0027] Preferably, after identifying the abnormal situation in the medium-accuracy satellite image, the image data is corrected. The image correction is achieved by weighted fusion of image data from different sources. The correction formula is: ; In the formula, is the corrected detail accuracy value at time h+k, The detail accuracy value of low-resolution images, is the detail accuracy value of the medium-resolution image, is the detail accuracy value of the high-resolution image, is the weight coefficient.

[0028] The present invention also provides a national land space planning data acquisition and processing system, including an image data acquisition module, a fidelity analysis module, a multi-scale comparison module, an accuracy prediction module, an image classification module and a correction optimization module;

[0029] Image data acquisition module: According to the application requirements of national land space planning, image data of different resolutions are obtained from different satellites, and the image data of different resolutions are divided into high-resolution images, medium-resolution images and low-resolution images according to the range of surface area represented by each pixel in the image data;

[0030] Fidelity analysis module: In images of different resolutions, the boundaries of objects are identified through image segmentation technology, the feature ratio of low-resolution images that matches the high-resolution images is calculated, and after analyzing the deviation changes of the feature ratio, the fidelity of details of low-resolution images at different scales is evaluated;

[0031] Multi-scale comparison module: compares multi-resolution images of high-detail fidelity and low-detail fidelity areas, analyzes the performance differences of low-resolution images in high-fidelity areas and low-fidelity areas, and determines the fluctuation state of low-resolution images' ability to capture details when detail fidelity changes;

[0032] Accuracy prediction module: Comprehensively analyze the detail fidelity of low-resolution images at different scales and the fluctuation of the detail capture ability of low-resolution images, and determine the accuracy of low-resolution satellite images in reflecting surface details based on the analysis results;

[0033] Image classification module: divides low-resolution satellite images into different levels according to the accuracy of reflecting surface details, and divides low-resolution satellite images into high-accuracy satellite images, medium-accuracy satellite images and low-accuracy satellite images, and performs corresponding processing;

[0034] Correction and Optimization Module: For medium-accuracy satellite images, the module predicts the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period, and performs image data correction and decision optimization based on the prediction results.

[0035] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0036] 1. In the national land space planning, the present invention combines the acquisition, analysis and processing process of images with different resolutions. The present invention introduces the ground feature ratio deviation anomaly index and the boundary complex fluctuation index to quantify the detail fidelity of the image. Through the comprehensive analysis of the image data by the machine learning model, the detail performance of low-resolution images in different scenes can be accurately predicted, and the images can be divided into three levels of high, medium and low according to the accuracy, so as to ensure the use of suitable image data in high-precision planning scenes and reduce error accumulation.

[0037] 2. The present invention provides a method for anomaly detection and image data correction for images of medium accuracy, and utilizes the fusion of multi-source images to improve the accuracy of image data. By predicting the anomaly index through an autoregressive model and performing weighted correction on the image data, the accuracy of low-resolution images in reflecting surface details can be effectively improved. Through intelligent analysis and correction processes, the present invention not only improves the accuracy and reliability of national land space planning, but also reduces the risk of error accumulation caused by long-term use of low-precision data, significantly improving the scientificity and executability of the planning system. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0039] Figure 1 The figure is a flow chart of the method of the present invention.

[0040] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Example 1, please refer to Figure 1 and Figure 2 As shown, the method for collecting and processing land space planning data described in this embodiment includes the following steps:

[0043] S1: According to the application requirements of national land space planning, image data of different resolutions are obtained from different satellites, and the image data of different resolutions are divided into high-resolution images, medium-resolution images and low-resolution images according to the range of the surface area represented by each pixel in the image data;

[0044] S2: In images of different resolutions, the boundaries of objects are identified through image segmentation technology, the feature ratios in low-resolution images that match those in high-resolution images are calculated, and after analyzing the deviation changes in feature ratios, the detail fidelity of low-resolution images at different scales is evaluated;

[0045] S3: Compare the multi-resolution images of high detail fidelity and low detail fidelity areas, analyze the performance differences of low-resolution images in high-fidelity areas and low-fidelity areas, and determine the fluctuation state of the low-resolution image's ability to capture details when the detail fidelity changes;

[0046] S4: Comprehensively analyze the detail fidelity of low-resolution images at different scales and the fluctuation of the detail capture ability of low-resolution images, and determine the accuracy of low-resolution satellite images in reflecting surface details based on the analysis results;

[0047] S5: classify the accuracy of low-resolution satellite images in reflecting surface details into different levels, and classify the low-resolution satellite images into high-accuracy satellite images, medium-accuracy satellite images and low-accuracy satellite images, and perform corresponding processing;

[0048] S6: For medium-accuracy satellite images, predict the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period, and perform image data correction and decision optimization based on the prediction results.

[0049] Among them, in S1, according to the application requirements of national land space planning, image data of different resolutions are obtained from different satellites, and according to the range of the surface area represented by each pixel in the image data of different resolutions, they are divided into high-resolution images, medium-resolution images and low-resolution images, specifically:

[0050] Determine the resolution requirements of the required image data based on the specific needs of national land space planning. Different application scenarios (such as urban planning, ecological protection, infrastructure construction, etc.) have different requirements for image resolution.

[0051] High-resolution images: Suitable for scenes with high detail requirements, such as urban planning, building identification, and road network design. The pixel resolution is generally between 0.3 meters and 2 meters.

[0052] Medium-resolution imagery: used for medium-scale analysis, such as land use classification and ecological assessment, with a resolution between 2 and 30 meters.

[0053] Low-resolution images: used for large-scale environmental monitoring and climate change research, with a resolution generally above 30 meters.

[0054] Image resolution standards are as follows: High-resolution images: 0.3 meters to 2 meters. Medium-resolution images: 2 meters to 30 meters. Low-resolution images: above 30 meters.

[0055] According to the above image requirements, data with appropriate resolutions can be obtained from different satellites. The following are some commonly used satellites and their resolution ranges:

[0056] High-resolution satellites: WorldView-3, GeoEye-1: Resolution 0.3 to 1 meter, for detailed urban and infrastructure analysis.

[0057] QuickBird, Pleiades: 1-meter resolution, used for accurate identification of buildings, roads, etc.

[0058] Medium-resolution satellite: Sentinel-2: with a resolution of 10 meters, used for large-scale land use classification and environmental monitoring.

[0059] Landsat 8: With a resolution of 30 meters, it is widely used for large-scale ecological monitoring and land use change analysis.

[0060] Low-resolution satellites: MODIS (Aqua / Terra): resolution 250 meters to 1000 meters, mainly used for large-scale ecological monitoring and climate research.

[0061] NOAA AVHRR: The resolution is about 1 km, suitable for large-scale environmental monitoring.

[0062] Download images from various online satellite data platforms. For example: High-resolution data: Commercial satellite image providers (such as DigitalGlobe) or National Satellite Data Center. Medium-resolution data: Copernicus program (Sentinel data) of the European Space Agency (ESA), Landsat program of the United States Geological Survey (USGS). Low-resolution data: MODIS data provided by NASA, downloaded through LAADS DAAC or Earthdata platform. When downloading, usually choose common GIS and remote sensing formats, such as GeoTIFF, HDF or JPEG2000, which are compatible with most GIS and remote sensing software.

[0063] After acquiring images of different resolutions, the data needs to be preprocessed to ensure that they can be used for further analysis. The preprocessing steps include:

[0064] Geometric correction: Perform geometric correction on the image to remove geometric distortion in satellite imagery, ensuring that the pixels in the image correctly correspond to the geographic coordinate system, usually using a global standard coordinate system (such as WGS 84).

[0065] Atmospheric correction: Perform atmospheric correction on images to eliminate the interference of the atmosphere on the images and ensure the consistent radiation accuracy of the images. It is applicable to all image resolutions.

[0066] Image cropping: Use GIS tools to crop images according to the specific areas of national land space planning to ensure that the area covered by each image data is consistent, which facilitates subsequent analysis and comparison.

[0067] Based on the surface area (resolution) represented by each image pixel, the image is divided into high resolution, medium resolution and low resolution. The specific steps are as follows:

[0068] Confirm the image resolution: Confirm the surface area corresponding to each pixel in the image metadata. For example, the resolution of WorldView-3 is 0.3 meters, the resolution of Sentinel-2 is 10 meters, and the resolution of MODIS is 250 meters.

[0069] Classification: images of different resolutions are divided into corresponding categories according to pixel size:

[0070] High-resolution images: Images with pixel resolutions between 0.3 meters and 2 meters are classified as high-resolution datasets.

[0071] Medium-resolution images: Images with pixel resolutions between 2 meters and 30 meters are classified as medium-resolution datasets.

[0072] Low-resolution images: Images with a pixel resolution of more than 30 meters are classified as low-resolution datasets.

[0073] Use GIS platforms or spatial databases (such as PostGIS, ArcSDE) to store and manage image data of different resolutions to ensure that they can be easily called and analyzed. Generate metadata for each image data, record key information such as image resolution, acquisition time, sensor type, coordinate system, etc., to facilitate subsequent retrieval and analysis.

[0074] S2: In images of different resolutions, the boundaries of objects are identified through image segmentation technology, the feature ratios in low-resolution images that match those in high-resolution images are calculated, and after analyzing the deviation changes in feature ratios, the detail fidelity of low-resolution images at different scales is evaluated.

[0075] Image segmentation is a key technology for identifying the boundaries of objects (such as buildings, roads, vegetation, etc.) in images. Images of different resolutions use different image segmentation algorithms to ensure that the boundaries of objects can be accurately identified. Commonly used image segmentation algorithms include:

[0076] Threshold segmentation: Applicable to simple image scenes, compares pixel values ​​with a certain threshold to distinguish different ground object areas.

[0077] Region-based segmentation: By aggregating adjacent pixels into regions based on similarity, different ground features in the image can be identified.

[0078] Edge detection algorithms (such as Canny edge detection): used to identify boundaries in an image where pixel values ​​change significantly, and are particularly suitable for detecting ground object boundaries.

[0079] Deep learning segmentation methods: such as U-Net or Mask R-CNN. These algorithms perform well in complex object recognition and high-precision boundary recognition, and are particularly suitable for high-resolution images.

[0080] Perform object segmentation on high-resolution images: Use the selected image segmentation algorithm to segment the high-resolution image to extract clear object boundaries. Since high-resolution images are rich in details, the segmentation accuracy is usually high, and the extracted objects (such as roads, buildings, vegetation, etc.) can be used as standard references.

[0081] Segmentation on low-resolution images: Use the same or adjusted segmentation algorithm to segment objects on low-resolution images. In low-resolution images, the boundaries of objects may not be as clear as those in high-resolution images, and the segmentation results may be a little blurry or distorted.

[0082] Matching of ground features: Find the same ground features in high-resolution and low-resolution images and match them. Usually, obvious ground features such as roads, buildings, and vegetation are selected. Use the registration tools or image processing tools in the GIS platform (such as ArcGIS, ENVI, etc.) to geo-register the two resolution images to ensure that they are aligned in the same coordinate system and spatial framework.

[0083] Matching based on shape and location: Identify the same object in high-resolution and low-resolution images based on its shape, size, and location characteristics.

[0084] Feature point matching algorithms: such as SIFT (Scale Invariant Feature Transform) and SURF (Speeded Up Robust Features), can automatically identify significant feature points in images and match them in images of different resolutions.

[0085] Calculate feature ratio: Calculate the feature ratio of the same feature in the low-resolution image and the high-resolution image by using the features of the feature such as area, shape, and boundary length. For example, if the area of ​​a building in the high-resolution image is A1 and the area of ​​the same building in the low-resolution image is A2, then the feature ratio is A2 / A1. Compare the feature ratios of the same feature in the high-resolution image and the low-resolution image, and calculate the deviation between the two. The smaller the deviation, the better the low-resolution image can preserve the details of the feature; the larger the deviation, the lower the fidelity of the detail in the low-resolution image. Calculate the difference in the area of ​​the feature in the high-resolution image and the low-resolution image. Map the deviation results to the spatial distribution map and mark the areas with large deviations. This helps identify which areas have serious detail loss in the low-resolution image, such as complex urban structures, densely built-up areas, and areas with high vegetation diversity. The error heat map can intuitively show which areas in the low-resolution image have poor detail fidelity.

[0086] The feature proportion deviation anomaly index of the feature is generated according to the feature proportion deviation of different features. The method for obtaining the feature proportion deviation anomaly index of the feature is as follows:

[0087] Determine the window size W, which represents the number of feature ratio deviation data points included in the time period. For example, assuming that each data point represents a feature ratio deviation for one month, then choosing W=3 means that we will consider 3 months of data in each window. Set the moving step size S, which represents the interval at which the window moves on the time series. The typical choice is S=1, which means moving one data point at a time. The smaller the step size, the more detailed the changes detected, and the larger the step size, the higher the computational efficiency, but some subtle changes may be missed.

[0088] Calculate the difference between the feature ratio in the low-resolution image and the feature ratio in the high-resolution image at time t , the expression is: ;in, Indicates the area of ​​the object in the low-resolution image. Indicates the area of ​​the feature in the high-resolution image. The closer the ratio is to 1, the higher the match between the low-resolution image and the high-resolution image. For each window, calculate the mean of the feature ratio difference within the window and standard deviation ; Set the standard fluctuation range of characteristic ratio difference under normal conditions according to historical data: ,in, ; ; In the formula, k is a constant used to control the sensitivity of detection, usually between 2 and 3; by calculating the absolute change of the deviation within the window, the abnormal index of the proportional deviation of the feature is calculated, and the expression is: ; In the formula, It is the abnormal index of the proportion deviation of land features.

[0089] The acquired ground feature proportion deviation anomaly index is compared with the preset ground feature proportion deviation anomaly index reference threshold. If the ground feature proportion deviation anomaly index is greater than or equal to the ground feature proportion deviation anomaly index reference threshold, it means that the ground feature proportion deviation anomaly degree is high, that is, the detail fidelity of the low-resolution image at different scales is low, and the corresponding area is divided into a low detail fidelity area; if the ground feature proportion deviation anomaly index is less than the ground feature proportion deviation anomaly index reference threshold, it means that the ground feature proportion deviation anomaly degree is low, that is, the detail fidelity of the low-resolution image at different scales is high, and the corresponding area is divided into a high detail fidelity area.

[0090] S3: Compare the multi-resolution images of high detail fidelity and low detail fidelity areas, analyze the performance differences of low-resolution images in high-fidelity areas and low-fidelity areas, and determine the fluctuation state of the low-resolution image's ability to capture details when the detail fidelity changes.

[0091] The high-resolution and low-resolution images are geometrically corrected and registered to ensure that they are aligned in the same coordinate system. This step is to ensure that the images are strictly aligned in space so that they can be accurately compared in the same area. Registration can be performed using GIS tools (such as ArcGIS, QGIS) or image processing software (such as ENVI, ERDAS).

[0092] For high detail fidelity areas and low detail fidelity areas, mark the compared areas on the image. You can select key areas (such as cities, mountains, forests, rivers, etc.) based on previous analysis or according to the needs of specific application scenarios. Mark these areas on the image in the form of polygons or grids to ensure that high-resolution and low-resolution images are compared in the same area. Compare the features in high-resolution and low-resolution images. Complex boundaries are usually simplified or blurred in low-resolution images, so by comparing the length or shape of the boundaries, you can intuitively assess the loss of details. For example, the high complexity expression is : ; The low complexity expression is: ;if If it is close to 1, it means that the boundary complexity changes little and the details are captured better; if it is much smaller than 1, it means that the boundary is greatly simplified in the low-resolution image.

[0093] In multiple regions, the changes in detail fidelity at different time points are calculated and compared. By analyzing the detail capture capabilities of high-detail fidelity areas and low-detail fidelity areas at different times, the fluctuation state of low-resolution images at different times or under different surface conditions can be found. If detail fidelity continues to decrease over a period of time, it indicates that the detail capture capabilities of low-resolution images are deteriorating.

[0094] The boundary complexity fluctuation index is generated by analyzing the detail capture capabilities of high detail fidelity areas and low detail fidelity areas at different times. The method for obtaining the boundary complexity fluctuation index is:

[0095] Perform image segmentation on images at different times (such as using Canny edge detection, threshold segmentation and other algorithms) to identify the boundaries of objects. Extract the two-dimensional coordinate sequence of the object boundary (i.e., boundary contour), and represent the boundary contour points as a set of coordinate points in two-dimensional space. ; Ensure that the boundary coordinates of different time periods are in the same coordinate system. Divide the high-detail fidelity area and the low-detail fidelity area into A square grid, for each scale , calculate the minimum number of boxes that can cover the boundary contour , gradually reducing the side length of the box , record the number of boxes at each scale By calculating different The value corresponds to The slope is obtained by regression fitting, which is the fractal dimension. , on a logarithmic scale, the fractal dimension Defined as: ; For each time point s, calculate the corresponding fractal dimension , and obtain the fractal dimension sequence at different times ; Where T is the length of the time series, which represents the fractal dimension at different times; the boundary complex fluctuation index is calculated based on the fractal dimension sequence at different times, and the expression is: ; Where BC is the boundary complex fluctuation index.

[0096] When the boundary complexity fluctuation index is large, it indicates that the low-resolution image has a large fluctuation in its ability to capture details. Specifically, the complexity of the boundary changes significantly at different times. This may be because the image resolution is too low to stably capture the shape of complex objects, resulting in inconsistent presentation of the boundaries of objects in time. Such large fluctuations usually occur in complex terrain areas, densely built-up urban areas, or changeable natural environments (such as forests, rivers, etc.). A high fluctuation index means that low-resolution images may miss key details at certain times, or may not accurately capture objects, especially when the boundaries of objects change significantly, the image may appear distorted or blurred.

[0097] When the boundary complexity fluctuation index is small, it indicates that the ability of low-resolution images to capture details is relatively stable, and the captured boundaries of objects change little between different times. A smaller fluctuation index indicates that low-resolution images can maintain high fidelity of details for simple or regular objects over a large range, which usually occurs in relatively homogeneous areas such as farmland, grassland, and plains. In these areas, the shape of objects changes more slowly, so low-resolution images can still capture the boundaries of objects stably despite insufficient detail resolution. A lower fluctuation index means that the details of the image are more consistent, and the boundary shape has basically not changed dramatically, making it suitable for long-term stable monitoring or large-scale environmental analysis.

[0098] S4: Comprehensively analyze the detail fidelity of low-resolution images at different scales and the fluctuation of their detail capture capabilities, and determine the accuracy of low-resolution satellite images in reflecting surface details based on the analysis results.

[0099] The feature proportion deviation anomaly index and the boundary complex fluctuation index are converted into the first eigenvector, and the first eigenvector is used as the input of the machine learning model. The machine learning model predicts the accuracy value label of the surface details reflected by the low-resolution satellite image for each group of first eigenvectors as the prediction target, and minimizes the sum of the prediction errors of the accuracy value labels of the surface details reflected by all low-resolution satellite images as the training target. The machine learning model is trained until the sum of the prediction errors converges and the model training is stopped. The accuracy value of the surface details reflected by the low-resolution satellite image is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0100] The method for obtaining the accuracy value of the surface details reflected by the low-resolution satellite image is as follows: from the first eigenvector training data of the trained machine learning model, the corresponding function expression is obtained: ; In the formula, is the output function of the model, is the abnormal index of the proportion deviation of the ground feature, BC is the boundary complex fluctuation index, The accuracy value of reflecting surface details in low-resolution satellite images.

[0101] S5: The accuracy of low-resolution satellite images in reflecting surface details is divided into different levels, and the low-resolution satellite images are divided into high-accuracy satellite images, medium-accuracy satellite images and low-accuracy satellite images, and are processed accordingly.

[0102] Comparing the accuracy value of the acquired low-resolution satellite image reflecting the surface details with the gradient standard threshold, the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold, and comparing the accuracy value of the low-resolution satellite image reflecting the surface details with the first standard threshold and the second standard threshold respectively;

[0103] If the accuracy value of the low-resolution satellite image in reflecting the surface details is greater than the second standard threshold, it means that the low-resolution satellite image has high accuracy in reflecting the surface details. In this case, a high-accuracy signal is generated, and the low-resolution satellite image is classified as a high-accuracy satellite image. The high-accuracy satellite image is directly used for high-precision land space planning and fine regional analysis.

[0104] If the accuracy value of the low-resolution satellite image in reflecting the surface details is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it means that the accuracy of the low-resolution satellite image in reflecting the surface details is average. In this case, a medium-accuracy signal is generated, and the low-resolution satellite image is classified as a medium-accuracy satellite image. Further analysis is required on the accuracy of the medium-accuracy satellite image in reflecting the surface details.

[0105] If the accuracy value of the low-resolution satellite image in reflecting the surface details is less than the first standard threshold, it means that the low-resolution satellite image has low accuracy in reflecting the surface details. At this time, a low-accuracy signal is generated, and the low-resolution satellite image is classified as a low-accuracy satellite image. The low-accuracy satellite image is used for rough global analysis, such as large-scale environmental monitoring or trend analysis.

[0106] S6: For medium-accuracy satellite images, predict the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period, and perform image data correction and decision optimization based on the prediction results.

[0107] For medium-accuracy satellite images, the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period is predicted, and the expression is: ; In the formula, Indicates at time The detail accuracy value predicted at time h, A(h) is the detail accuracy value at the current time h, is the detail accuracy value of the first p time steps, is the autoregressive coefficient, which indicates the weight of historical data on the current prediction. As the error term, the anomaly index of the accuracy of surface details reflected by low-resolution satellite images within a fixed time period is calculated, and the expression is: ; In the formula, For time The abnormal index at .

[0108] The obtained anomaly index is compared with the anomaly index reference threshold. If the anomaly index is greater than or equal to the anomaly index reference threshold, it means that the satellite image has a high degree of anomaly in reflecting the accuracy of surface details, and an early warning signal is generated at this time; if the anomaly index is less than the anomaly index reference threshold, it means that the satellite image has a low degree of anomaly in reflecting the accuracy of surface details, and no early warning signal is generated at this time.

[0109] After identifying anomalies in medium-resolution satellite images, the image data is corrected. The goal of the correction is to improve the detail capture ability of low-resolution images by fusing other high-resolution or medium-resolution image data.

[0110] Image correction is achieved by weighted fusion of image data from different sources. The correction formula is: ; In the formula, is the corrected detail accuracy value at time h+k, The detail accuracy value of low-resolution images, is the detail accuracy value of the medium-resolution image, is the detail accuracy value of the high-resolution image, is the weight coefficient, which is used to adjust the contribution of images with different resolutions to the correction results. is the weight of low-resolution data, which is usually lower because of its poor accuracy. is the weight of the medium-resolution data, if available, which is usually between the high-resolution and low-resolution data. is the weight of high-resolution data, which usually dominates due to its high accuracy.

[0111] After processing the data correction, the corrected image is analyzed to decide how to adjust the subsequent decision. The optimization process is expressed as: ; In the formula, To optimize the decision-making plan before is the optimized decision-making plan, H is the total optimization time, is the corrected detail accuracy value, for true detail accuracy (e.g. obtained through ground verification or high-resolution imagery), Is the loss function, which is used to measure the gap between the corrected data and the real data. The mean square error is usually used, and the expression is: ; By minimizing the loss function L, the subsequent decision-making process can be optimized to ensure accurate judgments based on the corrected data.

[0112] In this application, the degree of abnormality of medium-accuracy satellite images is predicted, and image data correction is performed based on this, ultimately optimizing subsequent decisions. The autoregressive model is used to predict abnormal fluctuations, correct the accuracy of details in low-resolution images, and finally the decision optimization model is used to ensure accurate decisions in different scenarios.

[0113] In this embodiment, according to the needs of national land space planning, image data of different resolutions are obtained from different satellites, and divided into high-resolution, medium-resolution and low-resolution images according to the resolution range. Then, the image segmentation technology is used to identify the boundary of the object, the deviation of the feature ratio of the low-resolution image and the high-resolution image is calculated, and the detail fidelity of the low-resolution image at different scales is analyzed. Then, by comparing the multi-resolution images of the high detail fidelity and low detail fidelity areas, the fluctuation state of the low-resolution image's ability to capture the details of the object is evaluated. On this basis, the detail fidelity and fluctuation state of the low-resolution image are comprehensively analyzed to determine its accuracy in reflecting the surface details. According to the analysis results, the detail accuracy of the image is divided into three levels: high, medium and low, and corresponding processing is performed. For the medium-accuracy image, the abnormal degree of its detail accuracy within a fixed time period is further predicted, and image data correction and decision optimization are performed based on the prediction results to improve the application accuracy of the image.

[0114] Embodiment 2, a land space planning data acquisition and processing system described in this embodiment includes an image data acquisition module, a fidelity analysis module, a multi-scale comparison module, an accuracy prediction module, an image classification module and a correction optimization module;

[0115] Image data acquisition module: According to the application requirements of national land space planning, image data of different resolutions are obtained from different satellites, and the image data of different resolutions are divided into high-resolution images, medium-resolution images and low-resolution images according to the range of surface area represented by each pixel in the image data;

[0116] Fidelity analysis module: In images of different resolutions, the boundaries of objects are identified through image segmentation technology, the feature ratio of low-resolution images that matches the high-resolution images is calculated, and after analyzing the deviation changes of the feature ratio, the fidelity of details of low-resolution images at different scales is evaluated;

[0117] Multi-scale comparison module: compares multi-resolution images of high-detail fidelity and low-detail fidelity areas, analyzes the performance differences of low-resolution images in high-fidelity areas and low-fidelity areas, and determines the fluctuation state of low-resolution images' ability to capture details when detail fidelity changes;

[0118] Accuracy prediction module: Comprehensively analyze the detail fidelity of low-resolution images at different scales and the fluctuation of the detail capture ability of low-resolution images, and determine the accuracy of low-resolution satellite images in reflecting surface details based on the analysis results;

[0119] Image classification module: divides low-resolution satellite images into different levels according to the accuracy of reflecting surface details, and divides low-resolution satellite images into high-accuracy satellite images, medium-accuracy satellite images and low-accuracy satellite images, and performs corresponding processing;

[0120] Correction and Optimization Module: For medium-accuracy satellite images, the module predicts the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period, and performs image data correction and decision optimization based on the prediction results.

[0121] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for collecting and processing land space planning data, characterized in that: The steps include: S1: According to the application requirements of national land space planning, image data of different resolutions are obtained from different satellites, and the image data of different resolutions are divided into high-resolution images, medium-resolution images and low-resolution images according to the range of the surface area represented by each pixel in the image data; S2: In images of different resolutions, the boundaries of objects are identified through image segmentation technology, the feature ratios in low-resolution images that match those in high-resolution images are calculated, and after analyzing the deviation changes in feature ratios, the detail fidelity of low-resolution images at different scales is evaluated; The feature ratio deviation anomaly index of the feature is generated according to the feature ratio deviation of different features. The method for obtaining the feature ratio deviation anomaly index is as follows: determine the window size W, which represents the number of feature ratio deviation data points contained in the time period, set the moving step size S, which represents the interval of the window moving in the time series, and calculate the feature ratio difference ΔP between the feature ratio in the low-resolution image and the feature ratio in the high-resolution image at time t. t , the expression is: Among them, A low (t) represents the area of ​​the object in the low-resolution image, A high (t) represents the area of ​​the object in the high-resolution image. For each window, the mean value μ of the feature ratio difference within the window is calculated. W and standard deviation σ W ; Set the standard fluctuation range of characteristic ratio difference under normal conditions according to historical data: T lower ~T upper , where T upper =μ W +kσ W ; T lower =μ W -kσ W ; In the formula, k is a constant. By calculating the absolute change of the deviation within the window, the abnormal index of the proportional deviation of the feature is calculated. The expression is: Where AQ is the abnormal index of the deviation of the ground feature ratio; The acquired ground feature proportion deviation anomaly index is compared with the preset ground feature proportion deviation anomaly index reference threshold. If the ground feature proportion deviation anomaly index is greater than or equal to the ground feature proportion deviation anomaly index reference threshold, it indicates that the ground feature proportion deviation anomaly degree is high, that is, the detail fidelity of the low-resolution image at different scales is low, and the corresponding area is divided into a low detail fidelity area; if the ground feature proportion deviation anomaly index is less than the ground feature proportion deviation anomaly index reference threshold, it indicates that the ground feature proportion deviation anomaly degree is low, that is, the detail fidelity of the low-resolution image at different scales is high, and the corresponding area is divided into a high detail fidelity area; S3: Compare the multi-resolution images of high detail fidelity and low detail fidelity areas, analyze the performance differences of low-resolution images in high-fidelity areas and low-fidelity areas, and determine the fluctuation state of the low-resolution image's ability to capture details when the detail fidelity changes; The boundary complexity fluctuation index is generated by analyzing the detail capture capabilities of high detail fidelity areas and low detail fidelity areas at different times. The method for obtaining the boundary complexity fluctuation index is as follows: perform image segmentation on images at different times to identify the boundaries of the objects, extract the two-dimensional coordinate sequence of the boundaries of the objects, and represent the boundary contour points as a set of coordinate points in two-dimensional space { ( x1,y 1) , ( x2,y 2) ,..., ( x n ,y n) }; Divide the high-detail fidelity area and the low-detail fidelity area into square grids with a side length of ∈. For each scale ∈, calculate the minimum number of boxes N(∈) that can cover the boundary contour, gradually reduce the side length ∈ of the box, and record the number of boxes N(∈) at each scale; Calculate the N(∈) value corresponding to different ∈ values, and get the slope through regression fitting. The slope is the fractal dimension D f , on a logarithmic scale, the fractal dimension D f Defined as: For each time point s, calculate the corresponding fractal dimension D f( s ) , we get the fractal dimension sequence D at different times f( 1 ) ,D f( 2 ) ,…,D f( T ) ; Where T is the length of the time series, which represents the fractal dimension at different times; the boundary complex fluctuation index is calculated based on the fractal dimension sequence at different times, and the expression is: Where BC is the boundary complex fluctuation index; S4: Comprehensively analyze the detail fidelity of low-resolution images at different scales and the fluctuation of the detail capture ability of low-resolution images, and determine the accuracy of low-resolution satellite images in reflecting surface details based on the analysis results; S5: classify the accuracy of low-resolution satellite images in reflecting surface details into different levels, and classify the low-resolution satellite images into high-accuracy satellite images, medium-accuracy satellite images and low-accuracy satellite images, and perform corresponding processing; S6: For medium-accuracy satellite images, predict the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period, and perform image data correction and decision optimization based on the prediction results; The step S6 specifically includes: for medium-accuracy satellite images, predicting the degree of abnormality of the accuracy of surface details reflected by low-resolution satellite images within a fixed time period, the expression is: In the formula, represents the detail accuracy value predicted at time h+k, A(h) is the detail accuracy value at the current time h, and A ( h-1 ) , …, A ( h-p+1 ) is the detail accuracy value of the first p time steps, α1, α2, …, α p is the autoregressive coefficient, which indicates the weight of historical data on the current prediction, s t As the error term, the anomaly index of the accuracy of surface details reflected by low-resolution satellite images within a fixed time period is calculated, and the expression is: In the formula, E ( h+k ) is the anomaly index at time h+k.

2. A method for collecting and processing land space planning data according to claim 1, characterized in that: In S4, the fidelity of details of low-resolution images at different scales and the fluctuation of the ability of low-resolution images to capture details are comprehensively analyzed. The accuracy of low-resolution satellite images in reflecting surface details is determined based on the analysis results. Specifically: The feature proportion deviation anomaly index and the boundary complex fluctuation index are converted into the first eigenvector, and the first eigenvector is used as the input of the machine learning model. The machine learning model predicts the accuracy value label of the surface details reflected by the low-resolution satellite image for each group of first eigenvectors as the prediction target, and minimizes the sum of the prediction errors of the accuracy value labels of the surface details reflected by all low-resolution satellite images as the training target. The machine learning model is trained until the sum of the prediction errors converges and the model training is stopped. The accuracy value of the surface details reflected by the low-resolution satellite image is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

3. A method for collecting and processing land space planning data according to claim 2, characterized in that: In S5, the accuracy of low-resolution satellite images in reflecting surface details is divided into different levels, and the low-resolution satellite images are divided into high-accuracy satellite images, medium-accuracy satellite images and low-accuracy satellite images, and corresponding processing is performed, specifically: Comparing the accuracy value of the acquired low-resolution satellite image reflecting the surface details with the gradient standard threshold, the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold, and comparing the accuracy value of the low-resolution satellite image reflecting the surface details with the first standard threshold and the second standard threshold respectively; If the accuracy value of the low-resolution satellite image in reflecting the surface details is greater than the second standard threshold, it means that the low-resolution satellite image has high accuracy in reflecting the surface details. In this case, a high-accuracy signal is generated, and the low-resolution satellite image is classified as a high-accuracy satellite image. The high-accuracy satellite image is directly used for high-precision land space planning and fine regional analysis. If the accuracy value of the low-resolution satellite image in reflecting the surface details is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it means that the accuracy of the low-resolution satellite image in reflecting the surface details is average. In this case, a medium-accuracy signal is generated, and the low-resolution satellite image is classified as a medium-accuracy satellite image. Further analysis is required on the accuracy of the medium-accuracy satellite image in reflecting the surface details. If the accuracy value of the low-resolution satellite image in reflecting the surface details is less than the first standard threshold, it means that the low-resolution satellite image has low accuracy in reflecting the surface details. At this time, a low-accuracy signal is generated, and the low-resolution satellite image is classified as a low-accuracy satellite image. The low-accuracy satellite image is used for rough global analysis.

4. The method for collecting and processing land space planning data according to claim 1, characterized in that: The obtained anomaly index is compared with the anomaly index reference threshold. If the anomaly index is greater than or equal to the anomaly index reference threshold, it means that the satellite image has a high degree of anomaly in reflecting the accuracy of surface details, and an early warning signal is generated at this time; if the anomaly index is less than the anomaly index reference threshold, it means that the satellite image has a low degree of anomaly in reflecting the accuracy of surface details, and no early warning signal is generated at this time.

5. A method for collecting and processing land space planning data according to claim 4, characterized in that: After identifying the anomalies in the medium-accuracy satellite image, the image data is corrected. Image correction is achieved by weighted fusion of image data from different sources. The correction formula is: A corrected (h+k)=w1A low (h+k)+w2A mid (h+k)+w3A high (h+k); where A corrected (h+k) is the corrected detail accuracy value. At time h+k, A low (h+k) Detail accuracy value of low-resolution image, A mid (h+k) is the detail accuracy value of the medium-resolution image, A high (h+k) is the detail accuracy value of the high-resolution image, and w1, w2, and w3 are weight coefficients.

6. A land space planning data acquisition and processing system, used to implement a land space planning data acquisition and processing method according to any one of claims 1 to 5, characterized in that: It includes image data acquisition module, fidelity analysis module, multi-scale comparison module, accuracy prediction module, image classification module and correction optimization module; Image data acquisition module: According to the application requirements of national land space planning, image data of different resolutions are obtained from different satellites, and the image data of different resolutions are divided into high-resolution images, medium-resolution images and low-resolution images according to the range of surface area represented by each pixel in the image data; Fidelity analysis module: In images of different resolutions, the boundaries of objects are identified through image segmentation technology, the feature ratio of low-resolution images that matches the high-resolution images is calculated, and after analyzing the deviation changes of the feature ratio, the fidelity of details of low-resolution images at different scales is evaluated; Multi-scale comparison module: compares multi-resolution images of high-detail fidelity and low-detail fidelity areas, analyzes the performance differences of low-resolution images in high-fidelity areas and low-fidelity areas, and determines the fluctuation state of low-resolution images' ability to capture details when detail fidelity changes; Accuracy prediction module: Comprehensively analyze the detail fidelity of low-resolution images at different scales and the fluctuation of the detail capture ability of low-resolution images, and determine the accuracy of low-resolution satellite images in reflecting surface details based on the analysis results; Image classification module: divides low-resolution satellite images into different levels according to the accuracy of reflecting surface details, and divides low-resolution satellite images into high-accuracy satellite images, medium-accuracy satellite images and low-accuracy satellite images, and performs corresponding processing; Correction and Optimization Module: For medium-accuracy satellite images, the module predicts the degree of abnormality in the accuracy of surface details reflected by low-resolution satellite images within a fixed time period, and performs image data correction and decision optimization based on the prediction results.

Images