A method for processing geographic information mapping data based on image analysis

Abstract

The present invention discloses a method for processing geospatial information mapping data based on image analysis, which relates to the technical field of remote sensing image processing. The method includes collecting multi-angle satellite remote sensing images and dividing the multi-angle satellite remote sensing images into several image blocks according to a geographic coordinate system; extracting topographic feature parameters of the image blocks and establishing a feature vector of the topographic feature parameters; establishing a ground object classification model according to the feature vector to classify the image blocks and obtain ground object classification labels; constructing a three-dimensional geospatial information model based on the ground object classification labels and the topographic feature parameters and marking ground object boundary points; and generating mapping data based on the three-dimensional geospatial information model and the ground object boundary points. The present invention adopts an improved three-dimensional geospatial information modeling algorithm, finely marks the ground object boundary points, and effectively solves the problems of great difficulty in ground object classification, fuzzy boundary recognition and poor stability of model training in complex terrains.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing, and in particular to a method for processing geographic information mapping data based on image analysis. Background Art

[0002] With the rapid development of geographic information technology and remote sensing image processing technology, geographic information mapping based on image analysis has become an important research direction in the modern mapping field. Traditional geographic information mapping methods mainly rely on ground mapping equipment to obtain data. Although these methods can provide relatively high-precision measurement results, due to limitations in terrain conditions, mapping scope, and the on-site work efficiency of operators, they have significant limitations when dealing with large-scale complex terrains or high-resolution requirements. In recent years, with the increasing maturity of satellite remote sensing technology, the acquisition of high-resolution multi-angle satellite image data has provided a new data source for geographic information mapping. Using multi-angle satellite remote sensing images for mapping can not only cover large-scale terrains but also provide rich three-dimensional geographic information through multi-view images. However, the processing of multi-angle satellite image data usually involves complex image preprocessing, feature extraction, and object classification algorithms, which pose higher requirements for algorithm efficiency and accuracy. At the same time, the rapid growth of data scale also poses challenges to data storage and processing capabilities.

[0003] Although existing geographic information mapping methods based on remote sensing images have made some progress in terrain feature extraction and three-dimensional model construction, there are still some deficiencies. First, in the process of terrain feature extraction, most existing technologies rely on traditional rule matching or algorithms based on shallow features, resulting in limited accuracy when dealing with complex terrains or detailed features. Second, existing object classification models are usually based on single-view image data, ignoring the perspective compensation information provided by multi-angle images and unable to comprehensively reflect the three-dimensional spatial characteristics of objects. Third, the construction of three-dimensional geographic information models and the accurate annotation of object boundaries are still difficult points. Existing methods often lack effective utilization of the correlation between terrain features and object classification, resulting in insufficient accuracy of mapping data generation. In addition, since multiple steps in the processing flow are independent of each other and fail to achieve efficient coordination, the data processing efficiency is low and it is difficult to meet the requirements of efficient mapping. Summary of the Invention

[0004] In view of the problems existing in the prior art, the present invention proposes a method for processing geographic information mapping data based on image analysis.

[0005] Therefore, the present invention provides a method for processing geographic information mapping data based on image analysis, which can solve the problems mentioned in the background art.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a method for processing geographic information mapping data based on image analysis, which includes:

[0008] Collect multi-angle satellite remote sensing images and divide the multi-angle satellite remote sensing images into several image blocks according to the geographic coordinate system;

[0009] Extract the terrain feature parameters of the image blocks and establish a feature vector of the terrain feature parameters;

[0010] According to the feature vector, establish a ground object classification model to classify the image blocks and obtain ground object classification labels;

[0011] Based on the ground object classification labels and the terrain feature parameters, construct a three-dimensional geographic information model and mark the ground object boundary points;

[0012] Generate mapping data based on the three-dimensional geographic information model and the ground object boundary points.

[0013] As a preferred solution of the method for processing geographic information mapping data based on image analysis according to the present invention, wherein: the terrain feature parameters include elevation values, slope values, and texture feature values;

[0014] Combine the elevation value, the slope value, and the texture feature value to construct the feature vector;

[0015] The calculation of the elevation value is shown in the following formula:

[0016] ,

[0017] wherein, represents the gray value at the coordinate point in the image, is an adaptive kernel function, is a calculation window, is a smoothing coefficient, represents the Laplace operator, is the elevation value at the coordinate point .

[0018] As a preferred solution of the method for processing geographic information mapping data based on image analysis according to the present invention, wherein: the slope value is calculated based on the spatial change rate of the elevation value, as shown in the following formula:

[0019] ,

[0020] wherein, and are respectively direction and direction elevation gradients, is the slope correction coefficient, is the elevation influence factor, represents the coordinate point at the slope value;

[0021] The texture feature value is extracted by using the gray-level co-occurrence matrix, as shown in the following formula:

[0022] ,

[0023] where, is the probability value at position in the gray-level co-occurrence matrix, is the pixel spacing, is the direction angle, is a small constant to avoid the logarithm being zero, is the distance weight function, is the direction weight function, represents a specific distance and direction under the texture feature value.

[0024] As a preferred solution of the method for processing geospatial information mapping data based on image analysis according to the present invention, wherein: obtaining the ground object classification label includes:

[0025] Calculating the weights of features at different scales through the attention mechanism;

[0026] Fusing the features of different branches of the ground object classification model with the weights of features at different scales obtained by the attention mechanism to obtain a comprehensive feature vector;

[0027] Outputting the ground object classification label according to the comprehensive feature vector, and optimizing the ground object classification model through a loss function.

[0028] As a preferred solution of the method for processing geospatial information mapping data based on image analysis according to the present invention, wherein: the classification confidence of the ground object classification label is calculated by the following formula:

[0029] ,

[0030] where, represents the current ground object category, is the classification weight of the current ground object category , is the bias term of the current ground object category , is the comprehensive feature vector, is the th bias term of the category, is the weight matrix of the i-th category;

[0031] The loss function is expressed by the following formula:

[0032] ,

[0033] where, is the true label, is the predicted probability, is the sample difficulty coefficient, is the balance factor, represents the final loss function value.

[0034] As a preferred solution of the method for processing geospatial information mapping data based on image analysis according to the present invention, wherein: the construction of the three-dimensional geospatial information model includes:

[0035] According to the ground object classification label and the terrain feature parameter, extract the ground object boundary feature through a bidirectional recurrent neural network;

[0036] Set a differentiated boundary transition zone for different ground object categories, and construct a three-dimensional grid model based on boundary constraints based on the ground object boundary feature;

[0037] In the transition zone area, perform smooth transition processing on the elevation value through the cubic spline interpolation algorithm, and optimize the vertex position of the three-dimensional model through the iterative least squares method to obtain the three-dimensional geospatial information model;

[0038] Obtaining the ground object boundary points includes:

[0039] Calculate the initial boundary point coordinates by using the region segmentation algorithm and the adaptive boundary point extraction algorithm;

[0040] Optimize and screen the initial boundary point coordinates;

[0041] Interpolate the reserved boundary points to obtain the ground object boundary points and generate a continuous boundary curve.

[0042] As a preferred solution of the method for processing geospatial information mapping data based on image analysis according to the present invention, wherein: the extraction of the ground object boundary feature is expressed by the following formula:

[0043] ,

[0044] where, and are the forward and backward LSTM units respectively, is the classification feature at position , is the feature fusion operator, is the class adaptive weight, is the position The ground object boundary feature value at

[0045] The three-dimensional grid model is represented by the following formula:

[0046] ,

[0047] where is the vertex coordinate, is the elevation gradient, is the boundary response function, is the edge smoothing factor, is the boundary feature of the vertex coordinate, is the three-dimensional grid model.

[0048] As a preferred solution of the method for processing geospatial mapping data based on image analysis according to the present invention, wherein: the calculation of the initial boundary point coordinates is shown by the following formula:

[0049] ,

[0050] where is the boundary saliency measure, is the local structure weight, is the classification confidence, is the balance coefficient, is the three-dimensional coordinate of the boundary point;

[0051] The interpolation of the retained boundary points is represented by the following formula:

[0052] ,

[0053] where is the cubic B-spline basis function, is the tension control factor, is the continuous boundary curve equation, is the th coordinate of the boundary point, is the parameter variable.

[0054] In a second aspect, an embodiment of the present invention provides a system for processing geospatial mapping data based on image analysis, which includes:

[0055] An acquisition module, configured to acquire multi-angle satellite remote sensing images and divide the multi-angle satellite remote sensing images into a plurality of image blocks according to a geographic coordinate system;

[0056] A feature extraction module, configured to extract terrain feature parameters of the image blocks and establish a feature vector of the terrain feature parameters;

[0057] The ground object classification module is used to establish a ground object classification model based on the feature vector to classify the image patches, and obtain ground object classification labels;

[0058] The boundary point marking module is used to establish a ground object classification model based on the feature vector to classify the image patches, and obtain ground object classification labels;

[0059] The surveying and mapping data generation module is used to generate surveying and mapping data based on the three-dimensional geographic information model and the ground object boundary points.

[0060] The beneficial effects of the present invention are the acquisition and refined division of multi-angle satellite remote sensing images, combined with high-precision terrain feature extraction technology, realizing the comprehensive analysis of ground object features, and introducing a multi-scale convolutional neural network (MSCNN) for ground object classification, greatly improving the classification accuracy and the accuracy of boundary recognition. By adopting an improved three-dimensional geographic information modeling algorithm, not only the ground object boundary points are accurately marked, but also high-quality topographic maps, contour maps and ground object distribution maps are generated through optimization processing, effectively solving the problems of difficult ground object classification, fuzzy boundary recognition and poor model training stability in complex terrains. In addition, through hierarchical quality control and hierarchical structure storage, it is ensured that the surveying and mapping data has high precision, high integrity and good compatibility, meeting the requirements of engineering applications and demonstrating significant technical advantages in the field of geographic information surveying and mapping. Description of the Drawings

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. Among them:

[0062] Figure 1 It is a flowchart of a geographic information surveying and mapping data processing method based on image analysis. Detailed Embodiments

[0063] To make the above objects, features and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0064] In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways than those specifically described herein. Those skilled in the art can make similar extensions without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0065] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other with other embodiments.

[0066] The present invention is described in detail in conjunction with schematic diagrams. When detailing the embodiments of the present invention, for the convenience of explanation, the cross-sectional views showing the device structure will be enlarged locally in a non-general proportion, and the schematic diagrams are only examples and should not limit the scope of protection of the present invention herein. In addition, in actual production, three-dimensional spatial dimensions including length, width, and depth should be included.

[0067] Meanwhile, in the description of the present invention, it should be noted that the orientation or positional relationships indicated by terms such as "upper, lower, inner, and outer" are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be construed as a limitation on the present invention. In addition, the terms "first, second, or third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0068] Unless otherwise clearly defined and limited in the present invention, the terms "mounted, connected, and coupled" shall be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may also be a mechanical connection, an electrical connection, or a direct connection, or may be indirectly connected through an intermediate medium, or may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0069] Embodiment 1

[0070] Referring to Figure 1 , which is the first embodiment of the present invention. This embodiment provides a method for processing geospatial information mapping data based on image analysis, including:

[0071] S1: Collect multi-angle satellite remote sensing images, and divide the multi-angle satellite remote sensing images into several image blocks according to the geographic coordinate system;

[0072] Collect multi-angle remote sensing images of the study area at different times by multiple low-orbit remote sensing satellites. The multi-angle remote sensing images include vertical view, 45-degree oblique view in the east-west direction, and 45-degree oblique view in the north-south direction. Divide the multi-angle remote sensing images into square image blocks of 500 meters × 500 meters according to the WGS84 geographic coordinate system, and set an overlapping area of 100 meters between adjacent image blocks for subsequent image stitching.

[0073] S2: Extract the topographic feature parameters of the image blocks, including elevation values, slope values, and texture feature values, and establish the feature vectors of the topographic feature parameters;

[0074] Extract the feature parameters of the image blocks. Obtain the elevation value by calculating the digital elevation model of each image block through the pixel gray value. More specifically, when extracting the feature parameters of the image blocks, first calculate the elevation value of the digital elevation model DEM based on the pixel gray gradient operator :

[0075] ,

[0076] where, represents the gray value of the pixel point , is the adaptive kernel function, is the calculation window, is the smoothing coefficient, represents the Laplace operator.

[0077] Subsequently, calculate the slope value using the spatial change rate of the elevation value :

[0078] ,

[0079] where, and are the elevation gradients in the direction and direction respectively, is the slope correction coefficient, is the elevation influence factor, represents the slope value at the coordinate point .

[0080] Further extract the texture feature values including energy, contrast, and entropy value using the gray-level co-occurrence matrix, as shown in the following formula:

[0081] ,

[0082] where, is the probability value at the position in the gray-level co-occurrence matrix, is the pixel spacing, is the direction angle, is a small constant to avoid a logarithm of zero, is the distance weight function, is the direction weight function, represents a specific distance and direction under the texture eigenvalue.

[0083] Combining the elevation value, the slope value, and the texture eigenvalue to construct a 15-dimensional feature vector , as shown in the following formula:

[0084] ,

[0085] where, , , are the normalized eigenvalues respectively, , , are the corresponding feature weight coefficients, where the elevation value occupies 3 dimensions, the slope value occupies 2 dimensions, and the texture eigenvalue occupies 10 dimensions.

[0086] S3: According to the feature vector, establish a ground object classification model to classify the image block and obtain a ground object classification label;

[0087] According to the 15-dimensional feature vector, construct a multi-scale convolutional neural network MSCNN for ground object classification. The MSCNN contains three parallel convolutional branches, and the convolutional kernel sizes of each branch are 3×3, 5×5, and 7×7 respectively. Each branch uses four convolutional layers, where the first layer uses 32 feature maps, the second layer uses 64 feature maps, the third layer uses 128 feature maps, and the fourth layer uses 256 feature maps. Batch normalization layers and ReLU activation functions are set between the convolutional layers, and the following attention mechanism is used to calculate the weights of features at different scales:

[0088] ,

[0089] where, and are learnable weight matrices, is the feature map at the and are bias terms.

[0090] Fuse the features of the three branches through attention weighting to obtain the comprehensive feature vector Fc:

[0091] ,

[0092] where, Denotes a convolution operation, is a scale adaptive function.

[0093] The final output includes land cover classification labels such as buildings, roads, vegetation, and water bodies. The classification confidence is calculated by the Softmax function:

[0094] ,

[0095] Among them, denotes the land cover class, is the corresponding classification weight, is the bias term.

[0096] An improved cross-entropy loss function L is used to optimize and train the model:

[0097] ,

[0098] Among them, is the true label, is the predicted probability, is the sample difficulty coefficient, is the balance factor.

[0099] It should be noted that the sample difficulty coefficient is calculated by an adaptive dynamic weighting mechanism. The specific calculation process is as follows:

[0100] First, calculate the historical prediction error Ei of the sample:

[0101] ,

[0102] Among them, is the number of historical predictions, is the true label at the th iteration, is the predicted probability at the th iteration.

[0103] Then, calculate the relative entropy of the sample based on the prediction error:

[0104] ,

[0105] Among them, is the maximum entropy value in the current batch;

[0106] Next, introduce the local spatial consistency index :

[0107] ,

[0108] Among them, and are the feature vectors of adjacent samples, is the spatial distance weight, is the normalization factor.

[0109] Finally, the sample difficulty coefficient is obtained through the weighted combination of the above three indicators.

[0110] Preferably, the sample difficulty coefficient is obtained by fusing the historical prediction error , relative entropy and the spatial consistency index The three-dimensional evaluation indicators of the three dimensions, and a dynamic weight adaptive adjustment mechanism is adopted, which significantly improves the limitations of the traditional single error evaluation method, and reflects the technical effects of an 8-12% increase in classification accuracy, a 25% increase in training convergence speed, and a more than 15% increase in the accuracy of ground object boundary recognition in the experiment. At the same time, it solves key technical problems such as complex ground object boundary recognition, small sample classification, and model training stability, and has strong adaptability, robustness, and generalization ability, showing obvious technical advantages in the practical application of complex terrain and ground object classification.

[0111] S4: Construct a three-dimensional geographic information model according to the ground object classification label and the terrain feature parameter, and label the ground object boundary points;

[0112] According to the ground object classification label and the terrain feature parameter, an improved three-dimensional reconstruction algorithm is used to construct a geographic information model, a reference plane grid is set, the grid spacing is 10 meters, and the grid nodes include elevation attributes and ground object type attributes. First, the ground object boundary feature B(x,y) is extracted through a bidirectional recurrent neural network BRNN:

[0113] ,

[0114] wherein, and are the forward and backward LSTM units respectively, is the classification feature at the position , is the feature fusion operator, is the category adaptive weight, is the position The ground object boundary feature value at.

[0115] Differentiated boundary transition zones are set for different ground object categories, where the building boundary transition zone width is 2 meters, the road boundary transition zone width is 5 meters, and the water body and vegetation boundary transition zone width is 3 meters, and a three-dimensional grid model based on boundary constraints is constructed:

[0116] ,

[0117] Among them, is the vertex coordinate, is the elevation gradient, is the boundary response function, is the edge smoothing factor, is the boundary feature of the vertex coordinate, is the 3D mesh model;

[0118] In the transition zone area, the elevation values are smoothed through the cubic spline interpolation algorithm, and the vertex positions of the 3D model are optimized by the iterative least squares method:

[0119] ,

[0120] Among them, is the energy function, is the learning rate, is the regularization coefficient, is the local curvature.

[0121] The region segmentation algorithm and the adaptive boundary point extraction algorithm are used to calculate the coordinates P(x, y, z) of the key boundary points, where the region with an elevation change rate greater than 0.3 and spanning different land cover types is marked as the boundary candidate region:

[0122] ,

[0123] Among them, is the boundary saliency measure, is the local structure weight, is the classification confidence, is the balance coefficient; One boundary point is sampled every 1 meter within the boundary candidate region; The extracted boundary points are optimized and screened to remove the redundant points with a distance between adjacent boundary points less than 0.5 meters and an elevation difference less than 0.3 meters. Finally, a continuous boundary curve L(t) is generated by boundary point interpolation:

[0124] ,

[0125] Among them, is the cubic B-spline basis function, is the tension control factor. Finally, the optimized boundary points are grouped and stored according to the land cover types, and each boundary point contains information such as spatial coordinates, land cover attributes, and adjacent land cover types.

[0126] S5: Generate mapping data based on the 3D geographic information model and the land cover boundary points, including topographic maps, contour maps, and land cover distribution maps.

[0127] Generate surveying and mapping data based on the three-dimensional geographic information model and the feature boundary points. First, generate topographic map data. Project the three-dimensional model onto a two-dimensional plane at a scale of 1:500, and use the layer coloring method to express topographic features. The color difference between adjacent elevation levels is set to 20%. The resolution of the topographic map is set to 0.5 meters per pixel. At the same time, mark the absolute elevation values at key geomorphic points, and vectorize artificial features such as buildings and roads using on-site survey standards.

[0128] Subsequently, generate contour map data. The basic contour interval is set to 1 meter, the main curve interval is 5 meters, and the index contour interval is 25 meters. In steep areas with a slope greater than 30 degrees, the contour interval is encrypted to 0.5 meters. For gentle areas with a slope less than 5 degrees, additional elevation annotation points are used for supplementary expression. All contour lines are fitted using smooth spline curves to ensure the continuity and smoothness of the curves. At the same time, elevation annotations are added at the turning points of the contour lines.

[0129] Next, construct a feature distribution map. Use a classification coloring scheme to express different feature types. Among them, buildings are represented in red, roads are represented in gray, vegetation is represented in green, and water bodies are represented in blue. A 0.5-meter-wide transition buffer is set at the feature boundaries, and the boundary recognition effect is enhanced through gradient tones and texture fills. At the same time, the attribute information of the main features is superimposed and displayed, including the floor area of buildings, the width grade of roads, the coverage of vegetation, and the water depth grade of water bodies.

[0130] During the generation process, hierarchical quality control is carried out on the surveying and mapping data. First, perform geometric accuracy inspection. The plane position error is controlled within 0.2 meters, and the elevation error is controlled within 0.1 meters. Subsequently, perform attribute integrity inspection to ensure that all feature elements have complete spatial information and attribute information. Finally, perform map symbol standardization inspection and strictly implement the schema regulations in the 1:500 topographic map surveying and mapping specification.

[0131] The generated surveying and mapping data is stored in a hierarchical structure. Each layer contains corresponding spatial data and attribute data, supports two-way conversion between vector format and raster format, and establishes a data update mechanism to record the generation time, update history, and quality assessment results of the surveying and mapping data. The surveying and mapping results meet the requirements of the surveying and mapping industry standards and can be directly applied to engineering planning, design, and construction.

[0132] Furthermore, this embodiment also provides a geographic information surveying and mapping data processing system based on image analysis, including:

[0133] An acquisition module, configured to acquire multi-angle satellite remote sensing images and divide the multi-angle satellite remote sensing images into several image blocks according to the geographic coordinate system;

[0134] A feature extraction module, configured to extract the terrain feature parameters of the image block and establish a feature vector of the terrain feature parameters;

[0135] A ground object classification module, configured to establish a ground object classification model according to the feature vector to perform ground object classification on the image block and obtain a ground object classification label;

[0136] A boundary point marking module, configured to establish a ground object classification model according to the feature vector to perform ground object classification on the image block and obtain a ground object classification label;

[0137] A surveying and mapping data generation module, configured to generate surveying and mapping data based on the three-dimensional geographic information model and the ground object boundary points.

[0138] This embodiment also provides a computer device, applicable to the case of a geographic information surveying and mapping data processing method based on image analysis, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the geographic information surveying and mapping data processing method based on image analysis as proposed in the above embodiment.

[0139] This computer device may be a terminal. This computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of this computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of this computer device may be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0140] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the geographic information surveying and mapping data processing method based on image analysis as proposed in the above embodiment.

[0141] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. Technical details not described in detail in this embodiment can be referred to in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0142] Embodiment 2

[0143] This is the second embodiment of the present invention. This embodiment provides a geographic information surveying and mapping data processing method based on image analysis. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0144] A 3 square kilometer area in the central area of ​​a city was selected as the study area, and three low-orbit remote sensing satellites (satellite numbers are RS-101, RS-102, and RS-103) were used to collect image data from September 15 to October 15, 2023. The satellite orbit altitude is 480 kilometers, the spatial resolution is better than 0.5 meters, and the sampling period is 3 days. The acquisition period is selected in the time window with clear weather and good atmospheric visibility to ensure image quality. The study area includes typical urban buildings, road networks, park green spaces, and river systems. The terrain undulations are obvious, and the maximum height difference reaches 76 meters.

[0145] First, according to step S1 of the invention, remote sensing images are collected at vertical viewing angles, 45-degree oblique viewing angles in the east-west direction, and 45-degree oblique viewing angles in the north-south direction. Customized image processing software is used to divide the study area into 12 standard image blocks of 500 meters by 500 meters, and a 100-meter overlap zone is set between adjacent image blocks. In the image preprocessing stage, atmospheric correction, geometric correction, and radiation calibration are performed to ensure that the registration accuracy of multi-phase and multi-angle images is better than 0.3 pixels.

[0146] Then, according to step S2, the elevation value, slope value and texture feature value of each image block are calculated respectively using the improved feature extraction algorithm. In the feature extraction process, an adaptive calculation window of 7×7 pixels is used, the smoothing coefficient is set to 0.15, and the elevation influence factor is 0.08. When extracting texture features, the pixel spacing is set to 4, and the direction angles include four directions: 0°, 45°, 90°, and 135°. Finally, a 15-dimensional feature vector is obtained, in which the elevation feature weight coefficient is 0.4, the slope feature weight coefficient is 0.3, and the texture feature weight coefficient is 0.3.

[0147] Next, according to step S3, a multi-scale convolutional neural network is constructed and trained. The network training adopts a batch size of 64, an initial learning rate of 0.001, an Adam optimizer, and 200 training rounds. During the model training process, the sample difficulty coefficient is dynamically adjusted through cross-validation, the spatial distance weight is set to 0.6, and the normalization factor is taken as 0.1.

[0148] In step S4, when constructing the three-dimensional geographic information model, the grid reference elevation is set to 50 meters above sea level, and an adaptive grid generation strategy is adopted to encrypt grid nodes in areas with drastic terrain changes. The threshold of the boundary response function is set to 0.25, the edge smoothing factor is set to 0.18, and the convergence threshold of the iterative least squares method is set to 0.001.

[0149] Finally, in step S5, surveying and mapping data with a scale of 1:500 is generated, using 24-bit true color mode, with a resolution of 300 dpi. The data output adopts a hierarchical structure and supports mainstream formats such as DWG and SHP. The experimental data shown in Table 1 below is obtained:

[0150] Table 1 Experimental Data Table

[0151] Sample area Planar position accuracy (m) Elevation accuracy (m) Boundary recognition accuracy rate (%) Overall classification accuracy (%) Processing time (min) Data storage capacity (GB) Traditional method A 0.45 0.32 82.5 85.7 156 2.8 Traditional method B 0.38 0.28 84.3 87.2 142 2.5 Method 1 of the present invention 0.21 0.15 95.8 94.3 98 1.9 Method 2 of the present invention 0.19 0.13 96.2 95.1 95 1.8 Method 3 of the present invention 0.18 0.12 96.5 95.4 92 1.7 Method 4 of the present invention 0.17 0.11 97.1 95.8 90 1.7

[0152] By comparing the experimental data, it can be found that the method of the present invention shows significant advantages over the traditional method in multiple key indicators:

[0153] The planar position accuracy is improved from 0.38 - 0.45 meters of the traditional method to 0.17 - 0.21 meters, with an improvement amplitude of 55%;

[0154] The elevation accuracy is improved from 0.28 - 0.32 meters of the traditional method to 0.11 - 0.15 meters, with an improvement amplitude of 60%;

[0155] The boundary recognition accuracy rate is improved from 82.5 - 84.3% to 95.8 - 97.1%, with an improvement amplitude of more than 13 percentage points;

[0156] The overall classification accuracy is improved from 85.7 - 87.2% to 94.3 - 95.8%, with an improvement amplitude of 9 percentage points.

[0157] The processing time is reduced from 142 - 156 minutes of the traditional method to 90 - 98 minutes, and the efficiency is improved by about 37%;

[0158] The data storage capacity is reduced from 2.5 - 2.8 GB to 1.7 - 1.9 GB, and the storage efficiency is improved by about 32%;

[0159] It is particularly worth noting that the improvement of the method of the present invention in the boundary recognition accuracy rate is the most significant, which solves the key technical problems in the boundary recognition of complex terrain and features. At the same time, the improvement of processing efficiency and the reduction of storage capacity also reflect the engineering value of the present invention in practical applications.

[0160] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A geographic information surveying and mapping data processing method based on image analysis, characterized in that: include: Collecting multi-angle satellite remote sensing images, and dividing the multi-angle satellite remote sensing images into a plurality of image blocks according to a geographic coordinate system; Extracting terrain characteristic parameters of the image block and establishing a characteristic vector of the terrain characteristic parameters; According to the feature vector, a ground object classification model is established to classify the ground objects in the image block to obtain a ground object classification label; Constructing a three-dimensional geographic information model based on the object classification labels and the terrain feature parameters, and marking the object boundary points; Generate surveying and mapping data based on the three-dimensional geographic information model and the boundary points of the land object; The terrain characteristic parameters include elevation value, slope value and texture characteristic value; Combining the elevation value, the slope value and the texture feature value to construct the feature vector; The calculation of the elevation value is shown in the following formula: ， in, Represents the coordinate points in the image The gray value at is the adaptive kernel function, is the calculation window, is the smoothing coefficient, represents the Laplace operator, is the coordinate point The elevation value at ; The slope value is calculated based on the spatial change rate of the elevation value, as shown in the following formula: ， in, and They are Direction and The elevation gradient of the direction, is the slope correction factor, is the elevation influencing factor, Indicates coordinate points The slope value at ; The texture feature value is extracted using the gray level co-occurrence matrix, as shown in the following formula: ， in, is the position in the gray-level co-occurrence matrix The probability value of is the pixel pitch, is the direction angle, To avoid small constants whose logarithms are zero, is the distance weight function, is the direction weight function, Indicates a specific distance and direction The texture feature value under ; The construction of the three-dimensional geographic information model includes: Extracting the boundary features of the object through a bidirectional recursive neural network according to the object classification label and the terrain feature parameters; Differentiated boundary transition zones are set for different types of land objects, and a three-dimensional grid model based on boundary constraints is constructed based on the boundary characteristics of the land objects; In the transition zone area, the elevation value is smoothly transitioned by using a cubic spline interpolation algorithm, and the vertex positions of the three-dimensional model are optimized by an iterative least squares method to obtain the three-dimensional geographic information model; Obtaining the boundary points of the ground feature includes: The region segmentation algorithm and the adaptive boundary point extraction algorithm are used to calculate the coordinates of the initial boundary points; Optimizing and screening the coordinates of the initial boundary points; Interpolating the retained boundary points to obtain the boundary points of the feature and generate a continuous boundary curve; The extraction of the boundary features of the ground object is expressed by the following formula: ， in, and are the forward and backward LSTM units, respectively. For location The classification characteristics of is the feature fusion operator, is the class adaptive weight, For location The feature value of the boundary of the object at ; The three-dimensional grid model is expressed by the following formula: ， in, are vertex coordinates, is the elevation gradient, is the boundary response function, is the edge smoothing factor, is the boundary feature of the vertex coordinates, It is a three-dimensional mesh model; The calculation of the initial boundary point coordinates is shown in the following formula: ， in, is the boundary saliency measure, is the local structure weight, is the classification confidence, is the balance coefficient, is the three-dimensional coordinate of the boundary point; The interpolation of the reserved boundary points is expressed by the following formula: ， in, is the cubic B-spline basis function, is the tension control factor, is the continuous boundary curve equation, For the The coordinates of the boundary points, is a parameter variable.

2. The method for processing geographic information surveying and mapping data based on image analysis according to claim 1, characterized in that: Obtaining the object classification label includes: Calculate the weights of features at different scales through the attention mechanism; The features of different branches of the object classification model are weighted and fused with the weights of the features of different scales obtained by the attention mechanism to obtain a comprehensive feature vector; The object classification label is output according to the comprehensive feature vector, and the object classification model is optimized through a loss function.

3. The method for processing geographic information surveying and mapping data based on image analysis according to claim 2, characterized in that: The classification confidence of the object classification label is calculated by the following formula: ， in, Indicates the current feature category. The current feature category The classification weights, The current feature category The bias term, is the comprehensive feature vector, For the The bias term of each category, It is The weight matrix of each category; The loss function is expressed by the following formula: ， in, is the true label, is the predicted probability, is the sample difficulty coefficient, is the balance factor, Represents the final loss function value.

4. A geographic information surveying and mapping data processing system based on image analysis, based on the geographic information surveying and mapping data processing method based on image analysis according to any one of claims 1 to 3, characterized in that: include: An acquisition module is used to acquire multi-angle satellite remote sensing images and divide the multi-angle satellite remote sensing images into a plurality of image blocks according to a geographic coordinate system; A feature extraction module, used to extract terrain feature parameters of the image block and establish a feature vector of the terrain feature parameters; A ground object classification module is used to establish a ground object classification model according to the feature vector to classify the image block and obtain a ground object classification label; A boundary point marking module is used to establish a ground object classification model according to the feature vector to classify the ground objects in the image block and obtain a ground object classification label; A surveying and mapping data generation module is used to generate surveying and mapping data based on the three-dimensional geographic information model and the boundary points of the land objects.

Images