A method and system for analyzing aerial images of unmanned aerial vehicles
By using traditional image processing techniques and drone aerial image analysis to extract the core area of buildings and perform texture comparison, the problem of distinguishing between building ancillary structures and road surfaces in rural areas has been solved, achieving efficient road recognition.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHENGDU RAINPOO TECH CO LTD
- Filing Date
- 2026-05-21
- Publication Date
- 2026-06-19
AI Technical Summary
Existing drone aerial photo analysis methods struggle to effectively distinguish between building attachments and actual road surfaces in rural areas. Deep learning methods have high computational costs, while traditional machine learning methods are prone to confusing attachments with actual road surfaces.
Using traditional image processing techniques, the initial building mask is obtained from UAV aerial photos. The core area of the building is extracted and expanded. Connectivity analysis and texture comparison are used to identify the texture similarity of candidate connected components and generate road and building regions.
It effectively avoids the problem of misclassifying auxiliary structures as roads, has low computational requirements, good adaptability, and is suitable for direct implementation by UAV equipment.
Smart Images
Figure CN122244735A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent image recognition technology, specifically to a method and system for analyzing aerial photographs taken by unmanned aerial vehicles (UAVs). Background Technology
[0002] With the popularization of drone technology, using drone aerial photographs to quickly extract rural road information has become an important means of rural road network surveys and navigation map updates. However, buildings in rural areas are densely packed, and there are also ancillary structures such as balconies, terraces, and awnings that protrude from the main buildings. In orthophotos, these ancillary structures are easily confused with the actual road surface. Among existing road extraction methods, deep learning methods require a large amount of pixel-level labeled data, resulting in high computational costs, while traditional machine learning methods are prone to confusing ancillary structures with the actual road surface.
[0003] In the prior art, Chinese patent application number CN202210994482.1 discloses a road recognition method, apparatus, device, and storage medium based on remote sensing images. The method includes: acquiring a remote sensing image of a remotely located area to be identified; determining a suspected road area image based on the pixels and road features of the remote sensing image; preprocessing the suspected road area image based on preset road soil information and spatial resolution to obtain a current suspected road area image; detecting the current suspected road area image using a target edge recognition algorithm to obtain road surface information and road edge contour information; and generating a target remote road based on the road surface information and road edge contour information. This method, which determines a suspected road area image based on pixels and road features, preprocesses the suspected road area image, and then generates a target remote road based on the road surface information and road edge contour information, is suitable for identifying remote roads but is difficult to apply to densely populated rural areas. Summary of the Invention
[0004] In order to at least overcome the above-mentioned shortcomings in the prior art, the purpose of this application is to provide a method and system for analyzing unmanned aerial vehicle (UAV) aerial photographs.
[0005] In a first aspect, embodiments of this application provide a method for analyzing aerial photographs of unmanned aerial vehicles (UAVs), including:
[0006] Acquire drone aerial photographs of the target area and obtain the initial building mask from the drone aerial photographs;
[0007] The area between adjacent initial building masks is used as the initial road candidate area, and the building core area is extracted based on the initial building masks;
[0008] The core area of the building is expanded, and the intersection between the expanded area and the initial road candidate area is used as the auxiliary structure candidate area.
[0009] The candidate regions of the subordinate structures are divided into multiple candidate connected regions through connected component analysis;
[0010] Each candidate connected component is compared with the texture of the corresponding candidate building core region, and the candidate connected components with a texture similarity exceeding a preset value are selected as building regions; the candidate building core region is the building core region adjacent to the candidate connected component.
[0011] The road area and building area are generated based on the core building area and the selected building area.
[0012] In one possible implementation, the extraction of the building's core area includes:
[0013] The initial building mask is eroded and expanded with a first radius to form the core area of the building; the first radius is determined based on the thickness of the building walls.
[0014] In one possible implementation, forming the core area of the building includes:
[0015] Construct a first circular structural element with a radius of a first radius, and slide the center of the first circular structural element pixel by pixel in the initial building mask;
[0016] When sliding, if there are no pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a candidate pixel; if there are pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a non-candidate pixel.
[0017] The center of the first circular structural element is slid across the non-candidate pixels and the candidate pixels. When there are pixels of the initial building mask in the first circular structural element, the pixel corresponding to the center of the first circular structural element is marked as the core pixel.
[0018] All core pixels are combined to form the core area of the building.
[0019] In one possible implementation, the candidate regions for the subordinate structure include:
[0020] The edge of the building core area facing the initial road candidate area is extended towards the initial road candidate area by a second radius to form an expanded core area; the second radius is determined according to the road width of the target area.
[0021] The overlapping area between the expansion core region and the initial road candidate region is used as the auxiliary structure candidate region.
[0022] In one possible implementation, the formation of multiple candidate connected components includes:
[0023] The pixels in the candidate region of the auxiliary structure are traversed and searched.
[0024] When any pixel's neighboring pixels are also in the candidate region of the subordinate structure, these pixels are divided into the same candidate connected region.
[0025] In one possible implementation, obtaining the selected building area includes:
[0026] The images of the candidate connected components are converted to the LAB color space to form a connected component histogram, and the images of the corresponding candidate building core regions are converted to the LAB color space to form a core histogram.
[0027] Calculate the Barcol distance between the distribution of the connected component histogram and the distribution of the core histogram, and determine the candidate connected component as the selected building region when the Barcol distance is lower than a preset value.
[0028] In one possible implementation, generating the road area and the building area based on the building core area and the selected building area includes:
[0029] The core building area and the adjacent selected building areas are merged to form a building area;
[0030] The area between adjacent building areas is defined as the road area.
[0031] Based on the same inventive concept, this application also provides an unmanned aerial vehicle (UAV) aerial photograph analysis system, comprising:
[0032] The identification unit is configured to acquire drone aerial photographs of the target area and acquire an initial building mask in the drone aerial photographs;
[0033] The core unit is configured to use the area between adjacent initial building masks as an initial road candidate area, and extract the building core area based on the initial building masks;
[0034] An expansion unit is configured to expand the core area of the building and use the intersection between the expanded area and the initial road candidate area as the auxiliary structure candidate area;
[0035] The connectivity unit is configured to divide the candidate region of the subordinate structure into multiple candidate connectivity regions through connectivity analysis;
[0036] The comparison unit is configured to compare the texture of each candidate connected component with the texture of the corresponding candidate building core region, and select the candidate connected components whose texture similarity exceeds a preset value as the selected building region; the candidate building core region is the building core region adjacent to the candidate connected component.
[0037] The generation unit is configured to generate a road area and a building area based on the building core area and the selected building area.
[0038] In one possible implementation, the core unit is further configured as follows:
[0039] The initial building mask is eroded and expanded with a first radius to form the core area of the building; the first radius is determined based on the thickness of the building walls.
[0040] In one possible implementation, the core unit is further configured as follows:
[0041] Construct a first circular structural element with a radius of a first radius, and slide the center of the first circular structural element pixel by pixel in the initial building mask;
[0042] When sliding, if there are no pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a candidate pixel; if there are pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a non-candidate pixel.
[0043] The center of the first circular structural element is slid across the non-candidate pixels and the candidate pixels. When there are pixels of the initial building mask in the first circular structural element, the pixel corresponding to the center of the first circular structural element is marked as the core pixel.
[0044] All core pixels are combined to form the core area of the building.
[0045] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0046] This invention provides a method and system for analyzing aerial photographs taken by unmanned aerial vehicles (UAVs). Based on traditional image processing techniques, it does not rely on deep learning models, has low computational requirements, and is easy to implement directly on UAV equipment. At the same time, it effectively avoids the problem of misclassifying auxiliary structures as roads, and has good adaptability. Attached Figure Description
[0047] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:
[0048] Figure 1 This is a schematic diagram of the method steps in an embodiment of this application;
[0049] Figure 2 This is a schematic diagram of township road identification in an embodiment of this application. Detailed Implementation
[0050] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.
[0051] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0052] Please refer to the following: Figure 1 This is a flowchart illustrating a UAV aerial photograph analysis method provided in an embodiment of the present invention. Further, the UAV aerial photograph analysis method may specifically include the content described in steps S1-S6.
[0053] S1: Acquire drone aerial photos of the target area and obtain the initial building mask from the drone aerial photos;
[0054] S2: The area between adjacent initial building masks is taken as the initial road candidate area, and the building core area is extracted based on the initial building masks;
[0055] S3: Expand the core area of the building and take the intersection between the expanded area and the initial road candidate area as the auxiliary structure candidate area;
[0056] S4: Divide the candidate region of the subordinate structure into multiple candidate connected regions through connected component analysis;
[0057] S5: Compare the texture of each candidate connected component with the texture of the corresponding candidate building core region, and select the candidate connected components whose texture similarity exceeds a preset value as the selected building region; the candidate building core region is the building core region adjacent to the candidate connected component;
[0058] S6: Generate a road area and a building area based on the core building area and the selected building area.
[0059] In implementing this application embodiment, it is first necessary to acquire corresponding drone aerial photographs of the target area using a drone. Orthophotos are preferred. Then, building identification is performed using common image recognition methods to form an initial building mask. It should be understood that the initial building mask can be formed using morphological building indices or some traditional machine learning algorithms such as support vector machines and random forest algorithms, which are very mature technologies and are not limited in this application. This mask covers the initially identified buildings. Please refer to [link to relevant documentation]. Figure 2 The diagram illustrates the initial identification process. It shows that a light green initial building mask covers the initially identified buildings, and most of the area between these initial building masks represents rural roads. However, it also shows that some building attachments, as indicated by the red box in the diagram, are easily misidentified as non-buildings during the initial identification process due to their elongated shape. Therefore, this embodiment aims to accurately identify these areas that resemble roads but are not roads, thereby accurately identifying the actual roads.
[0060] In this embodiment, the UAV aerial photograph is generally an RGB three-channel image. The identified initial building mask is marked as 1 in the mask layer, while the area between adjacent identified initial building masks is marked as 0, i.e., the initial road candidate area. However, in scientific practice, the inventors discovered that, for the identified initial building masks, since most rural houses are self-built, they often have many rough edges and corners. Therefore, the initial building masks need to be pre-processed to remove these elements and form the core building area. By expanding the core building area outwards, it can cover areas that may be roads and the areas of attached buildings identified as roads.
[0061] In this embodiment, for candidate regions of ancillary structures, it is necessary to segment them into multiple candidate connected regions for sequential discrimination. Therefore, connected component analysis is adopted, treating connected regions as independent regions for subsequent identification. In scientific practice, the inventors discovered that in rural areas, there are certain differences in texture between ancillary buildings and road areas, but these differences vary in different locations. Furthermore, the inventors found that for self-built houses in rural areas, the main building and ancillary buildings generally use similar waterproofing and top-floor paving. This results in similar textures such as moss and water accumulation on the tops of the main building and ancillary buildings after prolonged use. This similarity is completely different from that of roads, which are frequently walked on. Therefore, in this embodiment, comparing the textures of candidate connected regions with the corresponding candidate building core regions allows for more accurate identification of the actual ancillary buildings. Based on the identified ancillary buildings and building core regions, the actual rural roads can be accurately identified.
[0062] In one possible implementation, the extraction of the building's core area includes:
[0063] The initial building mask is eroded and expanded with a first radius to form the core area of the building; the first radius is determined based on the thickness of the building walls.
[0064] In one possible implementation, forming the core area of the building includes:
[0065] Construct a first circular structural element with a radius of a first radius, and slide the center of the first circular structural element pixel by pixel in the initial building mask;
[0066] When sliding, if there are no pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a candidate pixel; if there are pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a non-candidate pixel.
[0067] The center of the first circular structural element is slid across the non-candidate pixels and the candidate pixels. When there are pixels of the initial building mask in the first circular structural element, the pixel corresponding to the center of the first circular structural element is marked as the core pixel.
[0068] All core pixels are combined to form the core area of the building.
[0069] In the implementation of this application embodiment, the core area of the building is extracted by first processing the burrs on the outer boundary and then restoring the boundary. The first radius can be selected as the thickness of the building wall or half of the thickness of the building wall, which can be selected according to the actual local conditions. After traversing the pixels of the initial building mask by constructing the first circular structural element, the protruding burrs will be removed, but some of the outermost wall will also be removed. Therefore, the same first circular structural element needs to be used to restore the area again, so as to realize the extraction of the core area of the building.
[0070] In one possible implementation, the candidate regions for the subordinate structure include:
[0071] The edge of the building core area facing the initial road candidate area is extended towards the initial road candidate area by a second radius to form an expanded core area; the second radius is determined according to the road width of the target area.
[0072] The overlapping area between the expansion core region and the initial road candidate region is used as the auxiliary structure candidate region.
[0073] In the implementation of this application embodiment, after determining the candidate area of the auxiliary structure, a second radius greater than the first radius is used for expansion. It should be understood that in this application embodiment, the building core area is expanded towards the boundary of the initial road candidate area in the normal direction. The reason for this is that half of the normally formed building core area is a quadrilateral area. If it is expanded by the traditional expansion method, it will connect the two expanded areas at the corners of the building core area, which is not conducive to the subsequent segmentation of the candidate area of the auxiliary structure. Generally speaking, the second radius can be the statistical average of the road width of the target area, or it can be determined according to the possible width of the auxiliary building. This application embodiment does not impose any limitations.
[0074] In one possible implementation, the formation of multiple candidate connected components includes:
[0075] The pixels in the candidate region of the auxiliary structure are traversed and searched.
[0076] When any pixel's neighboring pixels are also in the candidate region of the subordinate structure, these pixels are divided into the same candidate connected region.
[0077] In the implementation of this application embodiment, when determining candidate connected components, candidate connected components can be effectively generated by judging the positional relationship between adjacent pixels. The adjacent refers to the eight pixels adjacent to the pixel, or it can be four pixels in the horizontal and vertical directions. This application embodiment does not impose any limitations.
[0078] In one possible implementation, obtaining the selected building area includes:
[0079] The images of the candidate connected components are converted to the LAB color space to form a connected component histogram, and the images of the corresponding candidate building core regions are converted to the LAB color space to form a core histogram.
[0080] Calculate the Barcol distance between the distribution of the connected component histogram and the distribution of the core histogram, and determine the candidate connected component as the selected building region when the Barcol distance is lower than a preset value.
[0081] In the implementation of this application embodiment, since the UAV aerial images are actually acquired as RGB three-channel images, which are not conducive to texture recognition, it is necessary to first convert them to the LAB color space. The LAB color space can separate the luminance of the L channel and the color channels A and B, thereby effectively performing texture analysis. In specific analysis, the data of the three channels L, A, and B need to be quantized, that is, the data is segmented, with each channel forming n data intervals, and the final data interval series is n cubed; then the connected component histogram and the core histogram are normalized to form the probability distribution of the data; then the distance between the distribution of the normalized connected component histogram and the distribution of the core histogram is calculated. This application embodiment uses Bach distance for measurement, as shown in the following formula:
[0082]
[0083] In the formula, D B Here, K represents the Bach distance, and K is the series of the data interval. Let be the normalized value of the j-th data interval in the connected component histogram. The normalized value for the j-th data interval of the core histogram. The mean of the histogram of connected components. The mean of the core histogram is used. The Bach distance ranges from [0,1]. The smaller the value, the more similar the distributions of the two histograms are. When the Bach distance is lower than the preset value, the two histograms are considered similar. In this case, the candidate connected component is selected as the building region. The preset value is generally 0.25.
[0084] In one possible implementation, generating the road area and the building area based on the building core area and the selected building area includes:
[0085] The core building area and the adjacent selected building areas are merged to form a building area;
[0086] The area between adjacent building areas is defined as the road area.
[0087] Based on the same inventive concept, this application also provides an unmanned aerial vehicle (UAV) aerial photograph analysis system, comprising:
[0088] The identification unit is configured to acquire drone aerial photographs of the target area and acquire an initial building mask in the drone aerial photographs;
[0089] The core unit is configured to use the area between adjacent initial building masks as an initial road candidate area, and extract the building core area based on the initial building masks;
[0090] An expansion unit is configured to expand the core area of the building and use the intersection between the expanded area and the initial road candidate area as the auxiliary structure candidate area;
[0091] The connectivity unit is configured to divide the candidate region of the subordinate structure into multiple candidate connectivity regions through connectivity analysis;
[0092] The comparison unit is configured to compare the texture of each candidate connected component with the texture of the corresponding candidate building core region, and select the candidate connected components whose texture similarity exceeds a preset value as the selected building region; the candidate building core region is the building core region adjacent to the candidate connected component.
[0093] The generation unit is configured to generate a road area and a building area based on the building core area and the selected building area.
[0094] In one possible implementation, the core unit is further configured as follows:
[0095] The initial building mask is eroded and expanded with a first radius to form the core area of the building; the first radius is determined based on the thickness of the building walls.
[0096] In one possible implementation, the core unit is further configured as follows:
[0097] Construct a first circular structural element with a radius of a first radius, and slide the center of the first circular structural element pixel by pixel in the initial building mask;
[0098] When sliding, if there are no pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a candidate pixel; if there are pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a non-candidate pixel.
[0099] The center of the first circular structural element is slid across the non-candidate pixels and the candidate pixels. When there are pixels of the initial building mask in the first circular structural element, the pixel corresponding to the center of the first circular structural element is marked as the core pixel.
[0100] All core pixels are combined to form the core area of the building.
[0101] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0102] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices or units, or may be electrical, mechanical or other forms of connection.
[0103] The units described as separate components may or may not be physically separate. As will be apparent to those skilled in the art, the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0104] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0105] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the 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 to cause a computer device (which may be a personal computer, server, or grid device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0106] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for analyzing aerial photographs taken by unmanned aerial vehicles, characterized in that, include: Acquire drone aerial photographs of the target area and obtain the initial building mask from the drone aerial photographs; The area between adjacent initial building masks is used as the initial road candidate area, and the building core area is extracted based on the initial building masks; The core area of the building is expanded, and the intersection between the expanded area and the initial road candidate area is used as the auxiliary structure candidate area. The candidate regions of the subordinate structures are divided into multiple candidate connected regions through connected component analysis; Each candidate connected component is compared with the texture of the corresponding candidate building core region, and the candidate connected components whose texture similarity exceeds a preset value are selected as building regions; The candidate building core area is the building core area adjacent to the candidate connected domain; The road area and building area are generated based on the core building area and the selected building area.
2. The method for analyzing UAV aerial photographs according to claim 1, characterized in that, The extraction of the core area of the building includes: The initial building mask is eroded and expanded with a first radius to form the core area of the building; the first radius is determined based on the thickness of the building walls.
3. The method for analyzing UAV aerial photographs according to claim 2, characterized in that, The core area of the building includes: Construct a first circular structural element with a radius of a first radius, and slide the center of the first circular structural element pixel by pixel in the initial building mask; When sliding, if there are no pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a candidate pixel; if there are pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a non-candidate pixel. The center of the first circular structural element is slid across the non-candidate pixels and the candidate pixels. When there are pixels of the initial building mask in the first circular structural element, the pixel corresponding to the center of the first circular structural element is marked as the core pixel. All core pixels are combined to form the core area of the building.
4. The UAV aerial photograph analysis method according to claim 2, characterized in that, Candidate regions for the formation of ancillary structures include: The edge of the building core area facing the initial road candidate area is extended towards the initial road candidate area by a second radius to form an expanded core area; the second radius is determined according to the road width of the target area. The overlapping area between the expansion core region and the initial road candidate region is used as the auxiliary structure candidate region.
5. The method for analyzing UAV aerial photographs according to claim 1, characterized in that, The formation of multiple candidate connected components includes: The pixels in the candidate region of the auxiliary structure are traversed and searched. When any pixel's neighboring pixels are also in the candidate region of the subordinate structure, these pixels are divided into the same candidate connected region.
6. The method for analyzing UAV aerial photographs according to claim 1, characterized in that, The acquisition of the selected building area includes: The images of the candidate connected components are converted to the LAB color space to form a connected component histogram, and the images of the corresponding candidate building core regions are converted to the LAB color space to form a core histogram. Calculate the Barcol distance between the distribution of the connected component histogram and the distribution of the core histogram, and determine the candidate connected component as the selected building region when the Barcol distance is lower than a preset value.
7. The method for analyzing UAV aerial photographs according to claim 1, characterized in that, Generating road areas and building areas based on the core building area and the selected building area includes: The core building area and the adjacent selected building areas are merged to form a building area; The area between adjacent building areas is defined as the road area.
8. A UAV aerial photograph analysis system, characterized in that, include: The identification unit is configured to acquire drone aerial photographs of the target area and acquire an initial building mask in the drone aerial photographs; The core unit is configured to use the area between adjacent initial building masks as an initial road candidate area, and extract the building core area based on the initial building masks; An expansion unit is configured to expand the core area of the building and use the intersection between the expanded area and the initial road candidate area as the auxiliary structure candidate area; The connectivity unit is configured to divide the candidate region of the subordinate structure into multiple candidate connectivity regions through connectivity analysis; The comparison unit is configured to compare each candidate connected component with the texture of the corresponding candidate building core region, and select the candidate connected components whose texture similarity exceeds a preset value as the selected building region; The candidate building core area is the building core area adjacent to the candidate connected domain; The generation unit is configured to generate a road area and a building area based on the building core area and the selected building area.
9. The UAV aerial photograph analysis system according to claim 8, characterized in that, The core unit is also configured as follows: The initial building mask is eroded and expanded with a first radius to form the core area of the building; the first radius is determined based on the thickness of the building walls.
10. The UAV aerial photograph analysis system according to claim 9, characterized in that, The core unit is also configured as follows: Construct a first circular structural element with a radius of a first radius, and slide the center of the first circular structural element pixel by pixel in the initial building mask; When sliding, if there are no pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a candidate pixel; if there are pixels in the first circular structural element that are not part of the initial building mask, the pixel corresponding to the center of the first circular structural element is marked as a non-candidate pixel. The center of the first circular structural element is slid across the non-candidate pixels and the candidate pixels. When there are pixels of the initial building mask in the first circular structural element, the pixel corresponding to the center of the first circular structural element is marked as the core pixel. All core pixels are combined to form the core area of the building.