An image data processing method for mapping

Abstract

The present invention discloses an image data processing method for surveying and mapping, specifically relating to the field of image data processing, including: collecting optical image data and near-infrared data of the surveyed building area, extracting core features for precise mapping, and establishing spectral comparison features of background elements; generating a background mask based on the comparison result of the spectral features and the spectral comparison features of the background elements, and segmenting the background area in combination with a dynamic segmentation model of region growing; marking the grid that does not fully label the background mask in the orthoimage as a building-background mixed area, performing secondary annotation of the background mask on the building-background mixed area, and suppressing background features; extracting the preliminary contour of the building by using an edge detection algorithm, and performing multi-level progressive fitting of curves on the extracted building contour edge; performing feature calibration and data layer superposition on the building contour based on curve fitting to form a comprehensive expression of the building and geographical feature elements.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and more specifically, to an image data processing method for surveying and mapping. Background Art

[0002] In the fields of building surveying and mapping and ground object segmentation, the accurate differentiation between buildings and complex backgrounds (such as vegetation, water bodies, shadows, etc.) has long been a technical difficulty. In complex background areas, due to the high similarity of texture, shape, and spectral characteristics, it is easy to be confused with buildings, resulting in limitations in accurately extracting building areas using traditional segmentation methods. Especially in urban dense areas or natural environments, buildings often overlap with background elements, such as parts of buildings covered by vegetation or water body reflections interfering with building edge features, further exacerbating the difficulty of differentiation.

[0003] Near-infrared (NIR) data has strong differentiation ability for background elements such as vegetation and water bodies in terms of reflection characteristics. For example, vegetation has high reflectance in the NIR band, while the reflectance of buildings is relatively low. However, segmentation methods that rely solely on near-infrared data often show insufficient accuracy when dealing with scenes with similar spectral characteristics or complex textures. In addition, there are usually problems of inaccurate annotation or mislabeling in the building boundary and background transition areas (such as parts of buildings blocked by vegetation) in complex backgrounds, and existing methods lack the ability to refine these areas.

[0004] How to propose an image data processing method that suppresses complex background features and then combines edge detection and curve fitting techniques to achieve accurate surveying and mapping is an urgent problem to be solved.

[0005] To solve the above problems, a technical solution is provided as follows. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an image data processing method for surveying and mapping to solve the problems raised in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solution:

[0008] S1: Collect and extract the core features of the optical image data and near-infrared data of the surveyed building area, perform precise two-dimensional mapping on the core features to form a planar grid structure, measure the near-infrared reflection spectral characteristic data of non-building elements, and establish a spectral comparison feature of background elements;

[0009] S2: Extract the spectral features in the grid, generate a background mask based on the comparison result between the spectral features and the spectral comparison feature of background elements, and segment the background area in combination with the dynamic segmentation model of region growing;

[0010] S3: Mark the grids in the orthoimage that do not have a fully labeled background mask as building-background mixed areas, perform secondary labeling of the background mask on the building-background mixed areas, and suppress the background features of the background mask areas after secondary labeling;

[0011] S4: Use the Canny edge detection algorithm to extract the preliminary outline of the building, and perform multi-level progressive fitting of the B-Spline curve on the extracted building outline edges;

[0012] S5: Calibrate the elements of the building outline after fitting with the B-Spline curve and overlay the data layers to form a comprehensive expression of the building and geographical feature elements.

[0013] In a preferred embodiment, in S1, collect and extract the core features of the optical image data and near-infrared data of the surveyed building area, perform precise two-dimensional mapping of the core features to form a planar grid structure, measure the near-infrared reflection spectral characteristic data of non-building elements, and establish the spectral reference features of background elements, specifically including:

[0014] Collect multi-modal data of the surveyed building area through a high-precision remote sensing platform, and the multi-modal data includes optical image data and near-infrared data;

[0015] Through multi-view image matching and geographical calibration, unify all modal data into the same coordinate system and projection standard to form an orthoimage, and perform core feature extraction and feature alignment on the optical image data and near-infrared data in the modal data;

[0016] Perform precise two-dimensional mapping of the core features of different modalities after alignment in the orthoimage to form a planar grid structure, and each grid in the planar grid structure contains at least one pixel;

[0017] Analyze each grid of the grid after modal feature mapping, and label the core features of the grid. The core features include the RGB and texture features of the building extracted from the optical image channel, and the reflection spectral features of the near-infrared channel;

[0018] Based on the multi-layer feature overlay strategy, integrate the RGB, texture features, and spectral features into multi-modal combined features, where each modality represents a corresponding physical feature attribute expression;

[0019] Select the grids or grid groups that contain all non-building object elements, measure the near-infrared reflection spectral characteristic data of non-building elements, and establish the spectral reference features of background elements.

[0020] In a preferred embodiment, in S2, spectral features in the grid are extracted, a background mask is generated based on the comparison result between the spectral features and the spectral comparison features of background elements, and the specific steps of segmenting the background region by combining the dynamic segmentation model of region growing include:

[0021] Select at least one spectral feature comparison point in each grid. Based on the hit situation of the spectral feature comparison point in the spectral comparison features of background elements, identify and generate a preliminary background mask for all non-building regions in the image;

[0022] Select the grid with dense preliminary background mask annotations as the initial point, recursively compare adjacent grids using the multi-modal combined features of optical images and near-infrared data, and perform background mask diffusion annotation based on the feature similarity, where the feature similarity is the feature cosine similarity value that fuses texture, brightness, and spectral reflection.

[0023] In a preferred embodiment, in S3, the grids in the orthoimage that do not have a completely annotated background mask are marked as building-background mixed regions, the background mask of the building-background mixed regions is re-annotated, and the background feature suppression of the re-annotated background mask region specifically includes:

[0024] Obtain the grid combination in the orthoimage where there is an unannotated background mask region, mark it as a building-background mixed region, and extract the multi-modal combined features of the optical image and near-infrared data of all grids in the building-background mixed region;

[0025] Based on the hit result of the spectral features of all pixel points in the grid in the spectral comparison features of background elements, re-annotate the background mask of the building-background mixed region;

[0026] Based on the feature similarity when the spectral features hit the spectral comparison features of background elements, perform equal-proportion feature weight attenuation on the multi-modal combined features at the pixel point positions of the re-annotated background mask, and weaken the feature expression of the background in the building-background mixed region;

[0027] Use the orthoimage after background feature suppression processing as the initial analysis image of the buildings in the survey area for contour fitting.

[0028] In a preferred embodiment, in S4, the Canny edge detection algorithm is used to extract the preliminary contour of the building, and the multi-level progressive fitting of the B-Spline curve is performed on the extracted building contour edge, specifically including:

[0029] Use the Canny edge detection algorithm to extract the edges of the initial analysis image of the buildings in the survey area, construct structured scene contour information, and introduce multi-scale convolutional features to extract and identify the local details and global morphology of the buildings;

[0030] According to the length and complexity of the original contour of the building, the original contour is simplified to obtain key nodes, and the key nodes in the parts with higher geometric complexity are extracted and marked as a sparse control point set. The preliminary B-Spline curve is constructed using these sparse control point sets;

[0031] The constructed preliminary B-Spline curve is segmented, and the local curvature distribution of each curve segment and the error between adjacent key nodes within the curve segment interval are calculated. Based on the error calculation results, the threshold comparison method is used to determine whether the detail fitting degree of the curve in this area meets the standard;

[0032] In the curve segments where the detail fitting degree does not meet the standard, the number of control points is increased by control point interpolation, and multi-level cyclic fitting of the B-Spline curve is performed. The number of control points is increased by an equal amount in each layer of fitting based on the previous layer, and at the same time, the order of the B-Spline curve is increased until the detail fitting degree of all segments of the B-Spline curve meets the standard, and then the loop ends to obtain the fitted building contour edge.

[0033] In a preferred embodiment, in S5, the building contour after B-Spline curve fitting is subjected to element calibration and data layer superposition to form a comprehensive expression of the building and geographical feature elements, which specifically includes:

[0034] The coordinate data corresponding to the building contour is converted into element information of the building to establish a geographical data layer;

[0035] All the fitted building contour boundaries are subjected to unified coordinate projection transformation and spatial position alignment, and at the same time, the geographical data layer is superposed, and attribute information is added to the building object to form a comprehensive expression of the building and geographical feature elements.

[0036] The technical effects and advantages of an image data processing method for surveying and mapping drawing of the present invention:

[0037] Through the feature extraction and fusion analysis of optical images and near-infrared data, the texture and shape advantages of optical images and the spectral difference characteristics of near-infrared data are effectively utilized to achieve the preliminary separation of buildings and the background. The background mask generated based on spectral feature comparison combined with the region growing dynamic segmentation model can accurately identify and suppress the interference of complex background areas (such as vegetation and water bodies), especially in the transition area of the building boundary. The mixed feature expression in the intersection area of the complex background and the building is further weakened through secondary annotation and background feature suppression. Finally, through the Canny edge detection and B-Spline curve fitting technologies, the building contour boundary is gradually optimized at multiple levels, making the shape expression of the building smoother and more realistic, providing high-quality contour data support for subsequent surveying and mapping drawing. Brief Description of the Drawings

[0038] Figure 1 Schematic diagram of an image data processing method for mapping in the present invention. Detailed implementation manners

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0040] Embodiment 1

[0041] Figure 1 An image data processing method for mapping in the present invention is provided, which includes the following steps:

[0042] S1: Collect and extract the core features of the optical image data and near-infrared data of the surveyed building area, perform precise two-dimensional mapping on the core features to form a planar grid structure, measure the near-infrared reflection spectral characteristic data of non-building elements, and establish the spectral comparison characteristics of background elements;

[0043] S2: Extract the spectral features in the grid, generate a background mask based on the comparison result between the spectral features and the spectral comparison characteristics of background elements, and segment the background area by combining the dynamic segmentation model of region growing;

[0044] S3: Mark the grids in the orthoimage that do not fully label the background mask as the building-background mixed area, perform secondary annotation on the building-background mixed area for the background mask, and suppress the background features of the background mask area after secondary annotation;

[0045] S4: Use the Canny edge detection algorithm to extract the preliminary contour of the building, and perform multi-level progressive fitting of the B-Spline curve on the extracted building contour edge;

[0046] S5: Perform element calibration and data layer superposition on the building contour fitted based on the B-Spline curve to form a comprehensive expression of the building and geographical feature elements.

[0047] In S1, collect and extract the core features of the optical image data and near-infrared data of the surveyed building area, perform precise two-dimensional mapping on the core features to form a planar grid structure, measure the near-infrared reflection spectral characteristic data of non-building elements, and establish the spectral comparison characteristics of background elements.

[0048] In the building area surveying and mapping task, a high-precision remote sensing platform (such as a drone or satellite) is used to collect optical image data and near-infrared (NIR) data. The optical image data is used to obtain the texture, shape, and color characteristics of buildings. The multi-view images are aligned through a multi-view image matching algorithm (such as SIFT key point matching), and the multi-modal data is unified under the same geographic coordinate system and projection standard using geographic calibration technology to generate a high-precision orthophoto image, ensuring the spatial consistency and registration accuracy of the multi-modal data.

[0049] Core feature extraction is performed on the optical image and near-infrared data in the orthophoto image. RGB features (such as color histograms) and texture features (such as gray-level co-occurrence matrices, LBP features) are extracted from the optical image; reflection spectral characteristics are extracted from the near-infrared data. Considering the differences in resolution and feature distribution of different modal data, a resolution matching algorithm is used to perform super-resolution reconstruction on the low-resolution near-infrared data, and through pixel-level feature alignment technology, it is ensured that the optical and near-infrared modal features accurately correspond in the same spatial grid.

[0050] The core features of the aligned optical image and near-infrared data are accurately mapped two-dimensionally in the orthophoto image, and a unified planar grid structure is generated through a pixel-level multi-modal feature superposition strategy. Each grid contains the fusion features of the optical image and near-infrared data, and at least contains one pixel point. The fusion features include RGB colors, texture features, and spectral features, and the features of each modality are represented through multiple channels to support subsequent analysis.

[0051] Grid-by-grid analysis is performed on the grids after modal feature mapping, and the core features of the grids are labeled. The core features include the RGB and texture features of the buildings extracted from the optical image channel, and the reflection spectral features of the near-infrared channel. Among them, the RGB features extract the color features of the optical image in the grid to represent the visual information of the building or background area. The texture features extract the edge and structure information of the optical image to enhance the description of the building shape. The spectral features label the reflection spectral characteristics of the near-infrared data to distinguish the background (such as vegetation, water bodies) and buildings.

[0052] Select samples in the grid or group of grids that completely contain non-building object elements (such as vegetation, water bodies, bare soil). By analyzing their near-infrared spectral reflection characteristics, the spectral reference characteristics of the background elements are established. The specific steps are as follows:

[0053] Background sample screening: Based on the color characteristics of the optical image and the spectral distribution of the near-infrared data, highly representative background grids are screened out (such as vegetation areas with high NIR reflectivity, water body areas with low reflectivity);

[0054] Spectral data determination: Calculate the reflection spectral curve for the screened background samples and record the reflection characteristics of the background elements in the near-infrared band;

[0055] Spectral contrast feature establishment: Cluster the spectral characteristics of the background samples (such as K-Means clustering) to generate a spectral contrast database of background elements for subsequent background differentiation and mask generation.

[0056] In S2, extract the spectral features in the grid, generate a background mask based on the comparison results between the spectral features and the spectral contrast features of background elements, and segment the background area by combining the dynamic segmentation model of region growing.

[0057] In each grid, extract at least one spectral feature point from the reflection characteristics in the near-infrared band, and preferably select the pixel points with the most representative reflection characteristics to ensure that the spectral characteristics of the grid can be accurately expressed. For areas with complex spectral features, multiple pixel points can be selected to improve the expression accuracy.

[0058] Match the spectral features of each comparison point with the pre-established spectral contrast feature library of background elements. Use Euclidean calculation to match the degree between the comparison point and the background spectral features, and mark the features with a matching degree exceeding the set standard as feature hits.

[0059] Select the grid group with a higher density of background annotations in the preliminary background mask as the initial point. These grid groups have a higher background confidence and can be used as the starting area for diffusion annotation. The selection criteria include that the annotation ratio of the background mask exceeds the set threshold (80%) and the cosine similarity of the spectral features hitting the background contrast features is higher than the set threshold (0.85). Recursively compare the adjacent grids using the multi-modal combined features of optical images and near-infrared data, and perform background mask diffusion annotation based on the feature similarity. The feature similarity is the cosine similarity value of the features fusing texture, brightness, and spectral reflection. The similarity value calculation formula is:

[0060]

[0061] In the formula, X i is the feature vector of the i-th initial point, Y i is the feature vector of the i-th adjacent grid, n is the set number of initial points. When a group of X i is compared with Y i after comparison, Y i is automatically converted to X i for comparison with the next adjacent grid. For the newly marked background grids, repeat the multi-modal feature extraction and adjacent grid comparison process to achieve the gradual diffusion annotation of the mask.

[0062] In S3, mark the grids in the orthoimage that do not have a fully annotated background mask as the building-background mixed areas, perform secondary annotation on the background mask of the building-background mixed areas, and suppress the background features of the background mask area after secondary annotation.

[0063] Identify the areas in the orthoimage where there are unlabeled background masks. These areas are usually located in the transition zones between buildings and the background, such as the edge zones where buildings are partially blocked by vegetation. By analyzing the multimodal mapping images of optical imagery and near-infrared data, identify these unlabeled areas (the areas that failed to spread during recursive diffusion labeling) as building-background mixed areas. These areas not only possess building features but are also affected by the surrounding complex background (such as vegetation, shadows, water bodies, etc.), and usually contain the spectral and texture features of the interweaving of buildings and the background. Extract the multimodal combined features of optical imagery and near-infrared data from the building-background mixed areas.

[0064] Compare the spectral features of each grid pixel in the building-background mixed area. Using the established spectral comparison feature library of background elements, conduct spectral feature hit detection for each pixel point in each grid. Calculate the cosine similarity of the features (simultaneously used as the feature attenuation ratio), judge the matching degree between the spectral features of each pixel point and the background element features, and mark the features with a matching degree exceeding the set standard as feature hits. According to the spectral feature hit results, perform secondary background mask labeling on the pixels in the building-background mixed area. For pixel points with a relatively high spectral similarity, mark them as background areas. For pixel points with a relatively low similarity, retain them as building areas. Through this step, accurately separate the transition zone between the building and the background to ensure the accuracy of the mask.

[0065] Based on the cosine similarity, perform feature weight attenuation on the multimodal combined features of the background mask pixel points after secondary labeling. Since the building-background mixed area contains the spectral features of complex background elements, it is necessary to weaken the feature expression of the background area. Adopt a feature attenuation strategy to reduce the weight of the background spectral features in the combined features, thereby weakening the influence of the background elements.

[0066] For the texture and light-shadow features of the background area, adopt low-pass filtering or other attenuation methods to reduce the interference of the background on the building. Use the orthoimage after background feature suppression processing as the initial analysis image of the building in the survey area for contour fitting.

[0067] In S4, adopt the Canny edge detection algorithm to extract the preliminary contour of the building, and perform multi-level progressive fitting of the B-Spline curve on the extracted building contour edge.

[0068] The Canny edge detection algorithm is used to extract the edges of building images in the surveying and mapping area. Through multi-stage processing, the Canny algorithm first performs Gaussian filtering to remove noise, then calculates the gradient intensity and direction, uses non-maximum suppression to refine the edges, and finally performs double-threshold detection to extract all potential edges in the image, especially the external edges of buildings. This step can effectively highlight the building contour and ignore background noise.

[0069] To further improve the recognition ability of building details, multi-scale convolutional feature extraction is introduced. By using convolutional kernels of different scales to perform convolutional operations on building images, local details and global morphological features of the images are extracted. This method can obtain the global contour and local complex shapes of buildings at different scales, enhancing the recognition ability of complex building structures, especially in parts such as edge details and complex roofs.

[0070] Based on the results of Canny edge detection and multi-scale feature extraction, the original contour of the building is simplified. By fitting polygons or smoothing curves, redundant edge points and details are removed to obtain a relatively concise preliminary contour. This simplification process mainly focuses on the structural features of the contour while retaining the basic features of the building form.

[0071] Based on the length and complexity of the original building contour, by analyzing the geometric morphology of the contour, representative key nodes are extracted. Usually, the position and distribution of the nodes reflect the complexity and geometric shape of the building contour. According to the geometric complexity of the contour, the parts with large local geometric changes are screened out and marked as "sparse control points". The specific annotation strategy for the control points is to calculate the contour curvature (second derivative) and find the parts with large curvature changes as control points. The expression is:

[0072]

[0073] where κ is the curvature (used to represent the set complexity), x(t), y(t) are the coordinates of the contour points, t is the actual parameter of the contour coordinates, x ′ (t), x″(t) and y ′ (t), y″(t) are the first derivative and second derivative of x(t), y(t) respectively.

[0074] This set of sparse control points will be used as the basis for subsequent B-Spline curve construction. Use this set of sparse control points to construct a preliminary B-Spline curve. The definition of the B-Spline curve is:

[0075]

[0076] where C(μ) is the B-Spline curve, N j,k(μ) is the B-Spline basis function, P j is the control point, k is the order of B-Spline, μ is a real number in the parameter domain, which determines the shape and position of the curve. The specific interval depends on the settings of the control point and B-Spline basis function.

[0077] The preliminary B-Spline curve is processed in segments, and the local curvature distribution of each segment is calculated. The curvature reflects the degree of bending of the curve, and the height of the local curvature can reflect the complexity of the building's form. Further, the error of each curve segment is calculated, that is, the distance error between each control point and its adjacent key nodes in the local area. This error value will be used as a standard for evaluating the accuracy of curve fitting. Based on the error calculation results, the threshold comparison method is used to judge the degree of detail fitting of each curve segment. If the error exceeds the preset tolerance threshold (0.5 pixels), it indicates that the degree of detail fitting in this area does not meet the standard and needs further optimization. For curve segments that meet the fitting accuracy requirements, the existing fitting structure is maintained.

[0078] For curve segments with unsatisfactory fitting accuracy, the number of control points is increased through control point interpolation to improve the fitting accuracy of the curve. Each time interpolation is performed, new control points are added equidistantly on the basis of the current curve segment, and the density of control points is gradually increased to refine the shape of the curve. On this basis, the order of the B-Spline curve can also be increased (from k to k+1) to make the curve fit the details of the building more accurately.

[0079] The process of adding control points and increasing the order is repeated in multiple levels until the detail fitting degree of all curve segments reaches the set standard. Each layer of fitting is performed on the basis of the previous layer, gradually optimizing the details of the contour to ensure the accuracy and smoothness of the building contour. When the detail fitting of all B-Spline curve segments reaches the predetermined accuracy requirement, the cycle is terminated. At this point, the obtained B-Spline curve accurately represents the shape of the building contour.

[0080] In S5, the building outline after B-Spline curve fitting is calibrated and the data layer is superimposed to form a comprehensive expression of buildings and geographical features.

[0081] Convert the spatial coordinates of the building outline in the two-dimensional coordinate system into coordinate data in the geographic coordinate system. To make these coordinates compatible with the existing geographic data layer, use coordinate projection transformation, such as converting from the geographic coordinate system (WGS84) to the local plane coordinate system. Combine the outline coordinate data of the building with relevant feature information to construct the attribute information of the building feature. For example, the height, use, structural type, etc. of the building can be part of the feature information. Associate this information with the outline coordinate data of each building to form a feature set containing spatial and attribute data.

[0082] Perform a unified coordinate projection transformation on all the fitted building outline coordinates so that the coordinates from different data sources are aligned in the same coordinate system. Through the projection transformation algorithm, convert the coordinate values of each building from the original space to the target space. By comparing the geographic coordinate information of different data sources, ensure that the building outlines are correctly matched in space and avoid geometric inconsistencies caused by errors or offsets.

[0083] Finally, overlay the outline data of the building with the existing geographic data layer. The geographic data layer includes geographic feature element data such as terrain, roads, vegetation, water bodies, etc., and the overlay operation can be performed through GIS (Geographic Information System). The spatial position and outline boundary of the building will be aligned with the geographic data layer to form a unified integrated data layer, supporting spatial analysis and image visualization display.

[0084] The above formulas are all dimensionless and take their numerical calculations. The formula is a formula obtained by collecting a large amount of data for software simulation to get the closest to the real situation. The preset parameters and threshold selection in the formula are set by those skilled in the art according to the actual situation.

[0085] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0086] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present application.

[0087] Those skilled in the art can clearly understand that for the sake of convenience and brevity of description, the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0088] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there can be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms.

[0089] The module described as a separation component may or may not be physically separated. The component shown as a module may or may not be a physical module, and it may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0090] In addition, in each embodiment of this application, each functional module can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0091] If the above functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of this application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0092] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0093] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An image data processing method for mapping and drawing, characterized in that, It includes the following steps: S1: Collect and extract the core features of the optical image data and near-infrared data of the surveyed building area, perform precise two-dimensional mapping on the core features to form a planar grid structure, measure the near-infrared reflection spectral characteristic data of non-building elements, and establish the spectral comparison characteristics of background elements; S2: Extract the spectral features in the grid, generate a background mask based on the comparison result between the spectral features and the spectral comparison characteristics of background elements, and segment the background area in combination with the dynamic segmentation model of region growing; S3: Mark the grids in the orthoimage that do not fully label the background mask as building-background mixed areas, perform secondary annotation on the background mask of the building-background mixed areas, and suppress the background features of the background mask area after secondary annotation; S4: Use the Canny edge detection algorithm to extract the preliminary contour of the building, and perform multi-level progressive fitting of the B-Spline curve on the extracted building contour edge; S5: Perform element calibration and data layer superposition on the building contour fitted based on the B-Spline curve to form a comprehensive expression of the building and geographical feature elements; In S4, using the Canny edge detection algorithm to extract the preliminary contour of the building and performing multi-level progressive fitting of the B-Spline curve on the extracted building contour edge specifically includes: Use the Canny edge detection algorithm to extract the edges of the initial analysis image of the building in the surveyed area, construct the structured scene contour information, and introduce multi-scale convolution feature extraction to identify the local details and global morphology of the building; According to the length and complexity of the original contour of the building, simplify the original contour to obtain key nodes, extract the key nodes in the parts with higher geometric complexity and mark them as a sparse control point set, and use these sparse control point sets to construct a preliminary B-Spline curve; Perform segmentation processing on the constructed preliminary B-Spline curve, calculate the local curvature distribution of each curve segment and the error between adjacent key nodes within the curve segment interval, and use the threshold comparison method based on the error calculation result to judge whether the detail fitting degree of the curve in this area meets the standard; In the curve segments where the detail fitting degree does not meet the standard, increase the number of control points by control point interpolation, perform multi-level loop fitting of the B-Spline curve, add an equal number of control points in each layer of fitting based on the previous layer, and at the same time increase the order of the B-Spline curve until the detail fitting degree of all segments of the B-Spline curve meets the standard and then end the loop to obtain the fitted building contour edge; 2. The image data processing method for mapping according to claim 1, wherein In S1, collecting and extracting the core features of the optical image data and near-infrared data of the surveyed building area, performing precise two-dimensional mapping on the core features to form a planar grid structure, measuring the near-infrared reflection spectral characteristic data of non-building elements, and establishing the spectral comparison characteristics of background elements specifically includes: Collect multi-modal data of the surveyed building area through a high-precision remote sensing platform, and the multi-modal data includes optical image data and near-infrared data; Through multi-view image matching and geocalibration, all modal data are unified into the same coordinate system and projection standard to form an orthoimage. Core feature extraction and feature alignment are performed on the optical image data and near-infrared data in the modal data. The aligned core features of different modalities are precisely two-dimensionally mapped in the orthoimage to form a planar grid structure, where each grid in the planar grid structure contains at least one pixel. The grids after modal feature mapping are analyzed grid by grid to label the core features of the grids. The core features include the RGB and texture features of buildings extracted from the optical image channel, and the reflectance spectral features of the near-infrared channel. Based on the multi-layer feature superposition strategy, the RGB, texture features, and spectral features are integrated into multi-modal combined features, where each modality represents a corresponding physical feature attribute expression. Select grids or groups of grids that contain all non-building object elements, measure the near-infrared reflectance spectral characteristic data of non-building elements, and establish a spectral reference feature for background elements.

3. The image data processing method for mapping according to claim 2, wherein In S2, the spectral features in the grids are extracted, and a background mask is generated based on the comparison result between the spectral features and the spectral reference feature for background elements. Combining with the dynamic segmentation model of region growing to segment the background region specifically includes: Select at least one spectral feature comparison point in each grid, and based on the hit situation of the spectral feature comparison point in the spectral reference feature for background elements, identify and generate a preliminary background mask for all non-building regions in the image. Select the grids with dense preliminary background mask annotations as the initial points, and recursively compare adjacent grids using the multi-modal combined features of the optical image and near-infrared data. Perform background mask diffusion annotation based on the feature similarity, where the feature similarity is the feature cosine similarity value that fuses texture, brightness, and spectral reflectance.

4. A method for processing image data for mapping according to claim 3, characterized in that, In S3, the grids in the orthoimage that have not been fully annotated with the background mask are marked as building-background mixed regions. Perform secondary annotation of the background mask on the building-background mixed regions, and suppress the background features of the region after secondary annotation of the background mask specifically includes: Obtain the grid combination in the orthoimage where there are regions without background mask annotation, mark it as a building-background mixed region, and extract the multi-modal combined features of the optical image and near-infrared data of all grids in the building-background mixed region. Based on the hit results of the spectral features of all pixel points in the grid in the spectral reference feature for background elements, perform secondary annotation of the background mask on the building-background mixed region. Based on the feature similarity when the spectral features hit the spectral reference feature for background elements, perform proportional feature weight attenuation on the multi-modal combined features at the pixel positions of the secondary annotation of the background mask, and weaken the feature expression of the background in the building-background mixed region. Use the orthoimage after background feature suppression processing as the initial analysis image of the buildings in the survey area for contour fitting.

5. A method for processing image data for mapping according to claim 3, characterized in that, In S5, perform feature calibration and data layer superposition on the building contours fitted based on the B-Spline curve to form a comprehensive expression of buildings and geographical feature elements specifically includes: Convert the coordinate data corresponding to the building contours into the element information of the buildings to establish a geographical data layer. Perform unified coordinate projection transformation and spatial position alignment on all the fitted building contour boundaries. At the same time, overlay the geographical data layer, add attribute information to the building objects, and form a comprehensive expression of the buildings and geographical feature elements.

Images