Method for displaying house type through fusion of three-dimensional model and aerial oblique photography

Abstract

The invention relates to the technical field of three-dimensional modeling, in particular to a method for displaying a house type through fusion of a three-dimensional model and aerial oblique photography, and the method comprises the steps: extracting building contour features of a target house through employing aerial image data, and carrying out the preliminary space registration of the building contour features of the target house; feature points are set according to the spatial position of the target house, house attribute checking is carried out by using each feature point and the building contour features, image partition information is extracted, the structure of the target house is divided according to the image partition information, and an iterative segmentation result is obtained; setting an iterative segmentation result as a plurality of graph units, and setting an extended progressive graph corresponding to the graph units; according to the extended progressive graph, performing statistics on overlapping parts of the graph units to obtain a house type difference quantity; performing three-dimensional stretching on the house type difference quantity to obtain different distribution of each house type, correcting the target house by using the different distribution of each house type, and recursively updating the display category of the target house; and the house display precision and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of 3D modeling, and specifically to a method for fusing 3D models with aerial oblique photography to display house floor plans. Background Art

[0002] Traditional house 3D modeling tasks are mainly achieved manually. Specifically, feature points are manually set on the floor-by-floor and household-by-household diagrams of the house to be modeled; based on the feature points in the floor-by-floor and household-by-household diagrams of each floor, manual calibration is performed; the 3D model of the house is generated using the data after manual calibration. The traditional method is easily affected by the surrounding environmental entities of the collected data in house floor plan display, resulting in insufficient modeling accuracy. At the same time, it is difficult to distinguish the differences in modeling of different houses, leading to the need to generate 3D models multiple times when performing house display, resulting in a reduction in display efficiency and accuracy.

[0003] For example, Chinese Patent Publication No. CN116580162A discloses a method, device, equipment, and storage medium for establishing a 3D model of a house. First, automatic calibration is performed on the floor-by-floor and household-by-household diagrams of the house to be modeled. Specifically, the initial floor-by-floor and household-by-household diagram of the bottommost floor is used as the initial calibration basis, and then for each non-bottommost current floor, the initial floor-by-floor and household-by-household diagram of the current floor is calibrated based on the target floor-by-floor and household-by-household diagram of the corresponding reference floor of the current floor to obtain the target floor-by-floor and household-by-household diagram of the current floor. Among them, the basis for the calibration process is that any line segment in the target floor-by-floor and household-by-household diagram of the current floor coincides with the associated line segment of the line segment in the target floor-by-floor and household-by-household diagram of the reference floor; by sequentially superimposing each target floor-by-floor and household-by-household diagram, the automatic superimposition task of the floor-by-floor and household-by-household data of the house to be modeled can be achieved; according to the attribute data of each floor, stretching processing is performed on the target floor-by-floor and household-by-household diagrams of each floor to generate the 3D model of the house to be modeled.

[0004] For example, Chinese Patent Publication No. CN118608381A discloses a method for stitching point cloud data of a house floor plan, including: dividing a house floor plan according to a preset survey station segmentation rule to obtain multiple measurement survey stations; constructing design feature data according to the design drawings of each measurement survey station and constructing scan feature data according to the point cloud data of the measurement survey station, and then matching the design feature data and scan feature data of each measurement survey station to obtain the house scaling ratio and the relative position relationship of the points in the point cloud data of each measurement survey station, and then using the relative position relationship of the measurement survey stations in the design drawings to perform point cloud stitching on the multiple measurement survey stations to obtain the point cloud model of the house floor plan.

[0005] In the prior art, there are descriptions of the image form for displaying stratified household division according to the line segments of each floor, and the combination of the positions of measurement stations to obtain the house type; these methods mainly display the static model of the house, and cannot display the differences in the display of each house in the three-dimensional model under the changes of the house over time or in other forms, resulting in the inability to dynamically update the display categories of each house, reducing the efficiency and relative accuracy of house display. Summary of the Invention

[0006] To solve the above technical problems, the technical solution adopted by the present invention is: a method for fusing a three-dimensional model and aerial oblique photography to display house types, including: S1, obtaining the aerial image data of the target house, extracting the architectural contour features of the target house using the aerial image data, and performing preliminary spatial registration on the architectural contour features of the target house.

[0007] S2, setting feature points based on the spatial position of the target house, using each feature point and the architectural contour features to check the house attributes, extracting multiple image partition information under oblique photography, dividing the structure of the target house based on the image partition information, and determining the iterative segmentation result at adjacent positions of the target house structure.

[0008] S3, setting the iterative segmentation result as multiple map units, querying the priority of the map units, selecting map units according to the priority, occurrence times, and area of the map units, constructing a connection channel between the selected map units, and setting an extended progressive map corresponding to the map units.

[0009] S4, statistically analyzing the overlapping parts of each map unit according to the extended progressive map to obtain the house type difference amount.

[0010] S5, performing three-dimensional stretching on the house type difference amount to obtain the distribution of each house type difference, using the distribution of each house type difference to correct the three-dimensional point cloud data of the target house, and recursively updating the display category of the target house.

[0011] The beneficial effects of the present invention are as follows: First, after aligning the target house, the present invention uses feature points and image partition information to describe the difference points and iterative segmentation results of the structure inside the current house, and can quickly generate a house structure diagram, improving the reusability of the three-dimensional model when generating the house structure.

[0012] Second, the present invention sets the iterative segmentation results as multiple graph units, queries the priorities of the graph units, and selects graph units according to the priorities, occurrence times, and areas of the graph units; it can query the associations between multiple house structures obtained by iterative segmentation and the connection relationships between different house structures, and generate an extended progressive graph corresponding to the graph units, which can comprehensively display the associations of the corresponding house structures and whether there are changes in the house structures within a preset period, reducing the problem that the accuracy of the house structure description in the house model is reduced due to environmental occlusion during generation and improving the display effect of the house itself image.

[0013] Third, the present invention constructs an extended progressive graph, counts the overlapping parts of each graph unit, and quickly calculates the house type difference through hierarchical division and analysis of the overlapping data volume; it can quickly quantify the house type difference, display each house type in different connection methods, which not only improves the setting accuracy of the 3D model but also enhances the comprehensiveness of the house type display, facilitating the tracking of the house status and preventing the occurrence of house risk situations; finally, through the distribution of its house type differences, it updates the house display categories, can describe the dynamic update effect of the house in aspects such as disaster warning, realizes the real-time display of each house type, and meets the requirements of house type display in various scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below with reference to the drawings and embodiments.

[0015] Figure 1 It is a schematic flowchart of a method for fusing a 3D model and aerial oblique photography to display house types.

[0016] Figure 2 It is a schematic flowchart of step S1 of a method for fusing a 3D model and aerial oblique photography to display house types.

[0017] Figure 3 It is a schematic flowchart of step S2 of a method for fusing a 3D model and aerial oblique photography to display house types.

[0018] Figure 4 It is a schematic flowchart of step S3 of a method for fusing a 3D model and aerial oblique photography to display house types.

[0019] Figure 5 It is a schematic flowchart of step S4 of a method for fusing a 3D model and aerial oblique photography to display house types.

[0020] Figure 6 It is a schematic flowchart of step S5 of a method for fusing a 3D model and aerial oblique photography to display house types. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] Embodiments of the present invention will be described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as a limitation to the present invention. For those not specified in the embodiments regarding specific technologies or conditions, they shall be carried out according to the technologies or conditions described in the literature in the field or according to the product specifications.

[0022] Referring to Figure 1 , a method for fusing a three-dimensional model and aerial oblique photography to display the house type, comprising: S1, obtaining the aerial image data of the target house, extracting the architectural contour features of the target house by using the aerial image data, and performing preliminary spatial registration on the architectural contour features of the target house.

[0023] S2, setting feature points based on the spatial position of the target house, using each feature point and the architectural contour features to check the house attributes, extracting multiple image partition information under oblique photography, dividing the structure of the target house according to the image partition information, and determining the iterative segmentation results at adjacent positions of the target house structure.

[0024] S3, setting the iterative segmentation results as multiple graphic units, querying the priorities of the graphic units, selecting graphic units according to the priorities, occurrence times and areas of the graphic units, constructing connection channels between the selected graphic units, and setting the extended progressive graphs corresponding to the graphic units.

[0025] S4, counting the overlapping parts of each graphic unit according to the extended progressive graph to obtain the house type difference amount.

[0026] S5, performing three-dimensional stretching on the house type difference amount to obtain the distribution of each house type difference, using the distribution of each house type difference to correct the three-dimensional point cloud data of the target house, and recursively updating the display category of the target house.

[0027] When obtaining the aerial image data of the target house, preprocessing the aerial image data, correcting the lens distortion using the camera calibration parameters, converting the aerial image data into sparse point clouds in the same spatial coordinate, and then determining the three-dimensional expression form of the target house for the sparse point clouds.

[0028] When performing preliminary spatial calibration, based on the actual environment of the area where the building is located, extracting multiple image control points and check points from the aerial image data one by one, performing preliminary spatial position calibration on the target house according to the spatial positions of the image control points, and at the same time evaluating whether the data after calibration processing can correspond to the current house type according to the check points.

[0029] At this time, the ground control points and check points can be adjusted to quickly verify the display results of the aerial image data. At the same time, according to the adjustment of the ground control points at the ground mark positions, the influence of vegetation and other obstacles on the model generation accuracy can be further reduced; the check points will represent the feature points of the marks during multiple image processes, and these feature points will represent the main recognized parts in the image processing. This part explains the key points and differences of the current house type display, facilitating the identification of whether the current house is a dangerous house, etc., and improving the accuracy of the house under multi-source integration.

[0030] As Figure 2 shown, the implementation method of step S1 further includes: S11, determining the perspectives of the aerial image data of the target house. Based on each perspective during shooting, the aerial image data of the target house is divided into multiple perspective partition information, and multiple ground control points and check points are arranged within each perspective partition information; the perspective partition information represents the image information of the current target house under the corresponding perspective. The ground control points are generally located at the edges of the building contour or significant feature points, such as the corners of the wall, the edge of the roof, and ground markers where the house contour and relative positions can be directly viewed. Further, the set ground control points are selected from the environmental entities around the target house and the house contour. The number of set ground control points can be 48 to obtain sufficient information about the surrounding environment at the location of the current target house and identify the relative positions of the target house and its surrounding area.

[0031] The check points are generally located at the key structures of the building, such as the centers of doors and windows, the turning points of stairs, or the house type difference areas, such as cracks and inclined walls. The check points will represent the prominent points in the house. By directly identifying this position, the structural entity of the entire house relative to the surrounding environment can be understood. Generally, the check points will be selected as different points from the ground control points. The number of selected check points can be 29 to obtain the integrity of the spatial alignment of the current target house under oblique shooting.

[0032] S12, obtaining the arrangement positions of the ground control points and check points under each perspective partition information, and aligning the positions of each perspective partition information according to the arrangement positions of the ground control points; the position alignment is achieved by calculating the distances between the pixel points within the perspective partition information and the ground control points, and determining that the calculated distances of the images under multiple perspectives are the same after the angle change, preventing the occurrence of offset in some positions. When aligning the positions of the check points, the distances between the ground control points and the check points are used to judge the relative positions of the current house, requiring that the distances between the ground control points and the check points in the images collected under multiple perspectives are the same.

[0033] S13. Use the aligned image control points and verification points as the building contour features of the target house to complete the preliminary spatial alignment of the target house. When the distances calculated between the image control points and verification points from multiple perspectives are the same, it is considered that the alignment is completed. These points that can maintain relative spatial consistency during multiple calibrations will be used for subsequent data verification of the target house. For example, the subsequent feature points will identify the distribution of each defect in spatial consistency according to the defects of the target house, and then identify the locations of these defects to find the differences between the current target house and the house standard model, so as to complete the rapid display of the house type based on these differences.

[0034] In an embodiment of the present invention, the iterative segmentation result shows what partition intersection points exist at the structural positions between the building contour of the target house and the pre-generated three-dimensional point cloud, indicating what forms of partitions and intersection point forms exist on the main structure of the house perimeter, so as to illustrate whether there are risk factors when the house changes, and whether the house is a dangerous house, and display the display forms of the house types in multiple scenarios.

[0035] When a risk occurs in the target house, due to the different internal and external representations, it will cause misidentification of the recognized image structure when generating its image structure. For example, when a wall crack appears due to a foundation problem, it mainly starts from the inside of the wall root. At this time, only for its external oblique photography image, it is easy to have the problem of incorrect risk identification, resulting in ignoring the problems existing in the house when displaying the house type, and causing corresponding problems. The feature points set at this time are the points that can assist in identifying whether there are defects in the house during house identification. These points are convenient for subsequent identification of whether there are differences between the current house and other houses in the current building; according to these differences, the differences between different house types can be quickly identified to reduce the problem of resource waste caused by multiple generations of its three-dimensional model when the difference points of the house type are too small. The house attributes represent various characteristic information of the house, such as the specific coordinates, geometric shape, contour area, etc. of the house.

[0036] Preferably, set the feature points according to the spatial position of the target house. The feature points represent the unique features of the house in terms of shape, size, height, structure, etc.; for example, some houses may have unique roof shapes, such as spire roofs, flat roofs, vaulted roofs, etc., or have special window layouts, balcony styles, etc., which can all be used as feature points to distinguish different houses.

[0037] Such as Figure 3As shown, the implementation of step S2 includes: S21. Establish the same number of split layers according to the set number of feature points. Each split layer contains at least one feature point and building outline features, which are used to compare whether there are differences under multiple layers into which the target house is split after the feature points are set. The differences may be reflected in aspects such as the shape, size, and structure of the house. For example, the outline of the house in a certain layer may be different from that in other layers, or the position and form of a certain feature point in the layer are different. After restoring the split layers, it is judged whether the information displayed in each layer is consistent with that before splitting to determine whether there are differences in the structure of the target house under each split layer; if there are differences, multiple image partition information is extracted to represent the positions where there are obvious differences in the current target house. If it is found that the restored information is inconsistent with that before splitting, or there are obvious differences between the split layers, it can be determined that there are differences in the structure of the target house under each split layer. The difference means that the structure of the house changes in the layers corresponding to different regions or different feature points, and this change may be caused by reasons such as damage, renovation, and measurement errors of the house itself. By extracting this information, the specific regions with differences in the house can be more accurately located; ultimately, the process of repeated modeling of house types can be reduced, and only the differences of each house need to be input into a standard model to complete the differential display of each house type.

[0038] S22. According to the feature points and building outline features of the split layer, obtain the target difference points corresponding to the split layer in reverse order, and retrieve the split layer with the target difference points to determine the house attributes corresponding to the target difference points. The house attributes represent the position, shape, and area of the target difference points.

[0039] S23. Based on the house attributes of the target difference points, restore the split layer to obtain the adjacent positions of the target difference points, and record the adjacent positions of the target difference points in the split layer to obtain image partition information.

[0040] Assume that the split layer is stored as a hierarchical data structure, such as a tree structure or a list form, with each layer corresponding to a layer. Compare the feature points and building outline features of each layer with those at other time points. When the difference exceeds 10%, it is recorded as a target difference point. These are compared sequentially after sorting each image in the aerial image data according to time; then, the positions and identifiers of the target difference points of each split layer are noted.

[0041] Starting from the topmost split layer, i.e., the last split layer, traverse layer by layer downward. Extract the information of the target difference points in each layer. The target difference points are represented by coordinates, shape features, and unique identifiers. During the traversal, filter according to the features of the target difference points to obtain multiple different types of target difference points. When filtering the target difference points, it is required that the target difference points be displayed one by one in the order of filtering.

[0042] After that, use the method of spatial index or mapping table to associate the target difference points with the house attributes to illustrate the position distribution and corresponding coordinates of the target difference points on the house. Finally, determine its position in the original complete layer according to the position and shape of the target difference point. Use image processing algorithms such as image fusion and boundary smoothing to merge the split layers and restore the area around the target difference point. During the merging process, ensure the topological relationships between houses, such as adjacent and inclusion, and keep them consistent. For each target difference point, analyze its adjacent area, such as in the form of 8-neighborhood or 4-neighborhood, to determine the adjacent positions; mark the adjacent positions in the split layer and record their attributes, such as whether they are other difference points and the distance from the target difference point. According to the marked information in the difference layer, divide the image into different regions. Each region corresponds to one or more target difference points and their adjacent positions; the attributes of each partition, such as region ID, the number of difference points included, and the total area, form the image partition information to complete the acquisition of the image partition information. The image partition information includes the boundaries of each split layer and the house attributes, such as house ID, area, and adjacent relationships.

[0043] After completing the acquisition of the image partition information, divide according to the structure of the image partition information within the target house to determine the significant differences of each image partition within the target house under the iteration of region growing for the content covered by the image partition information.

[0044] That is, the implementation method of the iterative segmentation result includes: dividing the house structure within each image partition information, starting from the components of the target house and the center point within the image partition information, and identifying the components of the target house according to the similarity criterion; at this time, the identification is based on similarity criteria such as color, texture, and gradient, and the features that can represent a single house structure are used as the starting point to identify the house structure in the aerial image data. At this time, the house structure includes information such as doors and windows, balconies, pipelines, house boundary lines, adjacent relationships between structures, and layouts. At this time, the identified house structure will be iterated to determine the house structure under the image partition information with differences, and the corresponding image data with the minimum boundary width at the adjacent positions after iteration to obtain the relatively accurate position of the difference points.

[0045] Retrieve the components of the target house using the image partition information to obtain the target house structure, which is used to characterize the content of the house structure with difference points in the image partition; perform scene integration based on the target house structure to obtain the iterative segmentation result.

[0046] After identifying the target house structure, describe the position corresponding to the house structure using multiple aerial image data to obtain the difference situation existing in the adjacent positions in the case of differences in the target house structure. This difference situation generally indicates that the house coordinates are offset, such as cracks appearing, differences in some features of the house compared to the difference points in the image during the previous image acquisition. These contents will be considered as the existing difference situation; combine these difference situations to illustrate the problem form that appears in the adjacent positions when the house structure has problems. Such as wall cracks, cracked layers, and the trend of cracks. These defects existing in the house under oblique photography using drones will be shown in the output iterative segmentation result.

[0047] In an embodiment of the present invention, after obtaining multiple problematic areas through the iterative segmentation result, identify the images of the house with risks, potential hazards, and progressive associations of each structure based on the images included in each map unit, record the visible areas and shaded parts in these images to further identify the differences existing between different houses, so as to obtain the difference distribution and extended images under progressive identification of different houses.

[0048] Such as Figure 4 shown, when implementing step S3, it further includes: S31, statistically analyze the data of each map unit according to a preset time period, and query the priority of each map unit according to the position of the map unit; the priority of the map unit will be set according to the position of the image corresponding to the map unit, and selected according to the number of occurrences of this position and the area corresponding to the map unit. When a house structure appears in each difference statistic and the recognized area gradually increases, it indicates that there are corresponding risks for this house type, or the same house may be partially blocked in the shootings at different times, resulting in inaccurate recognition of the house structure and other problems.

[0049] S32, select map units according to the priority, number of occurrences, and area of the map unit, and set at least one connection relationship between the selected map units. This connection relationship represents the spatial relationship, semantic relationship, etc. between the selected map units. The spatial relationship will represent the adjacent, inclusion, etc. relationships between the selected map units to illustrate whether there is a spatial correlation between the map units; the semantic relationship illustrates whether the positions of the house structures corresponding to the map units belong to the same type to illustrate what category the different parts of the target house belong to.

[0050] S33. Construct a connection channel based on the connection relationships of the selected graphic units, and count the graphic units as an extended progressive graph according to the connection order of the connection channel.

[0051] The extended progressive graph represents the connection relationships between the housing structures that show differences or risks within a preset time period under the connection relationships, and explains the structures with problems, which can assist in the subsequent display of the differences existing in the housing to comprehensively manage the housing.

[0052] Preferably, when selecting graphic units according to the priority, occurrence times, and area of the graphic units, query the occurrence times of the graphic units within a preset time period, then combine the area recognized each time of the graphic units according to the occurrence times, and then use the connection relationships between the graphic units to select each graphic unit according to time and priority, and complete the connection of its graphic units.

[0053] Preferably, the implementation method of selecting graphic units includes: querying the priorities of all graphic units, and determining the selection range according to the priorities and areas of the graphic units; the selection range explains the occurrence times of the same or similar housing structures retrieved under the value ranges of different priorities and the areas representing the structural differences of the graphic units, so as to identify the occurrence situations of the corresponding problems.

[0054] Set alternative graphic units according to the occurrence times of the graphic units within the selection range, and determine the selection probability of the alternative graphic units; at this time, the selection probability represents the probability value corresponding to the occurrence times of the graphic units within the selection range, and this probability value can be calculated by Gaussian distribution.

[0055] Calculate the area allocation ratio of each alternative unit according to the number of alternative graphic units, and the area allocation ratio represents the ratio of the area of the corresponding image of the corresponding graphic unit to the total area during statistics; select alternative graphic units in real time according to the area allocation ratio and the selection probability.

[0056] Preferably, the implementation method of selecting alternative graphic units in real time according to the area allocation ratio and the selection probability includes: comparing the selection probability with the selection probability threshold. When the selection probability is less than the selection probability threshold, classify the area allocation ratio according to the difference in the selection probability, and use the alternative graphic units corresponding to the classified area allocation ratio as the selected alternative graphic units.

[0057] When the selection probability is greater than the selection probability threshold, select the alternative graphic units with connection relationships in descending order of the value of the selection probability in turn to complete the selection of the graphic units.

[0058] The above content shows that when the number of occurrences of a certain housing difference point is small, output it according to the area of its difference. If the number of occurrences is large, select it according to its connection relationship to illustrate the corresponding relationship of the target housing when differences occur.

[0059] In an embodiment of the present invention, as Figure 5 shown, the implementation manner of step S4 includes: S41, counting the overlapping frequencies of each map unit, and dividing the extended progressive map into multiple levels based on the overlapping frequencies.

[0060] S42, determining the housing type difference amount according to the overlapping data volume within each level.

[0061] At this time, after obtaining the extended progressive map of the positions where the houses have differences changing with time and priority, calculate the overlapping parts of these continuously updated parts, and explain the differences between different housing types according to the overlapping parts. It is also possible to use the values corresponding to the pixel points during overlapping, and count the number of pixel points with different values to explain the housing type difference amount.

[0062] In the extended progressive map, each map unit represents a specific housing area or housing type; these housing areas will display the changes of the housing areas represented by the map units in a way of connecting multiple images progressively according to a pre-selected time period, the appearance times of the map units, area, priority, etc.

[0063] After that, count the overlapping parts, and use the overlapping data volume as the main content of the subsequent output, such as counting indicators such as the overlapping area and overlapping ratio between each map unit to quantify the overlapping degree between the map units; calculate the housing type difference amount by comparing the changes in the overlapping area and the changes in the housing type structure between the housing layers at different time points or from different sources. The final housing type difference amount can be expressed as the change amount of the overlapping area or in the form of a change ratio to describe the overlapping parts between multiple map units.

[0064] By counting the overlapping parts of the map units, the changes in the housing types of the houses at different time points or from different sources can be identified; this helps to understand information such as the usage status and renovation history of the houses. At the same time, by counting the overlapping parts, it is also possible to verify whether the data between different data sources is consistent. If data inconsistency is found, it can be corrected or adjusted in time to improve the accuracy and reliability of the data. According to the situation of this part of the difference amount, it helps to perform subsequent maintenance on the houses, and in combination with the form of the difference amount, quickly generate a three-dimensional model of the current house to improve the accuracy of the subsequent model generation of the house.

[0065] In an embodiment of the present invention, when performing three-dimensional stretching, as Figure 6As shown, step S5 includes the following implementation methods: S51, extract the spatial semantic information of each housing type difference amount, obtain at least one external boundary line corresponding to the target house, and perform multiple position comparisons on the interior of the house by combining the external boundary line with the spatial semantic information to determine the position comparison trend of each external boundary line. The external boundary line represents the line of the overall contour of the house and defines the overall contour of the housing type difference amount in the house area. The spatial semantic information of the housing type difference amount described here refers to the connection relationship between corresponding map units when extracting the housing type difference amount. The spatial relationship and semantic relationship in the connection relationship of the map units are used as the spatial semantic information to obtain the contour of the housing type difference amount corresponding to the house area. The position comparison trend of each external boundary line explains the position corresponding to the internal boundary of the house. For the data inside the house, a pre-set three-dimensional standard model can be used to obtain it, or after obtaining the internal data of the house by scanning, identify the position of the external boundary line relative to the interior to explain its position comparison trend; the position comparison trend will convert its coordinates into a vector form to explain the trend situation.

[0066] S52, based on the position comparison trend of each external boundary line, identify the relative position relationship between each external boundary line and the internal boundary line of the target house, determine the distance matrix under the relative position relationship, and verify the relative distances and distribution conditions of each boundary line with the distance matrix. The verified relative distances and distribution conditions of each boundary line are output as the housing type difference distribution. The internal boundary line represents the dividing line between each structure in the house and is used to describe the line segment situation of dividing the structure in the area corresponding to the housing type difference amount, so as to explain whether the spatial positions are accurately consistent when the housing type difference is identified.

[0067] When subsequently correcting the three-dimensional point cloud data of the target house, S53, splice each internal boundary line and external boundary line to generate a three-dimensional model corresponding to the target house, and correct the three-dimensional point cloud data with the housing type difference distribution under the three-dimensional model, and recursively update the display category of the target house with the plane where the target house is located during the correction.

[0068] After the above content uses the internal boundary line and the external boundary line to splice and form a three-dimensional model, map the data corresponding to each housing type difference distribution to the three-dimensional model to explain the different parts corresponding to the three-dimensional model. Then continuously update the type of the current house display according to the corrected content.

[0069] Preferably, verifying the relative distances and distribution conditions of each boundary line with the distance matrix means whether the distances between each internal boundary line and each external boundary line are consistent when using the data of multiple images for identification, to judge whether the currently obtained data is accurate. And when the data is verified to be accurate, use the distance matrix and the corresponding external boundary line and internal boundary line as the output housing type difference distribution.

[0070] Preferably, the implementation method of recursively updating the display category of the target house based on the plane where the target house is located during correction includes: extracting the house type boundary of the plane where the target house is located, and verifying whether the connection position is consistent with the orientation of the target house at the connection positions of all the walls on the house type boundary; if they are consistent, determining the overlapping line segments of the internal boundary line and the external boundary line of the target house, and recursively updating the display category of the target house according to the overlapping line segments.

[0071] The above description shows that when recursively updating the current target house, the wall information to be updated or queried is found based on the plane where the target house is located, and the positions of the line segments on these walls are compared with the current house to find the data that needs to be updated on the current house. At the same time, it is verified whether these areas to be updated belong to its internal boundary or external boundary. Finally, the parts of the house with differences are updated to show the differences in house type display on different planes. Eliminate the influence of the environment on the multi-size mapping of aerial images, realize the association and matching processing of the spatial semantics of each house type, and finally improve the generation speed of its 3D model and the accuracy of house type display.

[0072] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention, and still be covered by the protection scope of the present invention.

Claims

1. A method for displaying house types by integrating a three-dimensional model with aerial oblique photography, characterized in that: include: S1, obtaining aerial image data of a target house, extracting building outline features of the target house using the aerial image data, and performing preliminary spatial registration of the building outline features of the target house; S2, setting feature points based on the spatial position of the target house, using each feature point and the building outline feature to check the house attributes, extracting multiple image partition information under oblique photography, dividing the structure of the target house based on the image partition information, and determining the iterative segmentation results of the target house structure at adjacent positions; S3, setting the iterative segmentation result to multiple graph units, querying the priority of the graph unit, selecting the graph unit according to the priority, number of occurrences and area of ​​the graph unit, building a connection channel between the selected graph unit, and setting the extended progressive graph corresponding to the graph unit; the priority of the graph unit is set according to the position of the graph unit corresponding to the image, and the graph unit is selected according to the number of occurrences of the position and the area corresponding to the graph unit; S4, counting the overlapping parts of each graph unit according to the extended progressive graph to obtain the difference in apartment types; S5, three-dimensionally stretching the apartment type difference amount to obtain the difference distribution of each apartment type, using the difference distribution of each apartment type to correct the three-dimensional point cloud data of the target house, and recursively updating the display category of the target house.

2. The method for displaying house types by integrating a three-dimensional model with aerial oblique photography according to claim 1, characterized in that: The implementation of step S1 also includes: S11, determining the viewing angle of the aerial image data of the target house, dividing the aerial image data of the target house into a plurality of viewing angle partition information based on each viewing angle during shooting, and arranging a plurality of image control points and check points in each viewing angle partition information; S12, obtaining the layout positions of the image control points and the check points under each viewing angle partition information, and aligning the positions of each viewing angle partition information according to the layout positions of the image control points; S13, using the aligned image control points and check points as the building outline features of the target house, completing the preliminary spatial alignment of the target house.

3. The method for displaying house types by integrating a three-dimensional model with aerial oblique photography according to claim 1, characterized in that: The implementation of step S2 includes: S21, establishing an equal number of split layers according to the set number of feature points, wherein the split layers contain at least one feature point and a building outline feature; S22, according to the feature points of the split layer and the building outline features, the target difference points under the corresponding split layer are obtained in reverse order, the split layer is searched with the target difference points, and the house attributes corresponding to the target difference points are determined; S23, performing split layer restoration based on the house attributes of the target difference point to obtain the adjacent position of the target difference point, and recording the adjacent position of the target difference point in the split layer to obtain image partition information.

4. The method for displaying house types by integrating a three-dimensional model with aerial oblique photography according to claim 1, characterized in that: The implementation methods of iterative segmentation results include: The house structure in each image partition information is divided, and the components of the target house and the center point in the image partition information are used as starting points to identify the components of the target house according to the similarity criterion; The components of the target house are retrieved using the image partition information to obtain the target house structure, which is used to characterize the content of the house structure with differences in the image partitions. Scene integration is performed based on the target house structure to obtain an iterative segmentation result.

5. The method for displaying house types by integrating a three-dimensional model with aerial oblique photography according to claim 1, characterized in that: Step S3 also includes: S31, collecting statistics of data of each graph unit according to a preset time period, and querying the priority of each graph unit according to the position of the graph unit; S32, selecting graph units according to their priorities, occurrence times and areas, and setting at least one connection relationship between the selected graph units; S33, constructing a connection channel based on the connection relationship of the selected graph units, and counting the graph units into an extended progressive graph according to the connection sequence of the connection channel.

6. The method for displaying house types by integrating a three-dimensional model with aerial oblique photography according to claim 5, characterized in that: The implementation methods of selecting graph units include: Query the priorities of all graph units and determine the selection range based on the priorities and areas of the graph units; According to the number of occurrences of the graph unit in the selection range, the candidate graph unit is set, and the selection probability of the candidate graph unit is determined; According to the number of candidate map units, the area allocation ratio of each candidate unit is calculated, and the candidate map unit is selected in real time according to the area allocation ratio and the selection probability.

7. The method for displaying house types by integrating a three-dimensional model with aerial oblique photography according to claim 6, characterized in that: The implementation methods of selecting candidate map units in real time according to the area allocation ratio and the selection probability include: The selection probability is compared with the selection probability threshold. When the selection probability is less than the selection probability threshold, the area allocation ratio is classified according to the selection probability difference, and the candidate map unit corresponding to the area allocation ratio after classification is selected as the candidate map unit; When the selection probability is greater than the selection probability threshold, candidate graph units with connection relationships are selected in order from large to small according to the selection probability values.

8. The method for displaying house types by integrating a three-dimensional model with aerial oblique photography according to claim 1, characterized in that: The implementation of step S4 includes: S41, counting the overlapping frequency of each graph unit, and dividing the extended progressive graph into multiple levels based on the overlapping frequency; S42, determining the amount of apartment type differences according to the amount of overlapping data in each layer.

9. The method for displaying house types by integrating a three-dimensional model with aerial oblique photography according to claim 1, characterized in that: Step S5 includes the following implementations: S51, extracting spatial semantic information of the difference between each apartment type, obtaining at least one external boundary line corresponding to the target house, and performing multiple position comparisons of the external boundary line with the interior of the house in combination with the spatial semantic information to determine the position comparison direction of each external boundary line; S52, comparing the directions of the positions of the external boundary lines, identifying the relative position relationship between each external boundary line and the internal boundary line of the target house, determining a distance matrix under the relative position relationship, and verifying the relative distance and distribution of each boundary line with the distance matrix, and outputting the verified relative distance and distribution of each boundary line as the distribution of house type differences; S53, stitching the internal boundary lines and the external boundary lines to generate a three-dimensional model corresponding to the target house, and correcting the three-dimensional point cloud data according to the difference distribution of each apartment type under the three-dimensional model, so as to recursively update the display category of the target house when the plane where the target house is located is corrected.

10. The method for displaying house types by integrating a three-dimensional model with aerial oblique photography according to claim 9, characterized in that: The implementation method of recursively updating the display category of the target house based on the plane where the target house is located during correction includes: Extract the apartment boundary of the plane where the target house is located, and use the connection positions of all walls on the apartment boundary to verify whether the connection positions are consistent with the orientation of the target house; if they are consistent, determine the overlapping line segments of the internal boundary line and the external boundary line of the target house, and recursively update the display category of the target house according to the overlapping line segments.

Images