Single building appearance disease multi-mode edge detection method and system

Abstract

The invention discloses a multi-modal edge detection method based on common appearance diseases of a single building and an application system thereof, and the method comprises the steps: collecting building image data, and reconstructing a building white model; building a building white model UV grid; carrying out disease multi-modal initial detection on disease types in the building image data, and identifying and evaluating a disease danger level corresponding to a disease multi-modal initial detection result according to a preset disease evaluation rule; mapping the disease multi-modal initial detection result to a building white model UV grid in real time, and dynamically rendering a visual initial detection result mark; performing man-machine interaction review verification, and obtaining secondary shot image data based on a review verification result; extracting key features of the disease by using secondary shot image data, and accurately associating and marking the key features to a building white model based on a spatial position; according to the multi-mode edge detection method based on the common diseases of the single building appearance and the application system thereof provided by the invention, automatic identification and digital management of the building appearance diseases are realized by utilizing cooperative operation of the unmanned aerial vehicle.

Description

Technical Field

[0001] This invention belongs to the field of building intelligent inspection technology, specifically relating to a multimodal edge detection method and its application system based on common defects in the appearance of single buildings. Background Technology

[0002] For the detection and full lifecycle management of defects in various scenarios such as residential buildings, bridges, and historical buildings, traditional detection technologies rely on total station manual measurement or monocular vision reconstruction, achieving defect location accuracy only at the decimeter level (error > 30cm), which is insufficient to meet millimeter-level repair requirements. The detection process depends on manually annotating CAD drawings, resulting in low efficiency (inspection of a single building takes > 4 hours). While it can detect single cracks, it lacks a unified detection framework for all defect types (cracks, hollow areas, flaking, corrosion), and data storage is unstructured, leading to poor traceability. Traditional detection methods rely on manually annotating CAD drawings, resulting in vague spatial location of defects (error > 50cm), lack of real-time linkage with digital twin models, and low efficiency in expert decision-making (diagnosis of a single building takes > 1 hour). Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a multimodal edge detection method and its application system based on common defects in the appearance of single buildings.

[0004] To achieve the above objectives, the present invention is implemented using the following technical solution: In a first aspect, the present invention provides a multimodal edge detection method based on common defects in the exterior of single-building structures, comprising the following steps: Using drones, the building image data is collected layer by layer from the ground to the top at preset intervals. Based on the SFM algorithm and MVS technology, the building image point cloud is reconstructed into a white model of the building. UV parameter lines are defined on the surface of the architectural white model, and the UV mesh of the architectural white model is constructed using the Poisson disk sampling algorithm; The YOLOv8 algorithm is used to perform a multimodal preliminary detection of defects in building image data. The defect risk level corresponding to the multimodal preliminary detection results is identified and evaluated according to the preset defect assessment rules. The multimodal preliminary inspection results of the defects are mapped to the UV mesh of the building white model in real time, and the preliminary inspection results are dynamically rendered and marked according to different defect risk levels. The WebGL platform is used to retrieve the diseased areas marked by the initial inspection results and conduct human-computer interaction verification. Based on the verification results, the spiral approximation path of the UV grid corresponding to the fine detection diseased area is determined and the secondary image data is obtained perpendicular to the surface of the diseased area. Key features of the disease were extracted using secondary image data and accurately associated and labeled to the building white model based on spatial location.

[0005] Furthermore, it also includes deploying base stations around the building and calibrating them with the building's foundation reference points to establish a globally unified WGS84 coordinate system.

[0006] Furthermore, the types of damage include cracks, peeling, and water damage.

[0007] Furthermore, the preset disease assessment rules include a comprehensive comparative assessment of four parameters for each disease type: size, color, cumulative damage, and historical trend. Dimensional parameters: Crack width ≤ 1.0 mm and no deformation is judged as general danger; crack width 0.5-1.0 mm (inclusive) and no deformation is judged as medium danger; crack width > 1.0 mm or accompanied by deformation is judged as serious danger.

[0008] Color parameters: Based on HSV color space analysis, if the hue (H) shift of a building surface pixel is ≤15°, the saturation (S) change is ≤20%, and the lightness (V) change is ≤25%, or the continuous area is <0.5m², it is judged as a general hazard; if the hue (H) shift is >15°, the saturation (S) change is >20%, and the lightness (V) change is >25%, and the continuous area is 0.5-1m² (inclusive), it is judged as a moderate hazard; if the hue (H) shift is >15°, the saturation (S) change is >20%, and the lightness (V) change is >25%, and the continuous area is >1m², it is judged as a severe hazard.

[0009] Hazard stacking: When an abnormal color area (meeting the medium or severe level of color parameters) overlaps with an abnormal size parameter (meeting the medium or severe level of size parameters), the level is increased by 1 level (the upper limit of the level is severe danger), that is, general danger is adjusted to medium danger, medium danger is adjusted to severe danger, and severe danger remains unchanged.

[0010] Historical trend: In the last 3 tests, if the rate of disease spread is <0.2mm / month, it is judged as general risk; if the rate of spread is 0.2-0.5mm / month (inclusive), it is judged as medium risk; and if the rate of spread is ≥0.5mm / month, it is judged as serious risk.

[0011] Furthermore, the disease risk level adopts a three-level risk classification system, namely: Serious danger: Handle within 24 hours; visually indicated by a red marker. Moderate risk: Treat within 1 month; visually indicated by a yellow marker. General hazard: Temporarily postpone treatment; visually indicated by a green marker.

[0012] Furthermore, size parameters and historical trend parameters are prioritized as the core judgment criteria, and the higher of the two is taken as the initial level; color parameters are further modified, and if the color parameter level is higher than the core parameter level, the color parameter level shall prevail; the above results are adjusted according to the hazard superposition rules; the final level is locked as one of the following: severe hazard (red mark), moderate hazard (yellow mark), and general hazard (green mark), and corresponding visual mark is assigned.

[0013] Furthermore, during the acquisition of secondary image data, a secondary shooting task is automatically generated based on the verification results. The spiral approach path of the UV grid corresponding to the area requiring fine detection of the disease starts from 5m away from the target and progresses in steps of 0.5m. Laser ranging is enabled to calibrate the flight altitude in real time, and the camera pitch angle is adjustable by ±45° to ensure that the shooting angle is perpendicular to the surface of the disease and to obtain high-precision images with a resolution of ≥0.1mm / pixel.

[0014] Furthermore, it also includes: archiving secondary image data according to "building ID-UV grid-detection type", and accurately associating spatial location through bidirectional binding of EXIF ​​metadata and UV coordinates; Edge computing is used to preprocess images such as distortion correction and color correction, and key features of defects are automatically annotated to the building white model; Employing an edge-cloud hybrid architecture, the initial inspection results and visualization labels are completed at the edge with a latency of <200ms, while the white model is updated and synchronized to the cloud, ensuring a total response time of <5 minutes.

[0015] In a second aspect, the present invention provides a multimodal edge detection system based on common defects in the exterior of single-building structures, capable of executing the multimodal edge detection method based on common defects in the exterior of single-building structures as described in any one of the first aspects, the system comprising: The data acquisition module uses a drone to collect building image data by gradually layering the building from the ground to the top at preset intervals. Based on the SFM algorithm and MVS technology, the building image point cloud is reconstructed into a white model of the building. Additionally, UV parameter lines are defined on the surface of the architectural white model, and a UV mesh of the architectural white model is constructed using a Poisson disk sampling algorithm; The edge terminal processing module uses the YOLOv8 algorithm to perform multimodal preliminary detection of defects in building image data, and identifies and evaluates the defect risk level corresponding to the multimodal preliminary detection results according to preset defect assessment rules. In addition, the multimodal preliminary inspection results of the defects are mapped to the UV mesh of the building white model in real time, and the preliminary inspection results are dynamically rendered and marked according to different defect risk levels. The WebGL platform decision module retrieves the diseased areas marked by the initial inspection results through the WebGL platform and performs human-computer interaction verification. Based on the verification results, it determines the spiral approximation path of the UV grid corresponding to the fine detection diseased area and acquires secondary image data perpendicular to the surface of the diseased area. Furthermore, key features of the disease are extracted using secondary image data, and accurately associated and labeled to the building white model based on spatial location.

[0016] Thirdly, the present invention provides an electronic device, comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the multimodal edge detection method based on common defects in the exterior of single-building structures, as described in any one of the first aspects.

[0017] Compared with existing technologies, the beneficial effects achieved by this invention are as follows: The multimodal edge detection method and its application system based on common defects in the appearance of single buildings provided by this invention utilize drone collaborative operations to achieve automated identification and digital management of building appearance defects. It integrates high-precision spatial positioning, intelligent pre-detection algorithms, and a human-machine collaborative decision-making platform. Through multi-source data fusion and UV mesh division, it establishes a precise spatial coordinate system with an error of ≤±15cm. From initial defect detection and detailed imaging to 4D model generation, the end-to-end processing provides full lifecycle data support for defect management through seamless integration with the UV parametric mesh and BIM modeling system. It is applicable to defect detection and full lifecycle management in multiple scenarios such as residential buildings, bridges, and historical buildings. Attached Figure Description

[0018] Figure 1 A flowchart of a multimodal edge detection method based on common defects in the appearance of single buildings provided in an embodiment of the present invention.

[0019] Figure 2 This is a block diagram of a multimodal edge detection system based on common defects in the exterior of single-building structures, provided in an embodiment of the present invention.

[0020] Figure 3 This is a flowchart illustrating the decomposition process of fusing UV parametric meshes and BIM models in the multimodal edge detection method provided in this embodiment of the invention.

[0021] Figure 4 This is a schematic diagram of the human-computer collaborative decision-making interface of the WebGL platform provided in an embodiment of the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0023] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are used only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0024] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances. Example

[0025] like Figure 1 As shown, this embodiment of the invention provides a multimodal edge detection method based on common defects in the exterior of single buildings, including the following steps: Using drones, the building image data is collected layer by layer from the ground to the top at preset intervals. Based on the SFM algorithm and MVS technology, the building image point cloud is reconstructed into a white model of the building. UV parameter lines are defined on the surface of the architectural white model, and the UV mesh of the architectural white model is constructed using the Poisson disk sampling algorithm; The YOLOv8 algorithm is used to perform a multimodal preliminary detection of defects in building image data. The defect risk level corresponding to the multimodal preliminary detection results is identified and evaluated according to the preset defect assessment rules. The multimodal preliminary inspection results of the defects are mapped to the UV mesh of the building white model in real time, and the preliminary inspection results are dynamically rendered and marked according to different defect risk levels. The WebGL platform is used to retrieve the diseased areas marked by the initial inspection results and conduct human-computer interaction verification. Based on the verification results, the spiral approximation path of the UV grid corresponding to the fine detection diseased area is determined and the secondary image data is obtained perpendicular to the surface of the diseased area. Key features of the disease were extracted using secondary image data and accurately associated and labeled to the building white model based on spatial location.

[0026] In some embodiments, 2-3 dedicated 5.8GHz frequency band base stations (model: HY-M-300-AP) are installed within a 50-meter radius around the building. Each base station has a coverage radius of 20 meters, is equipped with an NVIDIA Jetson Orin Nano edge computing unit (40TOPS computing power, 5-15W power consumption), and is configured with a 100Mbps wired backhaul link to achieve real-time transmission of 4K video (10fps, bitrate ≥100Mbps) and RTK positioning data (accuracy ±2cm), with an end-to-end latency ≤50ms.

[0027] The base station integrates a 17dBi 90-degree dual-polarized sector antenna (transmit power 1000mw), and establishes a globally unified WGS84 coordinate system through GPS + BeiDou dual-mode positioning and building foundation benchmark calibration.

[0028] The drones circled the building in layers at 1m intervals (from the ground to the top), capturing images at 10fps using a visible light camera. Based on the SFM algorithm and MVS technology, the image point cloud was reconstructed into a triangular mesh model. Noise was removed by voxel filtering (5cm voxel size) to generate a white building model with a precision of 5cm. UV parameter lines were divided on the surface of the white model at 1m intervals, and the mesh was evenly distributed using the Poisson disk sampling algorithm. Each UV mesh was assigned a unique spatial coordinate (X,Y,Z,UV) to construct the UV mesh of the white building model.

[0029] In some embodiments, the YOLOv8 algorithm is used to detect defects with distinctive characteristics, including cracks, peeling, and water immersion. The defect risk level is then calculated based on the following parameters.

[0030] The following criteria are used to determine the risk level: cracks with a width ≤ 1.0 mm and no deformation are classified as generally dangerous; cracks with a width of 0.5-1.0 mm (inclusive) and no deformation are classified as moderately dangerous; cracks with a width > 1.0 mm or with deformation are classified as seriously dangerous.

[0031] The following color parameters are used for assessment: Based on HSV color space analysis, if the hue (H) shift of a building surface pixel is ≤15°, the saturation (S) change is ≤20%, and the lightness (V) change is ≤25%, or the continuous area is <0.5m², it is judged as a general hazard; if the hue (H) shift is >15°, the saturation (S) change is >20%, and the lightness (V) change is >25%, and the continuous area is 0.5-1m² (inclusive), it is judged as a moderate hazard; if the hue (H) shift is >15°, the saturation (S) change is >20%, and the lightness (V) change is >25%, and the continuous area is >1m², it is judged as a severe hazard.

[0032] For overlapping hazards, the following assessment is made: When an abnormal color area (meeting the medium or severe level of color parameters) overlaps with an abnormal size parameter (meeting the medium or severe level of size parameters), the hazard level is increased by 1 level (the upper limit of the hazard level is severe danger), that is, general hazard is adjusted to medium danger, medium danger is adjusted to severe danger, and severe danger remains unchanged.

[0033] Judgment based on historical trends: In the last 3 tests, if the rate of disease spread is <0.2mm / month, it is judged as general risk; if the rate of spread is 0.2-0.5mm / month (inclusive), it is judged as medium risk; if the rate of spread is ≥0.5mm / month, it is judged as serious risk.

[0034] The disease risk level adopts a three-level risk classification system, namely: Red (Severe Danger): Address within 24 hours (e.g., through-cracks, risk of large-area hollowing and detachment). Yellow (Moderate Risk): Requires treatment within 1 month; Green (General Hazard): Temporarily postpone treatment.

[0035] After the corresponding disease risk level is determined, dynamic color coding and disease marking map the initial inspection results to the white model UV mesh in real time, and dynamically render the visual markings according to the risk level: Red marker: Bold border + 1Hz flashing effect to highlight the emergency handling area; Yellow marker: Semi-transparent fill with dashed border, indicating a stage of processing requirements; Green markers: solid line with thin border and light-colored fill, used for routine monitoring records. Each marker is associated with a details card; clicking on it displays the size of the defect, the location of the component, the type of component, and a trend chart of historical monitoring data.

[0036] In some embodiments, technicians use the WebGL platform to retrieve the diseased areas marked in the initial inspection, and conduct video tracking and viewing of key or questionable disease points to confirm the accuracy and completeness of the initial inspection results, thus forming a human-computer interaction verification process.

[0037] Based on the review results, the system automatically generates a second shooting task and plans a "spiral approximation" path for the UV grid that needs to be finely inspected, starting from 5m away from the target and progressing in 0.5m steps. The laser ranging module is enabled to calibrate the flight altitude in real time and control the camera pitch angle (adjustable ±45°) to ensure that the shooting angle is perpendicular to the diseased surface and to obtain high-precision images with a resolution of ≥0.1mm / pixel.

[0038] Secondary, precisely captured images are archived according to "Building ID-UV Grid-Detection Type," and precise spatial location is achieved through bidirectional binding of EXIF ​​metadata and UV coordinates. Image preprocessing (distortion removal, color correction) is performed using an edge computing terminal, and key features (crack width, peeling boundary, etc.) are automatically annotated to the white model; Adopting an edge-cloud hybrid architecture, initial inspection and labeling are completed at the edge (latency <200ms), while complex model calculations (such as white model updates) are synchronized to the cloud, ensuring that the entire process response time is <5 minutes, and realizing efficient data flow and real-time processing.

[0039] like Figure 2 As shown, this embodiment of the invention also provides a multimodal edge detection system based on common defects in the exterior of single-building structures, which can execute the above-described multimodal edge detection method based on common defects in the exterior of single-building structures. The system includes: The data acquisition module uses a drone to collect building image data by gradually layering the building from the ground to the top at preset intervals. Based on the SFM algorithm and MVS technology, the building image point cloud is reconstructed into a white model of the building. Additionally, UV parameter lines are defined on the surface of the architectural white model, and a UV mesh of the architectural white model is constructed using a Poisson disk sampling algorithm; The edge terminal processing module uses the YOLOv8 algorithm to perform multimodal preliminary detection of defects in building image data, and identifies and evaluates the defect risk level corresponding to the multimodal preliminary detection results according to preset defect assessment rules. In addition, the multimodal preliminary inspection results of the defects are mapped to the UV mesh of the building white model in real time, and the preliminary inspection results are dynamically rendered and marked according to different defect risk levels. The WebGL platform decision module retrieves the diseased areas marked by the initial inspection results through the WebGL platform and performs human-computer interaction verification. Based on the verification results, it determines the spiral approximation path of the UV grid corresponding to the fine detection diseased area and acquires secondary image data perpendicular to the surface of the diseased area. Furthermore, key features of the disease are extracted using secondary image data, and accurately associated and labeled to the building white model based on spatial location.

[0040] This invention also provides an electronic device, comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the multimodal edge detection method based on common defects in the appearance of single-building structures, as described above.

[0041] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A multimodal edge detection method based on common defects in the exterior of single buildings, characterized in that, Including the following steps: Using drones, the building image data is collected layer by layer from the ground to the top at preset intervals. Based on the SFM algorithm and MVS technology, the building image point cloud is reconstructed into a white model of the building. UV parameter lines are defined on the surface of the architectural white model, and the UV mesh of the architectural white model is constructed using the Poisson disk sampling algorithm; The YOLOv8 algorithm is used to perform a multimodal preliminary detection of defects in building image data. The defect risk level corresponding to the multimodal preliminary detection results is identified and evaluated according to the preset defect assessment rules. The multimodal preliminary inspection results of the defects are mapped to the UV mesh of the building white model in real time, and the preliminary inspection results are dynamically rendered and marked according to different defect risk levels. The WebGL platform is used to retrieve the diseased areas marked by the initial inspection results and conduct human-computer interaction verification. Based on the verification results, the spiral approximation path of the UV grid corresponding to the fine detection diseased area is determined and the secondary image data is obtained perpendicular to the surface of the diseased area. Key features of the disease were extracted using secondary image data and accurately associated and labeled to the building white model based on spatial location.

2. The multimodal edge detection method based on common defects in the exterior of single buildings according to claim 1, characterized in that, It also includes deploying base stations around the building and calibrating them with the building's foundation reference points to establish a globally unified WGS84 coordinate system.

3. The multimodal edge detection method based on common defects in the exterior of single-building structures according to claim 1, characterized in that, The types of damage include cracks, peeling, and water damage.

4. The multimodal edge detection method based on common defects in the exterior of single-building structures according to claim 1, characterized in that, The preset disease assessment rules include a comprehensive comparative assessment of four parameters for each disease type: size, color, cumulative damage, and historical trend. The following criteria are applied to assess the crack width: if the crack width is ≤1.0mm and there is no deformation, it is classified as a general hazard; if the crack width is 0.5-1.0mm (inclusive) and there is no deformation, it is classified as a moderate hazard; if the crack width is >1.0mm or there is deformation, it is classified as a severe hazard. The following color parameters are used for assessment: Based on HSV color space analysis, if the hue (H) shift of a building surface pixel is ≤15°, the saturation (S) change is ≤20%, and the lightness (V) change is ≤25%, or the continuous area is <0.5m², it is judged as a general hazard; if the hue (H) shift is >15°, the saturation (S) change is >20%, and the lightness (V) change is >25%, and the continuous area is 0.5-1m² (inclusive), it is judged as a moderate hazard; if the hue (H) shift is >15°, the saturation (S) change is >20%, and the lightness (V) change is >25%, and the continuous area is >1m², it is judged as a severe hazard. For the assessment of overlapping hazards: when an area with abnormal color and a color parameter that meets the medium or severe level overlaps with an area with abnormal size and a size parameter that meets the medium or severe level, the level is increased by 1 level and the upper limit of the level is severe hazard. That is, general hazard is adjusted to medium hazard, medium hazard is adjusted to severe hazard, and severe hazard remains unchanged. Judgment based on historical trends: In the last 3 tests, if the rate of disease spread is <0.2mm / month, it is judged as general risk; if the rate of spread is 0.2-0.5mm / month (inclusive), it is judged as medium risk; if the rate of spread is ≥0.5mm / month, it is judged as serious risk.

5. The multimodal edge detection method based on common defects in the exterior of single-building structures according to claim 4, characterized in that, The disease risk level adopts a three-level risk classification system, namely: Serious danger: Handle within 24 hours; visually indicated by a red marker. Moderate risk: Treat within 1 month; visually indicated by a yellow marker. General hazard: Temporarily postpone treatment; visually indicated by a green marker.

6. The multimodal edge detection method based on common defects in the exterior of single-building structures according to claim 5, characterized in that, The size parameter and historical trend parameter are used as the core judgment criteria first, and the higher of the two is taken as the initial level; the color parameter is used for further correction. If the color parameter level is higher than the core parameter level, the color parameter level shall prevail; the above results are adjusted according to the hazard superposition rule; the final hazard level is locked into one of the following: severe hazard (red mark), moderate hazard (yellow mark), and general hazard (green mark), and the corresponding visual mark is assigned.

7. The multimodal edge detection method based on common defects in the exterior of single-building structures according to claim 1, characterized in that, During the acquisition of secondary image data, a secondary shooting task is automatically generated based on the verification results. The spiral approach path of the UV grid corresponding to the area of ​​disease that needs to be finely inspected starts from 5m away from the target and progresses in steps of 0.5m. Laser ranging is enabled to calibrate the flight altitude in real time, and the camera pitch angle is adjustable by ±45° to ensure that the shooting angle is perpendicular to the surface of the disease and to obtain high-precision images with a resolution of ≥0.1mm / pixel.

8. The multimodal edge detection method based on common defects in the exterior of single-building structures according to claim 1, characterized in that, Also includes: Secondary image data is archived according to "Building ID-UV grid-Detection type" and accurately associated with spatial location through bidirectional binding of EXIF ​​metadata and UV coordinates; Edge computing is used to preprocess images such as distortion correction and color correction, and key features of defects are automatically annotated to the building white model; Employing an edge-cloud hybrid architecture, the initial inspection results and visualization labels are completed at the edge with a latency of <200ms, while the white model is updated and synchronized to the cloud, ensuring a total response time of <5 minutes.

9. A multimodal edge detection system based on common defects in the exterior of single-building structures, characterized in that, The system, capable of executing the multimodal edge detection method based on common defects in the exterior of single-building structures as described in any one of claims 1 to 8, comprises: The data acquisition module uses a drone to collect building image data by gradually layering the building from the ground to the top at preset intervals. Based on the SFM algorithm and MVS technology, the building image point cloud is reconstructed into a white model of the building. Additionally, UV parameter lines are defined on the surface of the architectural white model, and a UV mesh of the architectural white model is constructed using a Poisson disk sampling algorithm; The edge terminal processing module uses the YOLOv8 algorithm to perform multimodal preliminary detection of defects in building image data, and identifies and evaluates the defect risk level corresponding to the multimodal preliminary detection results according to preset defect assessment rules. In addition, the multimodal preliminary inspection results of the defects are mapped to the UV mesh of the building white model in real time, and the preliminary inspection results are dynamically rendered and marked according to different defect risk levels. The WebGL platform decision module retrieves the diseased areas marked by the initial inspection results through the WebGL platform and performs human-computer interaction verification. Based on the verification results, it determines the spiral approximation path of the UV grid corresponding to the fine detection diseased area and acquires secondary image data perpendicular to the surface of the diseased area. Furthermore, key features of the disease are extracted using secondary image data, and accurately associated and labeled to the building white model based on spatial location.

10. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the multimodal edge detection method based on common defects in the appearance of single-building structures as described in any one of claims 1 to 8.

Images