A wall defect detection system for high-rise building wall evaluation

Abstract

This invention discloses a wall defect detection system for evaluating the walls of high-rise buildings. This invention relates to the field of building inspection technology and solves the technical problems of inaccurate flight path planning leading to blind spots in image acquisition and low recognition efficiency and accuracy in intelligent defect identification. This invention fully utilizes BIM full-dimensional data to accurately extract the building facade outline, component distribution, and obstacle information. The flight path type is precisely matched with the building form and inspection requirements, solving the problem of missed detections in complex areas. Flight parameters are calculated using precise formulas based on sensor performance and detection accuracy requirements. Combined with side overlap rate, it ensures the clarity and stitching integrity of image acquisition, improving acquisition efficiency. A unified defect quantitative evaluation system is established. By combining RTK positioning data and the BIM model, the defect pixel coordinates are accurately converted into world coordinates. Clear quantitative grading standards are formulated for defects such as cracks, hollow areas, and leaks, improving recognition accuracy.

Description

Technical Field

[0001] This invention relates to the field of building inspection technology, specifically to a wall defect detection system for evaluating the walls of high-rise buildings. Background Technology

[0002] With the advancement of urbanization, the number of high-rise buildings and old buildings continues to grow. The facades of these buildings are exposed to the natural environment for a long time, which can easily lead to defects such as cracks, hollows, leaks, and peeling. These defects not only affect the aesthetics and functionality of the buildings, but may also cause structural safety hazards and threaten the safety of people's lives and property.

[0003] According to patent application number CN202311663778.6, a method and system for detecting defects in building walls are disclosed. The method includes: acquiring structural parameters of the wall to be tested, including crack parameters; establishing a three-dimensional model of the wall to be tested based on the structural parameters; marking suspected leakage points on one side of the wall of the three-dimensional model according to the crack parameters; marking leakage areas with the suspected leakage points as centers and the crack propagation length as a radius; applying a water leakage environment to the leakage areas and recording the time it takes for the water to reach the other side of the wall of the three-dimensional model; and outputting the leakage level based on the leakage time.

[0004] However, when existing wall defect detection systems are in use, the flight path planning does not make full use of BIM full-dimensional data, the flight path type is not well adapted to the building form and detection requirements, and complex areas are prone to missed detection; the flight parameter calculation accuracy is insufficient, which makes it difficult to balance the clarity and coverage integrity of the acquired images. Secondly, the rule base matching has weak anti-interference ability and poor adaptability to complex scenarios. The deep learning model relies on a large number of labeled samples, and the identification effect of small samples and rare defects is not good. It fails to achieve the complementary advantages of the two, and the false positive rate and false negative rate are too high. The defect assessment lacks a unified quantitative standard and grading system. It does not effectively integrate RTK positioning data and BIM model. The accuracy of world coordinate transformation of defect location is insufficient. The severity judgment is highly subjective. The assessment results are difficult to directly support engineering repair decisions. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a wall defect detection system for evaluating the walls of high-rise buildings. This system solves the problems of blind spots in image acquisition due to inaccurate flight path planning and low recognition efficiency and accuracy in intelligent defect identification.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a wall defect detection system for evaluating the walls of high-rise buildings, comprising: The building information acquisition module is used to acquire the full-dimensional BIM model of the building, assess the requirements and UAV parameter information, and transmit them to the flight route planning module; The flight path planning module is used to plan the flight path of the UAV based on the information transmitted by the building information acquisition module, extract the facade outline, mark the detection area and obstacles, select the corresponding flight path type based on the evaluation requirements, obtain the initial flight path, calculate flight parameters including flight altitude, flight path spacing and flight speed, input them into the initial flight path for optimization, generate a standard flight path, and transmit it to the image information acquisition module. The image information acquisition module is used to control the UAV to fly according to the standard flight path to acquire images of the building facade, and to perform image correction, noise suppression and feature enhancement processing on the facade images to obtain preprocessed images. At the same time, it extracts the corresponding edge features, texture features, shape features and grayscale statistical features from the preprocessed images, and transmits the extracted features to the defect recognition and processing module. The defect identification and processing module is used to match the acquired features with the defect rule base, calculate the rule satisfaction degree and weighted voting score of the corresponding defect type, comprehensively identify defects, generate defect identification results, and transmit them to the defect evaluation and analysis module. The defect assessment and analysis module is used to perform parameter quantification analysis on the acquired defect identification results. It also combines the defect type and quantification parameters with preset grading standards to determine the severity of the defect and generate defect assessment results.

[0007] As a further aspect of the present invention, the initial flight path is obtained in the flight path planning module as follows: If the assessment requirement is a rapid screening of rectangular building facades, select a vertical grid flight path and fly parallel to the building facade to form a vertical grid to achieve full facade coverage; If the assessment requirement is the inspection of a circular building, a building with a cantilever structure, or the inspection of a building's corner area, select the circular flight path and fly around the building in a circular pattern. If the assessment requirement is to inspect a super high-rise building with more than 30 floors, a spiral flight path should be selected, flying in a spiral upward motion from the bottom to the top. If the assessment requirement is to inspect key areas marked after the initial drone scan, select a fixed-point detailed inspection route and take close-up, multi-angle photos of suspected defective areas.

[0008] As a further aspect of the present invention, the standard flight route is generated in the flight route planning module in the following way: Calculate the flight altitude using the formula: Altitude = Camera Focal Length × Pixel Size / Distance to Target Ground. Then, use the formula: Spacing = 2 × Altitude × × (1 - side overlap rate) to calculate the flight path spacing. Different flight speeds are set based on different shooting conditions. The calculated parameters are input into the initial flight path for optimization. For complex parts such as building corners, cantilevered structures, and arcs, the flight path nodes are manually adjusted to obtain the standard flight path. The field of view is obtained by using the formula: vertical coverage H = target ground sampling distance × P. v , where P v The vertical pixel count of the camera is then used, along with the calculated coverage area, and the formula VFOV = 2 × The field of view (VFOV) is calculated, where D is the flight distance.

[0009] As a further aspect of the present invention, the image information acquisition module obtains the preprocessed image in the following manner: Image correction: Radial / tangential distortion is corrected through camera calibration, and image coordinate alignment is performed based on RTK positioning data; Noise suppression: Adaptive median filtering is used to process salt-and-pepper noise and discrete noise caused by drone jitter. The window size of the adaptive median filter is automatically adjusted according to the noise density of the pixel neighborhood. Bilateral filtering is used to process wall texture noise and smoothing noise caused by uneven lighting. Feature enhancement: The Laplacian operator and grayscale morphological operations are used to enhance the features of the facade image to obtain a preprocessed image.

[0010] As a further aspect of the present invention, the extraction method of edge features, texture features, shape features, and grayscale statistical features in the image information acquisition module is as follows: Edge features were extracted using the improved Canny operator, texture features were extracted using LBP features, and shape features were obtained based on edge features using the findContours function to obtain defect contours, eliminating noisy contours with an area of ​​less than 5 pixels. Gray-level statistical features were analyzed using local gray-level mean / variance. Z-score normalization is applied to the extracted image features to eliminate the dimensional differences between different feature dimensions. Mutual information between features is calculated, and highly correlated features with mutual information > 0.8 are removed.

[0011] As a further aspect of the present invention, the defect rule base includes: Crack rules: continuous edge length ≥ 20 pixels, aspect ratio ≥ 10, roundness ≤ 0.1, grayscale mean of candidate region ≤ background × 0.7, exclude regions with width > 5 pixels or breakpoint rate > 30%; Hollow area rules: Infrared temperature difference ≥3℃, connected area ≥1000 pixels, circularity ≥0.4, excluding areas with temperature gradient >1℃ / pixel; Leakage rules: Grayscale difference ≥30, GLCM contrast ≤0.6 of background, diffused shape (aspect ratio ≤5), excluding areas <500 pixels or less than 5cm away from doors and windows; The rules for detachment are: LBP entropy ≥ background × 1.5, edge gradient ≥ background × 2.0, area ≥ 250 pixels, and areas with an overlap rate with building structure > 80% are excluded.

[0012] As a further aspect of the present invention, the defect identification processing module generates defect identification results by comprehensively identifying defects in the following manner: Input image features and calculate the rule satisfaction and weighted voting score for the corresponding defect type, where rule satisfaction = (number of satisfied positive rules / total number of rules for this defect) × 100%, and weighted voting score = ... At the same time, the number of counterexample rules must be 0. If the rule satisfaction is ≥80% and the weighted voting score is ≥0.7, it is judged as a complete match, and the corresponding defect type, matching rule details and initial confidence are directly output, and the confidence score is equal to the weighted voting score. If the rule satisfaction rate is less than 60%, the weighted voting score is less than 0.5, or the number of counterexample rule satisfactions is greater than or equal to 2, then it is judged as a mismatch. In this case, the image features are input into a pre-trained deep learning model, which further analyzes and judges whether there is a defect and the specific type of defect, and generates a defect identification result.

[0013] As a further aspect of the present invention, the defect assessment and analysis module generates defect assessment results in the following manner: For crack defects, those with a width <0.2mm and a length <1m are classified as minor; those with a width ≤0.5mm or a length ≤3m are classified as moderate; and those with a width ≥0.5mm or a length ≥3m are classified as severe. For hollow areas with an area <0.1m² 2 A temperature difference of less than 4℃ is considered a minor issue; 0.1m 2 Area ≤ 0.5m² 2 Or, a temperature difference of 4℃ to 6℃ is considered a moderate level; the area must be ≥0.5m². 2 A temperature difference ≥6℃ or a large-scale distribution indicates a severe condition; For leakage defects with an area <0.05m² 2 And the lack of diffusion indicates a slight level; 0.05m 2 Area ≤ 0.2m² 2 Or, the localized spread is classified as moderate; the area is ≥0.2m². 2 Or it may spread vertically along the wall, which is classified as a severe level.

[0014] As a further embodiment of the present invention, the Laplacian operator is a 5×5 improved Laplacian operator, and the grayscale morphological operation is a combination of opening and closing operations, wherein the opening operation uses a 3×3 rectangular kernel and the closing operation uses a 3×3 circular kernel.

[0015] As a further aspect of the present invention, the improved Canny operator employs a dynamic dual threshold, with the high threshold being equal to the local grayscale mean × 1.6 and the low threshold being equal to the local grayscale mean × 0.4. The local window size is 11 × 11, and post-processing removes edge fragments with a length < 5 pixels.

[0016] This invention provides a wall defect detection system for evaluating the walls of high-rise buildings. Compared with the prior art, it has the following advantages: This invention fully utilizes BIM full-dimensional data to accurately extract the building facade outline, component distribution, and obstacle information. The flight path type is precisely matched with the building form and inspection requirements to solve the problem of missed detection in complex areas. The flight parameters are calculated using precise formulas based on sensor performance and inspection accuracy requirements. Combined with the side overlap rate, the clarity of image acquisition and the integrity of the stitching are ensured, thereby improving acquisition efficiency. Secondly, a unified defect quantitative assessment system is established. By combining RTK positioning data and BIM models, defect pixel coordinates are accurately converted into world coordinates. Clear quantitative grading standards are formulated for defects such as cracks, hollow areas, and leaks. The assessment results include defect type, quantitative indicators, severity level, grading basis, and BIM visualization annotations, which directly support the formulation of repair plans and project acceptance, and solve the problems of insufficient qualitative and authoritative assessments in the current system. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the wall defect detection system of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] First Embodiment Please see Figure 1 This application provides a wall defect detection system for high-rise building wall assessment, including: a building information acquisition module, a flight path planning module, an image information acquisition module, a defect identification and processing module, and a defect assessment and analysis module, and according to... Figure 1 It can be seen that the information between the above functional modules is transmitted in one direction only.

[0020] The building information acquisition module is used to acquire the BIM model of the building. It can acquire the full-dimensional BIM model of the building through three methods: local import, cloud database retrieval, or on-site lightweight scanning and reconstruction, and obtain the corresponding evaluation requirements. The evaluation requirements include the inspection object, accuracy requirements, and area priority. At the same time, it acquires the UAV parameter information, specifically the maximum flight altitude, maximum flight speed, and sensor parameters of the UAV. Then, the acquired information is transmitted to the flight path planning module.

[0021] The flight path planning module is used to plan the drone's flight path based on the acquired information. Based on the acquired BIM model, it extracts the building's facade outline and marks the detection area and obstacles on the model. The obstacles include no-fly zones. Based on the assessment requirements, the corresponding flight path type is selected to obtain the initial flight path. Specifically, for the assessment requirement of rapid screening of rectangular building facades, a vertical grid flight path is selected, flying parallel to the building facade to form a vertical grid and ensure full coverage of the facade. For circular buildings, buildings with cantilever structures, or buildings that need to be inspected at corner areas, a circular flight path is selected, flying in a ring around the building, which is suitable for curved / irregular buildings. For super high-rise buildings with more than 30 floors, a spiral flight path is selected, flying spirally upward from the bottom to the top. For key areas marked after the initial drone scan, a fixed-point detailed inspection flight path is selected to take close-up, multi-angle pictures of suspected defect areas. Next, different flight parameters are calculated, including flight altitude, flight path spacing, and flight speed. Flight altitude is determined by sensor resolution, using the formula: Altitude = Camera focal length × Pixel size / Target ground distance. For example, to detect a 0.1mm crack, with a camera pixel size of 3.45µm and a focal length of 24mm, the altitude would be 24 × 3.45 / 0.1 ≈ 8.28m. Flight path spacing is calculated based on the camera's field of view and side overlap rate, using the formula: Spacing = 2 × Altitude × × (1 - lateral overlap rate) is used to calculate the flight path spacing, with the lateral overlap rate set to the minimum of 80%. The field of view is calculated as follows: The physical coverage area corresponding to the sensor pixels is calculated based on the target ground sampling distance and the flight distance. The vertical coverage area H is calculated using the formula: Vertical Coverage H = Target Ground Sampling Distance × P v , where P v The vertical pixel count of the camera is then used, along with the calculated coverage area, and the formula VFOV = 2 × The field of view (VFOV) is calculated, where D is the flight distance; At the same time, different flight speeds are set based on different shooting conditions: ≤5m / s for high-definition camera shooting and ≤3m / s for infrared detection. The calculated parameters are input into the initial flight path for optimization. For complex parts such as building corners, cantilevered structures, and arcs, the flight path nodes are manually adjusted to obtain a standard flight path, which is then transmitted to the image information acquisition module.

[0022] Second Embodiment As a second embodiment of the present invention, it is implemented based on the first embodiment, and the difference from the first embodiment is as follows: The image information acquisition module is used to fly the UAV according to the acquired standard flight path. The UAV follows the standard flight path and triggers image acquisition based on RTK positioning data and IMU attitude sensor to obtain the exterior facade image of the building. At the same time, the obtained exterior facade image is preprocessed, and the specific preprocessing operations include the following: First, image correction is performed on the facade image. Radial / tangential distortion is corrected through camera calibration, and image coordinate alignment is performed based on RTK positioning data. Then, adaptive median filtering and bilateral filtering are used. Specifically, for salt-and-pepper noise and discrete noise caused by drone jitter, adaptive median filtering with dynamic windows is used, and the window size is automatically adjusted according to the noise density of the pixel neighborhood. For wall texture noise and smoothing noise caused by uneven lighting, bilateral filtering is used to suppress noise while maintaining the sharpness of defect edges. Noise suppression is performed on the facade image. Finally, Laplacian operator and grayscale morphological operations are used to enhance the features of the facade image. The Laplacian operator is a 5×5 modified Laplacian operator, and the grayscale morphological operations are a combination of opening and closing operations. The opening operation uses a 3×3 rectangular kernel, and the closing operation uses a 3×3 circular kernel to obtain the preprocessed image. Image features are extracted from the preprocessed image, including edge features, texture features, shape features, and grayscale statistical features. For edge features, an improved Canny operator is used, employing dynamic dual thresholds: a high threshold equal to 1.6 times the local grayscale mean and a low threshold equal to 0.4 times the local grayscale mean, with a local window size of 11×11. Post-processing removes edge fragments shorter than 5 pixels. For texture features, LBP feature extraction is used. For shape features, the findContours function is used to obtain defect contours based on edge features, removing noisy contours with an area less than 5 pixels. For grayscale statistical features, local grayscale mean / variance is analyzed, and the resulting image features are standardized. Z-score standardization is applied to all extracted features to eliminate dimensional differences between different feature dimensions. Mutual information between features is calculated, and highly correlated features with mutual information greater than 0.8 are removed to reduce feature dimensionality. The resulting image features are then transmitted to the defect recognition processing module.

[0023] The defect identification and processing module is used to identify defects based on the acquired image features. First, a defect rule base is constructed, including rules for cracks, hollow areas, leaks, and detachments. Specifically, the crack rule indicates a continuous edge length ≥ 20 pixels, aspect ratio ≥ 10, circularity ≤ 0.1, and candidate region grayscale mean ≤ background × 0.7, excluding areas with a width > 5 pixels or a breakpoint rate > 30%. The hollow area rule indicates an infrared temperature difference ≥ 3℃, connected area ≥ 1000 pixels, and circularity ≥ 0.4, excluding areas with a temperature gradient > 1℃ / pixel. The leak rule indicates a grayscale difference ≥ 30, GLCM contrast ≤ background × 0.6, and a diffused appearance (aspect ratio ≤ 5), excluding areas with an area < 500 pixels or a distance < 5cm from doors and windows. The detachment rule indicates an LBP entropy ≥ background × 1.5, edge gradient ≥ background × 2.0, and area ≥ 250 pixels, excluding areas with an overlap rate > 80% with the building structure. The obtained image features are then matched against the defect rule base, with the specific matching method as follows: Input image features and calculate the rule satisfaction and weighted voting score for the corresponding defect type, where rule satisfaction = (number of satisfied positive rules / total number of rules for this defect) × 100%, and weighted voting score = ... Simultaneously, the number of counterexample rules must be 0. If the rule satisfaction rate is ≥80% and the weighted voting score is ≥0.7, it is determined to be a complete match, and the corresponding defect type, matching rule details, and initial confidence score are directly output, with the confidence score equal to the weighted voting score. There is no need to enter the deep learning model. If the rule satisfaction rate is less than 60%, the weighted voting score is less than 0.5, or the number of counterexample rule satisfactions is greater than or equal to 2, it is judged as a mismatch. The image features are then input into a pre-trained deep learning model. The deep learning model further analyzes and judges whether there is a defect and the specific type of defect, generates a defect identification result, and transmits it to the defect assessment and analysis module.

[0024] The defect assessment and analysis module is used to determine and assess the defect based on the acquired defect identification results. It obtains the defect region in the preprocessed image and segments it. At the same time, it performs morphological processing on the segmented defect region. Based on the defect coordinates and area, it removes duplicate annotations of the same defect in multiple frames of images. Combining RTK positioning data and BIM model coordinates, it converts the defect pixel coordinates into world coordinates and performs quantitative analysis based on the defect identification results. For defects identified as cracks, the corresponding crack width and crack length are obtained, and the severity is determined by combining the corresponding grading standards. If the width is <0.2mm and the crack length is <1m, it is classified as a minor defect. If 0.2mm≤crack width<0.5mm or 1m≤crack length<3m, it is classified as a moderate defect. If the crack width is ≥0.5mm or the crack length is ≥3m, it is classified as a severe defect. For defects identified as hollow areas, their corresponding area and temperature difference are obtained. If the area is <0.1m², the area is considered hollow. 2 If the temperature difference is less than 4℃, it is classified as a minor level; if it is 0.1m 2 Area ≤ 0.5m² 2 If the temperature difference is between 4℃ and 6℃, it is classified as a moderate level. If the area is ≥0.5m², it is classified as a medium level. 2 If the temperature difference is ≥6℃ or the distribution is widespread, it is classified as a severe level. For defects identified as leakage, their area and diffusion range are obtained. If the area is <0.05m², the area is considered a leakage defect. 2 If there is no diffusion, it is classified as a slight level; if 0.05m 2 Area ≤ 0.2m² 2 If it is localized or spreads, it is classified as a moderate level; if the area is ≥0.2m². 2 If it spreads vertically along the wall, it is classified as a severe level; At the same time, defect assessment results are generated and displayed to the relevant managers.

[0025] Third Embodiment As a third embodiment of the present invention, the focus is on combining the implementation processes of the first and second embodiments.

[0026] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0027] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A wall defect detection system for evaluating the walls of high-rise buildings, characterized in that, include: The building information acquisition module is used to acquire the full-dimensional BIM model of the building, assess the requirements and UAV parameter information, and transmit them to the flight route planning module; The flight path planning module is used to plan the flight path of the UAV based on the information transmitted by the building information acquisition module, extract the facade outline, mark the detection area and obstacles, select the corresponding flight path type based on the evaluation requirements, obtain the initial flight path, calculate flight parameters including flight altitude, flight path spacing and flight speed, input them into the initial flight path for optimization, generate a standard flight path, and transmit it to the image information acquisition module. The image information acquisition module is used to control the UAV to fly according to the standard flight path to acquire images of the building facade, and to perform image correction, noise suppression and feature enhancement processing on the facade images to obtain preprocessed images. At the same time, it extracts the corresponding edge features, texture features, shape features and grayscale statistical features from the preprocessed images, and transmits the extracted features to the defect recognition and processing module. The defect identification and processing module is used to match the acquired features with the defect rule base, calculate the rule satisfaction degree and weighted voting score of the corresponding defect type, comprehensively identify defects, generate defect identification results, and transmit them to the defect evaluation and analysis module. The defect assessment and analysis module is used to perform parameter quantification analysis on the acquired defect identification results. It also combines the defect type and quantification parameters with preset grading standards to determine the severity of the defect and generate defect assessment results.

2. The wall defect detection system for high-rise building wall assessment according to claim 1, characterized in that, The initial flight path is obtained in the flight path planning module as follows: If the assessment requirement is a rapid screening of rectangular building facades, select a vertical grid flight path and fly parallel to the building facade to form a vertical grid to achieve full facade coverage; If the assessment requirement is the inspection of a circular building, a building with a cantilever structure, or the inspection of a building's corner area, select the circular flight path and fly around the building in a circular pattern. If the assessment requirement is to inspect a super high-rise building with more than 30 floors, a spiral flight path should be selected, flying in a spiral upward motion from the bottom to the top. If the assessment requirement is to inspect key areas marked after the initial drone scan, select a fixed-point detailed inspection route and take close-up, multi-angle photos of suspected defective areas.

3. The wall defect detection system for high-rise building wall assessment according to claim 1, characterized in that, The standard flight route is generated in the flight route planning module as follows: Calculate the flight altitude using the formula: Altitude = Camera Focal Length × Pixel Size / Distance to Target Ground. Then, use the formula: Spacing = 2 × Altitude × × (1 - side overlap rate) to calculate the flight path spacing. Different flight speeds are set based on different shooting conditions. The calculated parameters are input into the initial flight path for optimization. For complex parts such as building corners, cantilevered structures, and arcs, the flight path nodes are manually adjusted to obtain the standard flight path. The field of view is obtained by using the formula: vertical coverage H = target ground sampling distance × P. v , where P v The vertical pixel count of the camera is then used, along with the calculated coverage area, and the formula VFOV = 2 × The field of view (VFOV) is calculated, where D is the flight distance.

4. A wall defect detection system for evaluating high-rise building walls according to claim 1, characterized in that, The image information acquisition module obtains the preprocessed image in the following way: Image correction: Radial / tangential distortion is corrected through camera calibration, and image coordinate alignment is performed based on RTK positioning data; Noise suppression: Adaptive median filtering is used to process salt-and-pepper noise and discrete noise caused by drone jitter. The window size of the adaptive median filter is automatically adjusted according to the noise density of the pixel neighborhood. Bilateral filtering is used to process wall texture noise and smoothing noise caused by uneven lighting; Feature enhancement: The Laplacian operator and grayscale morphological operations are used to enhance the features of the facade image to obtain a preprocessed image.

5. A wall defect detection system for evaluating the walls of high-rise buildings according to claim 1, characterized in that, The extraction methods for edge features, texture features, shape features, and grayscale statistical features in the image information acquisition module are as follows: Edge features were extracted using the improved Canny operator, texture features were extracted using LBP features, and shape features were obtained based on edge features using the findContours function to obtain defect contours, eliminating noisy contours with an area of ​​less than 5 pixels. Gray-level statistical features were analyzed using local gray-level mean / variance. Z-score normalization is applied to the extracted image features to eliminate the dimensional differences between different feature dimensions. Mutual information between features is calculated, and highly correlated features with mutual information > 0.8 are removed.

6. A wall defect detection system for evaluating the walls of high-rise buildings according to claim 1, characterized in that, The defect rule base includes: Crack rules: continuous edge length ≥ 20 pixels, aspect ratio ≥ 10, roundness ≤ 0.1, grayscale mean of candidate region ≤ background × 0.7, exclude regions with width > 5 pixels or breakpoint rate > 30%; Hollow area rules: Infrared temperature difference ≥3℃, connected area ≥1000 pixels, circularity ≥0.4, excluding areas with temperature gradient >1℃ / pixel; Leakage rules: Grayscale difference ≥30, GLCM contrast ≤0.6 of background, diffused shape (aspect ratio ≤5), excluding areas <500 pixels or less than 5cm away from doors and windows; The rules for detachment are: LBP entropy ≥ background × 1.5, edge gradient ≥ background × 2.0, area ≥ 250 pixels, and areas with an overlap rate with building structure > 80% are excluded.

7. A wall defect detection system for evaluating high-rise building walls according to claim 1, characterized in that, The defect identification and processing module generates defect identification results by comprehensively identifying defects in the following way: Input image features and calculate the rule satisfaction and weighted voting score for the corresponding defect type, where rule satisfaction = number of satisfied positive rules / total number of rules for that defect × 100%, and weighted voting score = At the same time, the number of counterexample rules must be 0. If the rule satisfaction is ≥80% and the weighted voting score is ≥0.7, it is judged as a complete match, and the corresponding defect type, matching rule details and initial confidence are directly output, and the confidence score is equal to the weighted voting score. If the rule satisfaction rate is less than 60%, the weighted voting score is less than 0.5, or the number of counterexample rule satisfactions is greater than or equal to 2, then it is judged as a mismatch. In this case, the image features are input into a pre-trained deep learning model, which further analyzes and judges whether there is a defect and the specific type of defect, and generates a defect identification result.

8. A wall defect detection system for evaluating the walls of high-rise buildings according to claim 1, characterized in that, The defect assessment and analysis module generates defect assessment results in the following way: For crack defects, those with a width <0.2mm and a length <1m are classified as minor; those with a width ≤0.5mm or a length ≤3m are classified as moderate; and those with a width ≥0.5mm or a length ≥3m are classified as severe. For hollow areas with an area <0.1m² 2 A temperature difference of less than 4℃ is considered a minor issue; 0.1m 2 Area ≤ 0.5m² 2 Or, a temperature difference of 4℃ ≤ temperature difference < 6℃ is considered a moderate level. Area ≥ 0.5m² 2 A temperature difference ≥6℃ or a large-scale distribution indicates a severe condition; For leakage defects with an area <0.05m² 2 And the lack of diffusion indicates a slight level; 0.05m 2 Area ≤ 0.2m² 2 Or, the localized spread is classified as moderate; the area is ≥0.2m². 2 Or it may spread vertically along the wall, which is classified as a severe level.

9. A wall defect detection system for evaluating the walls of high-rise buildings according to claim 4, characterized in that, The Laplacian operator is a 5×5 improved Laplacian operator, and the grayscale morphological operations are a combination of opening and closing operations, where the opening operation uses a 3×3 rectangular kernel and the closing operation uses a 3×3 circular kernel.

10. A wall defect detection system for evaluating the walls of high-rise buildings according to claim 5, characterized in that, The improved Canny operator uses a dynamic dual threshold: the high threshold is equal to the local grayscale mean × 1.6, and the low threshold is equal to the local grayscale mean × 0.

4. The local window size is 11×11, and post-processing removes edge fragments with a length of less than 5 pixels.

Images