Intelligent detection method and system for building outer wall falling off

Abstract

The application provides a kind of building outer wall falling off intelligent detection method and system, the method comprises: collection and extraction building body outer facade feature, formulate the flight trajectory of unmanned aerial vehicle;Control unmanned aerial vehicle to fly according to flight trajectory, and carry out data acquisition, obtain outer facade infrared image data and outer facade visual image data;Based on infrared thermal imaging analysis model, the outer facade infrared image data is analyzed, and the infrared threat area is screened out;Based on visual analysis model, the outer facade visual image data and infrared threat area position data are analyzed and handled, and the visual dangerous area of interference area is excluded Secondary judgment is carried out, and dangerous data is obtained;Output dangerous data and send early warning signal.The application can carry out multi-angle, full coverage, close-range falling detection on building outer wall, so that staff can comprehensively analyze building outer wall and take preventive measures to avoid safety accidents.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology, specifically to an intelligent detection method and system for the detachment of building exterior walls. Background Technology

[0002] With the continuous expansion of university towns and the increase in student population, the problems existing in university buildings are becoming increasingly prominent. Generally, the service life of university teaching buildings is about 30-50 years, but most university dormitories and teaching buildings were built in the last century, and their service life has reached or even exceeded their expected lifespan. This has led to aging, cracking, hollowing, and tile detachment in many buildings. Accidents involving falling exterior wall tiles injuring people are common, posing a serious threat to the personal safety of teachers and students. Therefore, regular inspection of the exterior wall finishes of campus buildings is essential, and finding a quick and easy way to eliminate these safety hazards is key to preventing such incidents from recurring.

[0003] Currently, commonly used inspection methods include visual inspection and hammering, but these methods suffer from drawbacks such as reliance on scaffolding or cranes, high cost, long cycle time, and low efficiency. Infrared thermal imaging can analyze changes in the external temperature of the wall surface, but it is easily limited by the observation distance, elevation angle, and the influence of the material and color of the facade, resulting in high errors and uncertainties. Computer vision methods can quickly analyze and process large amounts of data, thus saving time and resources, but computer vision technology is still affected by lighting and viewing angle, causing significant errors. Faced with the complex field of building facade damage inspection, the above methods are clearly insufficient to meet the new demands of development trends.

[0004] Therefore, how to accurately and efficiently detect the peeling of building exterior walls has become an urgent problem for professionals in this field. Summary of the Invention

[0005] In view of this, it is necessary to provide an intelligent detection method and system for building exterior wall detachment, so as to achieve the purpose of accurate and efficient detection of the detachment of building exterior walls.

[0006] To achieve the above objectives, this invention provides an intelligent detection method for building exterior wall detachment, comprising:

[0007] Collect image data of building surface, extract building facade features, and formulate drone flight trajectory based on building facade features;

[0008] The drone is controlled to fly along the flight path and collect data to obtain infrared image data and visual image data of the facade.

[0009] Based on the infrared thermal imaging analysis model, the infrared image data of the facade is analyzed. If a threat exists, the infrared threat area is screened out and the location data of the infrared threat area is determined.

[0010] Based on the visual analysis model, the visual image data of the facade is analyzed and processed. If there is a danger, the visual danger area is screened out. The visual danger area after removing the interference area is judged again to obtain the danger data. The danger data includes the data of the detachment danger area and the location data of the detachment danger area.

[0011] Output dangerous data and issue warning signals.

[0012] In one possible implementation, image data of the building surface is acquired, building facade features are extracted, and the drone's trajectory is determined based on these features, including:

[0013] Collect image data of the building surface and extract the features of the building facade;

[0014] Based on the characteristics of the building's exterior facade, the building's exterior facade is divided into several rectangular areas;

[0015] Obstacle removal is performed on the rectangular region to obtain the target rectangular region;

[0016] The drone flight trajectory is established based on a grid method to traverse the target rectangular area.

[0017] In one possible implementation, facade image data is acquired based on infrared and visual camera components mounted on a drone.

[0018] In one possible implementation, the infrared thermal imaging analysis model is a deep learning model based on the support vector machine algorithm, and the kernel function of the support vector machine is the radial basis function kernel function.

[0019] In one possible implementation, visual analysis models are used to analyze and process exterior facade visual image data, and a secondary assessment is performed on threat areas after removing interfering regions to obtain hazard data, including:

[0020] The visual image data of the facade is filtered and processed to remove the visual image data of the facade with a resolution lower than a preset threshold, so as to obtain the first visual image data.

[0021] The first visual image data is filtered and enhanced to obtain the second visual image data.

[0022] Feature extraction is performed on the second visual image data to obtain the visual hazard area. Interference areas in the visual hazard area are removed to obtain the third visual image data.

[0023] A second assessment of visually hazardous areas in the third-vision image data is performed to obtain hazard data.

[0024] In one possible implementation, the first visual image data is filtered and enhanced to obtain the second visual image data, including:

[0025] The first visual image data is filtered based on adaptive smoothing filtering.

[0026] The first visual image data is contrast-enhanced using the histogram grayscale stretching method to obtain the second visual image data.

[0027] In one possible implementation, feature extraction is performed on the second visual image data to remove interfering regions from the threat region, resulting in third visual image data, including:

[0028] Feature extraction is performed on the second visual image data based on image edge detection, and interference areas in the visual danger area are removed based on curve feature extraction to obtain the third visual image data.

[0029] In one possible implementation, interference regions in visually hazardous areas are removed based on curve feature extraction, including:

[0030] The crack skeleton in the second visual image is obtained based on the image thinning algorithm. The burr area and fracture area in the crack skeleton are removed by the image closing operation to obtain the target crack skeleton.

[0031] Extract the pixel coordinates of the target crack skeleton, and calculate the linear feature fitting equation based on the pixel coordinates of the target crack skeleton;

[0032] By using curvature to fit the linear feature equation, the visual danger area is obtained. The straight line interference area in the target crack skeleton is removed to obtain the third visual image data.

[0033] In one possible implementation, a secondary assessment is performed on the visually hazardous areas in the third-vision image data to obtain hazardous data, including:

[0034] Based on the infrared threat region, a secondary judgment is made on the visual hazard region in the third visual image data to determine the difference between the infrared threat region and the visual hazard region, thereby obtaining hazard data.

[0035] To achieve the above objectives, the present invention also provides an intelligent detection system for building exterior wall detachment, comprising:

[0036] The flight trajectory planning module is used to collect image data of the building surface, extract the building facade features, and formulate the drone's flight trajectory based on the building facade features.

[0037] The data acquisition module is used to control the drone to fly along the flight path and to collect data, obtaining infrared image data and visual image data of the facade.

[0038] The infrared thermal imaging analysis module is used to analyze the infrared image data of the facade based on the infrared thermal imaging analysis model. If a threat exists, the infrared threat area is screened out and the location data of the infrared threat area is determined.

[0039] The visual analysis module is used to analyze and process the visual image data of the facade based on the visual analysis model. If there is a danger, the danger area is screened out. The visual danger area after removing the interference area is judged again to obtain the danger data. The danger data includes the data of the detachment danger area and the location data of the detachment danger area.

[0040] The output module is used to output dangerous data and issue warning signals.

[0041] The beneficial effects of the above embodiments are as follows: Image data of the building surface is collected, building facade features are extracted, and the drone's trajectory is determined based on these features; the drone is controlled to fly along the trajectory and collect data, obtaining infrared image data and visual image data of the facade; the infrared image data is analyzed based on an infrared thermal imaging analysis model to identify infrared threat areas and determine their location; the visual image data and infrared threat area location data are analyzed and processed based on a visual analysis model, and visually hazardous areas with eliminated interference areas are re-evaluated to obtain hazard data; the hazard data is output and a warning signal is issued. This invention enables multi-angle, full-coverage, and close-range detection of detachment from building facades, allowing staff to conduct comprehensive analysis and take preventative measures to avoid safety accidents. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 A flowchart illustrating an embodiment of an intelligent detection method for detachment of building exterior walls provided by the present invention;

[0044] Figure 2 This is a flowchart illustrating step S11 in one embodiment of the intelligent detection method for detachment of building exterior walls provided by the present invention.

[0045] Figure 3 This is a flowchart illustrating step S14 in one embodiment of the intelligent detection method for detachment of building exterior walls provided by the present invention.

[0046] Figure 4 This is a structural schematic diagram of an embodiment of an intelligent detection system for building exterior wall detachment provided by the present invention. Detailed Implementation

[0047] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0048] Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor systems and / or microcontroller systems.

[0049] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0050] Figure 1 This is a flowchart illustrating an embodiment of an intelligent detection method for building exterior wall detachment provided by the present invention.

[0051] Reference Figure 1 This invention provides an intelligent detection method for building exterior wall detachment, comprising:

[0052] S11. Collect image data of the building surface, extract the features of the building facade, and formulate the flight trajectory of the UAV based on the features of the building facade.

[0053] S12. Control the drone to fly along the flight path and collect data to obtain infrared image data and visual image data of the facade.

[0054] S13. Analyze the infrared image data of the facade based on the infrared thermal imaging analysis model. If there is a threat, filter out the infrared threat area and determine the location data of the infrared threat area.

[0055] S14. Based on the visual analysis model, analyze and process the visual image data of the facade. If there is a danger, screen out the dangerous areas, and make a second judgment on the visual dangerous areas after removing the interference areas to obtain the danger data. The danger data includes the data of the detachment dangerous areas and the location data of the detachment dangerous areas.

[0056] S15, Output dangerous data and issue a warning signal.

[0057] The beneficial effects of the above embodiments are as follows: Image data of the building surface is collected, building facade features are extracted, and the drone's trajectory is determined based on these features; the drone is controlled to fly along the trajectory and collect data, obtaining infrared image data and visual image data of the facade; the infrared image data is analyzed based on an infrared thermal imaging analysis model to identify infrared threat areas and determine their location; the visual image data and infrared threat area location data are analyzed and processed based on a visual analysis model, and visually hazardous areas with eliminated interference areas are re-evaluated to obtain hazard data; the hazard data is output and a warning signal is issued. This invention enables multi-angle, full-coverage, and close-range detection of detachment from building facades, allowing staff to conduct comprehensive analysis and take preventative measures to avoid safety accidents.

[0058] Understandably, this embodiment mainly includes two aspects: data acquisition and data processing. Data acquisition can be carried out by an UAV equipped with infrared and visual camera components. The data processing part refers to analyzing and judging the acquired infrared data through an infrared imaging analysis model. When defects are found, an infrared threat area is output, and the location of the infrared threat area is extracted. The acquired visual data is analyzed and judged through a visual analysis model. When the visual analysis model determines that a curve exists, a visual danger area is output. By comprehensively judging the infrared threat area and the visual danger area, the danger area is determined, and danger data and warning signals are output.

[0059] Figure 2 This is a flowchart illustrating step S11 in one embodiment of the intelligent detection method for building exterior wall detachment provided by the present invention.

[0060] Reference Figure 2 Step S11 includes:

[0061] S111. Collect image data of the building surface and extract the features of the building facade;

[0062] S112. Divide the building's exterior facade into several rectangular areas based on its external facade features;

[0063] S113. Perform obstacle removal processing on the rectangular area to obtain the target rectangular area;

[0064] S114. Establish the UAV flight trajectory that traverses the target rectangular area based on the grid method.

[0065] Understandably, the drone's flight path can be set using a component equipped with a GPS system. The setting rules include: for a vertical, flat facade, the flight path starts from a predetermined corner, moves vertically upwards to the next inflection point, then downwards, repeating this pattern until the entire facade is recorded and the drone moves to the next facade in a similar manner; for a horizontal, flat facade, the path should start from a predetermined corner and continue moving to the right, then move upwards to an inflection point and linearly to the left, repeating this pattern until the entire facade is recorded; after capturing the building surface image data, the drone can capture thermal imaging images of the roof using a similar grid pattern, starting from a corner and moving horizontally or vertically along the superimposed grid until the entire roof area is scanned.

[0066] Preferably, before data collection, a flight path needs to be planned, taking care to avoid obstacles or interference. The flight path should also be as continuous as possible to ensure the traceability of the data collection location. Simultaneously, the relationship between the drone and its surrounding environment should be checked in real time. When the wind speed is suitable, the drone can be unlocked and taken off after confirming that safe takeoff conditions have been met.

[0067] Furthermore, the facade image data was acquired based on infrared and visual camera components mounted on the drone.

[0068] In one embodiment, the infrared thermal imaging analysis model is a deep learning model based on the support vector machine algorithm, and the kernel function of the support vector machine is the radial basis function kernel function.

[0069] Understandably, infrared camera components utilize infrared detectors and optical imaging lenses to receive the infrared radiation energy distribution pattern of the measured area, and then project this pattern onto the photosensitive element of the infrared detector, thereby obtaining an infrared thermal image corresponding to the surface heat distribution field of the measured area. By converting the invisible infrared radiation from the exterior wall surface into a thermal image (different colors represent different temperature regions on the measured surface) through external temperature changes, and then using an infrared thermal imaging analysis model to analyze the overall temperature distribution of the thermal image, thereby determining whether a threat exists.

[0070] Specifically, the infrared thermal imaging analysis model includes a deep learning model based on the support vector machine (SVM) algorithm. In this embodiment, a small-sample surface defect identification dataset is established to train the automatic defect identification model. The radial basis function (RBF) kernel, which has strong learning capabilities and a wide adaptability range, is selected as the kernel function of the SVM. The SVM algorithm can construct an optimal segmentation plane, mapping the shape feature vectors of the selected defect regions to a high-dimensional space, thereby achieving automatic selection between defective and non-defective regions.

[0071] Figure 3 This is a flowchart illustrating step S14 in one embodiment of the intelligent detection method for building exterior wall detachment provided by the present invention.

[0072] Reference Figure 3 Step S14 includes:

[0073] S141. Filter the facade visual image data and remove facade visual image data with a clarity lower than a preset threshold to obtain the first visual image data.

[0074] S142. Filter and enhance the first visual image data to obtain the second visual image data;

[0075] S143. Extract features from the second visual image data to obtain the visual danger area, remove the interfering area in the visual danger area to obtain the third visual image data.

[0076] S144. Perform a secondary assessment of the visual hazard areas in the third visual image data to obtain hazard data.

[0077] The analysis and processing of facade visual image data and infrared threat area location data through visual analysis models can be divided into three steps: image screening, preprocessing, and target extraction and recognition, which will be introduced separately below.

[0078] In step S141, the exterior facade visual image data is filtered to select building exterior wall images that meet quality requirements, deleting images with severe blurriness or low clarity, thus obtaining the first visual image data. Step S142 uses adaptive smoothing filtering instead of the four-neighborhood average smoothing algorithm in the SFC combined method to extract features from the first visual image data, obtaining the second visual image data. This automatically adjusts the filtering parameters for different situations, making the algorithm more applicable. Step S143 extracts features from the second visual image data, specifically identifying areas where cracks may exist, thus obtaining visual hazard areas. Interference areas within these visual hazard areas are then removed. These interference areas can include wall dividing lines, some windows, railings, and other structures. Removing these structures effectively improves crack recognition accuracy. Step S144 performs a second judgment on the visual hazard areas in the third visual image data to obtain hazard data, further improving crack recognition accuracy.

[0079] In one embodiment, step S142 includes:

[0080] The first visual image data is filtered based on adaptive smoothing filtering.

[0081] The first visual image data is contrast-enhanced using the histogram grayscale stretching method to obtain the second visual image data.

[0082] In one application scenario, histogram grayscale stretching is used to enhance image contrast; the image is then subjected to adaptive filtering smoothing once and Laplacian sharpening once, repeated four times; an improved histogram stretching method is used to perform two contrast enhancement processes on the crack image; finally, adaptive filtering smoothing is applied twice to the crack image, and the improved histogram stretching method is used to enhance contrast five times. This method can effectively remove noise and improve image contrast while preserving image quality, highlighting crack features and thus facilitating further crack identification.

[0083] In one embodiment, feature extraction is performed on the second visual image data to remove interfering regions in the threat region, resulting in third visual image data, including:

[0084] Feature extraction is performed on the second visual image data based on image edge detection, and interference areas in the visual danger area are removed based on curve feature extraction to obtain the third visual image data.

[0085] Furthermore, the target extraction part in this embodiment employs an image edge detection method. This method extracts edge information from the high-frequency components of the image using differential operations, and then performs edge detection on the second visual image data based on a gradient operator template edge detection algorithm. The edge detection operator has high signal-to-noise ratio and detection accuracy. This embodiment uses an algorithm combining the Otsu's method and gradient histograms to calculate the high threshold, and sets the low threshold to half of the high threshold. This embodiment uses an improved algorithm to enhance the adaptive capability of the edge detection operator.

[0086] It needs to be explained that Otsu's method is an automatic thresholding method that maximizes the inter-class variance. Assume the image has N pixels, a grayscale range of [0, L-1], and n pixels corresponding to grayscale level i. i The probability of it occurring is:

[0087]

[0088] The pixels in the second visual image are divided into two categories, C0 and C1, based on their grayscale values ​​using a threshold T. C0 consists of pixels with grayscale values ​​in the range [0, T], and C1 consists of pixels with grayscale values ​​in the range [T+1, L-1]. The mean grayscale distribution probability of the second visual image is:

[0089]

[0090] u T Let C0 and C1 represent the mean probability distribution of the gray levels in the second visual image. Then, the expressions for the mean probability distributions of the gray levels in C0 and C1 are:

[0091]

[0092]

[0093] u0 represents the mean probability of the gray-level distribution of C0, and u1 represents the mean probability of the gray-level distribution of C1.

[0094]

[0095]

[0096]

[0097] Furthermore, the expression for the inter-class variance is:

[0098]

[0099] in, This represents the variance between classes.

[0100] In this embodiment, the optimal threshold for the Otsu method is obtained by taking values ​​of T sequentially within the range [0, L-1] to maximize the inter-class variance. After obtaining the threshold using the Otsu method, since the gradient histogram represents information such as edge strength, it is used to more accurately observe whether the currently obtained threshold will cause the loss of edge information in order to reduce the loss of some necessary crack edge points.

[0101] In one embodiment, images of building surface cracks acquired by a drone typically contain target areas suspected to be cracks, as well as non-crack interference areas. Therefore, the extracted edge information includes not only crack edges in the image but also edge information such as wall dividing lines and some windows, which complicates crack identification and measurement. Observation reveals that these interference areas are generally window frames, railing structural seams, etc., often exhibiting regular straight-line characteristics, while cracks are mostly irregular curves with weak directionality. Therefore, this invention utilizes the curvature of crack lines as a standard for eliminating interference areas.

[0102] In one embodiment, removing interfering regions from visually hazardous areas based on curve feature extraction includes:

[0103] The crack skeleton in the second visual image is obtained based on the image thinning algorithm. The burr area and fracture area in the crack skeleton are removed by the image closing operation to obtain the target crack skeleton.

[0104] Extract the pixel coordinates of the target crack skeleton, and calculate the linear feature fitting equation based on the pixel coordinates of the target crack skeleton;

[0105] By using curvature to fit the linear feature equation, the visual danger area is obtained. The straight line interference area in the target crack skeleton is removed to obtain the third visual image data.

[0106] Understandably, to obtain the linear features of the cracks and further acquire their curvature information, an image thinning algorithm is first used to obtain the crack skeleton based on the dual-vision image. Next, image closing operations are used to eliminate some small burrs or breaks. Specifically, line segment structural elements are selected as templates for morphological transformation based on the image features of the cracks. Morphological transformation and thinning of the building exterior wall cracks convert the binary image after edge detection into single-pixel lines. Based on this, the coordinates of all pixels constituting the line in the binary image need to be extracted, and mathematical equations are fitted using these coordinates to characterize the linear features of these lines. After fitting the polynomial equation of the line, the curvature at any point on the line is calculated. Based on the results, if the fitted line is a straight line, it can be deleted; if the fitted line is a curve, and there are no other special interfering elements, it can be determined that there are cracks in the building exterior wall crack image that need to be processed.

[0107] Furthermore, a secondary assessment is performed on the visually hazardous areas in the third-vision image data to obtain hazard data, including:

[0108] Based on the infrared threat region, a secondary judgment is made on the visual hazard region in the third visual image data to determine the difference between the infrared threat region and the visual hazard region, thereby obtaining hazard data.

[0109] Figure 4 This is a structural schematic diagram of an embodiment of an intelligent detection system for building exterior wall detachment provided by the present invention.

[0110] Reference Figure 4 The present invention also provides an intelligent detection system for building exterior wall detachment, comprising:

[0111] Flight trajectory planning module 41 is used to collect image data of the building surface, extract the building facade features, and formulate the flight trajectory of the UAV based on the building facade features.

[0112] The data acquisition module 42 is used to control the UAV to fly along the flight path and to collect data, obtaining infrared image data and visual image data of the facade.

[0113] The infrared thermal imaging analysis module 43 is used to analyze the infrared image data of the facade based on the infrared thermal imaging analysis model. If there is a threat, the infrared threat area is screened out and the location data of the infrared threat area is determined.

[0114] The visual analysis module 44 is used to analyze and process the visual image data of the facade based on the visual analysis model. If there is a danger, the danger area is screened out, and the visual danger area after removing the interference area is judged again to obtain the danger data. The danger data includes the data of the detachment danger area and the location data of the detachment danger area.

[0115] Output module 45 is used to output dangerous data and issue warning signals.

[0116] The beneficial effects of the above embodiments are as follows: the flight trajectory planning module 41 collects image data of the building surface, extracts the building facade features, and formulates the drone's flight trajectory based on the building facade features; the data acquisition module 42 controls the drone to fly according to the flight trajectory and collects data to obtain infrared image data and visual image data of the facade; the infrared thermal imaging analysis module 43 analyzes the infrared image data of the facade based on the infrared thermal imaging analysis model, filters out infrared threat areas, and determines the location data of the infrared threat areas; the visual analysis module 44 analyzes and processes the visual image data of the facade and the location data of the infrared threat areas based on the visual analysis model, performs a secondary judgment on the visually dangerous areas after eliminating interference areas, and obtains danger data; the output module 45 outputs the danger data and issues a warning signal. This invention can perform multi-angle, full-coverage, and close-range detection of detachment from building facades, enabling staff to conduct comprehensive analysis of building facades and take preventative measures to avoid safety accidents.

[0117] The intelligent detection system for building exterior wall detachment provided in the above embodiments can realize the technical solutions described in the embodiments of the intelligent detection method for building exterior wall detachment. The specific implementation principle of each module or unit can be based on the corresponding content in an embodiment of the intelligent detection method for building exterior wall detachment, which will not be repeated here.

[0118] The above provides a detailed description of the intelligent detection method and system for building exterior wall detachment provided by the present invention. Specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of ​​the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An intelligent detection method for building exterior wall detachment, characterized in that, include: Collect image data of the building surface, extract the building facade features, and formulate the drone's trajectory based on the building facade features; The drone is controlled to fly along the flight path and collect data to obtain infrared image data and visual image data of the facade. The infrared image data of the facade is analyzed based on the infrared thermal imaging analysis model. If a threat exists, the infrared threat area is screened out and the location data of the infrared threat area is determined. The visual image data of the facade is analyzed and processed based on a visual analysis model. If a hazard exists, a visual hazard area is selected. Based on the infrared threat area, a second judgment is made on the visual hazard area after removing interference areas. The visual hazard area is then used to obtain hazard data, which includes data on detachment hazard areas and their locations. Specifically, the visual image data of the facade is filtered to remove visual image data with a clarity lower than a preset threshold, resulting in first visual image data. The first visual image data is then filtered and enhanced to obtain second visual image data. A crack skeleton in the second visual image is obtained based on an image thinning algorithm. The burr and fracture areas in the crack skeleton are removed using image closing operations to obtain the target crack skeleton. The pixel coordinates of the target crack skeleton are extracted, and a linear feature fitting equation is calculated based on the pixel coordinates of the target crack skeleton. The curvature is used to judge the linear feature fitting equation to obtain the visual danger area. The straight line interference area in the target crack skeleton is removed to obtain the third visual image data. Based on the infrared threat region, a secondary judgment is made on the visual danger region in the third visual image data to determine the difference between the infrared threat region and the visual danger region, thereby obtaining danger data; Output the danger data and issue a warning signal.

2. The intelligent detection method for building exterior wall detachment according to claim 1, characterized in that, The process of collecting building surface image data, extracting building facade features, and determining the drone's trajectory based on these features includes: Collect image data of the building surface and extract the features of the building facade; Based on the aforementioned building facade features, the building facade is divided into several rectangular areas; Obstacle removal is performed on the rectangular region to obtain the target rectangular region; The UAV flight trajectory traversing the target rectangular area is established based on a grid method.

3. The intelligent detection method for building exterior wall detachment according to claim 1, characterized in that, The facade image data is acquired based on infrared and visual camera components mounted on a drone.

4. The intelligent detection method for building exterior wall detachment according to claim 1, characterized in that, The infrared thermal imaging analysis model is a deep learning model based on the support vector machine algorithm, and the kernel function of the support vector machine is the radial basis function. A second assessment is performed on the visually hazardous areas in the third visual image data to obtain hazardous data.

5. The intelligent detection method for building exterior wall detachment according to claim 1, characterized in that, The step of filtering and enhancing the first visual image data to obtain the second visual image data includes: The first visual image data is filtered based on adaptive smoothing filtering. The first visual image data is contrast-enhanced using the histogram grayscale stretching method to obtain the second visual image data.

6. An intelligent detection system for building exterior wall detachment, used to implement the intelligent detection method for building exterior wall detachment as described in any one of claims 1-5, characterized in that, The system includes: The flight trajectory planning module is used to collect image data of the building surface, extract the building facade features, and formulate the UAV's flight trajectory based on the building facade features. The data acquisition module is used to control the UAV to fly along the flight trajectory and to collect data, obtaining infrared image data and visual image data of the exterior facade. The infrared thermal imaging analysis module is used to analyze the infrared image data of the facade based on the infrared thermal imaging analysis model. If a threat exists, the infrared threat area is screened out and the location data of the infrared threat area is determined. The visual analysis module analyzes and processes the visual image data of the facade based on the visual analysis model. If there is a danger, the visual danger area is screened out. Based on the infrared threat area, the visual danger area that has eliminated interference areas is judged again. The visual danger area obtains danger data, which includes data on the detachment danger area and the location data of the detachment danger area. The output module is used to output the danger data and issue a warning signal.

Images