Concrete building outer wall detection method and system, computer equipment and storage medium

Abstract

The invention discloses a concrete building outer wall detection method and system, computer equipment and a storage medium, and relates to the technical field of prefabricated building detection, and the method comprises the steps: obtaining multi-angle image data of a concrete building outer wall; preprocessing the image data, and extracting a target area of the outer wall of the concrete building; extracting feature information representing outer wall defects from the preprocessed image data; inputting the feature information into an artificial intelligence detection model, and judging whether an outer wall defect exists or not; and according to an output result of the detection model, determining distribution positions of the defects on the surface of the outer wall, and classifying the detected defects. According to the concrete building outer wall detection method provided by the invention, the unmanned aerial vehicle avoids low efficiency and safety risks of high-altitude operation in manual detection. Meanwhile, the convolutional neural network can automatically extract multi-level features from the image, various defects are accurately recognized, and the accuracy problem of traditional infrared detection under the change of environmental factors is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of prefabricated building detection, and in particular to a method, system, computer equipment and storage medium for detecting exterior walls of concrete buildings. Background Art

[0002] With the acceleration of urbanization, the number of concrete buildings is increasing. In the long-term use of these buildings, they are affected by the natural environment (such as climate change, wind erosion, rain erosion), physical effects (such as thermal expansion and contraction caused by temperature changes), chemical effects (such as air pollution and acid rain) and human factors (such as construction quality, unreasonable use and maintenance), and their exterior walls are prone to various defects.

[0003] Exterior wall defects not only affect the appearance of the building, but more importantly, may reduce the stability and safety of the structure. For example, cracks may cause rainwater to penetrate, which in turn corrodes the internal steel bars and weakens the structural bearing capacity; large-scale peeling may injure pedestrians or damage surrounding facilities. Therefore, people's requirements for building quality and safety are increasing, and relevant building regulations and standards are becoming more and more stringent. In order to protect the safety of public life and property, it is particularly important to detect defects in the exterior walls of concrete buildings in a timely and accurate manner and take effective repair measures.

[0004] Traditional methods for detecting defects in concrete building exterior walls include manual detection, infrared detection, and sound wave and radar detection methods:

[0005] Manual inspection is the most commonly used traditional inspection method, but it has obvious inefficiency problems. Manual inspection requires a lot of manpower input and has a slow inspection speed, especially in high-rise buildings and large-area exterior walls. In addition, manual high-altitude operations have great safety risks, and inspectors need to work in a high-altitude environment, which increases the danger of operators. At the same time, the results of manual inspections are highly dependent on the experience and subjective judgment of inspectors, which may lead to misjudgment or missed judgment, thus affecting the accuracy and reliability of the inspection.

[0006] Infrared testing uses temperature differences to identify defects in concrete exterior walls, such as voids and moisture penetration. While this method is effective in some cases, its applicability is limited. Infrared testing is less effective for defects such as small cracks, where the temperature difference is not obvious. In addition, infrared testing is easily interfered by external environmental factors, such as changes in external temperature and solar radiation, which can affect the accuracy and reliability of the test results. Therefore, infrared testing cannot fully cover all types of exterior wall defects.

[0007] Acoustic wave and radar detection methods have certain advantages in detecting internal defects of concrete, but the equipment and operation of these methods are complex and require professional technicians to operate and analyze data. This not only increases the technical threshold of detection, but also increases the cost of detection. In addition, acoustic wave and radar detection can usually only detect local areas of a building, and it is difficult to achieve comprehensive coverage of the entire exterior wall. Therefore, in practical applications, these methods can often only be used as auxiliary means, and cannot undertake the task of comprehensive detection alone, which limits their application in large-scale building detection.

[0008] In summary, traditional detection methods each have their own limitations and are unable to comprehensively, quickly and efficiently detect various defects in concrete building exterior walls. Summary of the invention

[0009] In view of the above-mentioned problems, the present invention is proposed.

[0010] Therefore, the technical problem solved by the present invention is that although the traditional infrared detection method can identify some defects, it is not effective in detecting small cracks and some defects that are difficult to produce obvious temperature differences, and is easily interfered by environmental factors, resulting in inaccurate detection results. In addition, although sound wave and radar detection can detect defects inside concrete, it is difficult to carry out large-scale and efficient detection due to the complexity of the equipment, high difficulty in operation, and only partial coverage. Existing detection methods have certain limitations in detection accuracy, scope of application, and efficiency.

[0011] To solve the above technical problems, the present invention provides the following technical solutions: a method for detecting exterior walls of concrete buildings, comprising: acquiring multi-angle image data of exterior walls of concrete buildings; preprocessing the image data to extract target areas of exterior walls of concrete buildings; extracting feature information representing defects of exterior walls from the preprocessed image data; inputting the feature information into an artificial intelligence detection model to determine whether there are defects in the exterior walls; determining the distribution positions of defects on the exterior wall surface according to the output results of the detection model, and classifying the detected defects.

[0012] As a preferred solution of the concrete building exterior wall detection method described in the present invention, the acquisition of multi-angle image data of the concrete building exterior wall includes controlling the drone to fly to different collection positions in sequence along the path to collect image data covering the concrete building exterior wall; during the image collection process, the drone's lens exposure parameters and flight tilt angle are adjusted according to real-time lighting conditions.

[0013] As a preferred solution of the concrete building exterior wall detection method of the present invention, the preprocessing includes denoising the image data using a noise suppression algorithm; enhancing the image data using a contrast enhancement method;

[0014] Crop the image to keep the relevant part of the concrete building exterior wall;

[0015] Convert the color space of image data.

[0016] As a preferred solution of the concrete building exterior wall detection method of the present invention, wherein: the preprocessing also includes performing image registration when detecting multiple images;

[0017] The image registration adjusts the image positions according to the geometric relationships among the multiple images;

[0018] After registration, the images form a unified reference coordinate system.

[0019] As a preferred solution of the concrete building exterior wall detection method of the present invention, wherein: the characteristic information characterizing the exterior wall defects includes crack characteristics, peeling characteristics and hole characteristics; the crack characteristics include shape characteristics, texture characteristics and gradient characteristics;

[0020] The peeling characteristics include regional characteristics and color characteristics;

[0021] The hole features include geometric features and grayscale features;

[0022] The crack characteristics include shape characteristics, texture characteristics and gradient characteristics;

[0023] The peeling characteristics include regional characteristics and color characteristics;

[0024] The hole features include geometric features and grayscale features.

[0025] As a preferred solution of the concrete building exterior wall detection method of the present invention, wherein: the shape feature extracts the geometric shape parameters of the crack by performing edge detection and morphological processing on the crack area;

[0026] The texture feature uses an image processing method to analyze the difference in texture between the crack area and the surrounding normal area, and extracts the texture information of the crack area;

[0027] The gradient feature detects the edge and gradient changes in the image through a gradient operator, and extracts the gradient information of the crack area;

[0028] The regional features are obtained by calculating geometric parameters of the peeling region and evaluating the shape complexity;

[0029] The color feature extracts the color information of the peeling area by calculating the color difference between the peeling area and the normal area;

[0030] The geometric features extract the geometric shape parameters of the holes by image processing method;

[0031] The grayscale feature extracts grayscale difference information inside and outside the hole by analyzing the grayscale distribution of the hole area.

[0032] As a preferred solution of the concrete building exterior wall detection method of the present invention, wherein: the artificial intelligence detection model is trained;

[0033] Collecting images of concrete building exterior walls containing defects, and annotating the defects in each image, wherein the annotations include the defect type and its location;

[0034] A convolutional neural network is selected as the detection model, wherein the output layer of the convolutional neural network has a classification unit for classifying defect types;

[0035] Divide the training set and validation set, adjust the model's weights and biases through the back-propagation algorithm, use the optimization algorithm to update the model parameters, perform iterative training, and obtain the trained artificial intelligence detection model.

[0036] As a preferred solution of the concrete building exterior wall detection method of the present invention, wherein: the determination of the distribution position of defects on the exterior wall surface includes setting a defect threshold, comparing it with the output result of the artificial intelligence detection model, and if the output result exceeds the defect threshold, it is determined that there is a defect;

[0037] Determine the location of defects on the exterior wall surface and calibrate the specific location of the defect area through coordinates or bounding boxes.

[0038] As a preferred embodiment of the concrete building exterior wall detection method of the present invention, the classification of the detected defects includes comparing the detected defect features with the known defect feature patterns learned during the training process to determine the defect type;

[0039] Generate a probability value for each defect type, and select the defect category with the highest probability value as the final classification result.

[0040] Another object of the present invention is to provide a concrete building exterior wall detection system, which can solve the problems of low efficiency, low accuracy and partial coverage in existing infrared detection and acoustic wave radar detection by constructing a concrete building exterior wall detection system.

[0041] To solve the above technical problems, the present invention provides the following technical solutions: a concrete building exterior wall detection system, comprising: an image acquisition module, used to obtain multi-angle image data of the concrete building exterior wall; a data processing module, used to preprocess the image data and extract the target area of ​​the concrete building exterior wall; a feature extraction module, used to extract feature information representing exterior wall defects from the preprocessed image data; a defect detection module, used to input the feature information into an artificial intelligence detection model to determine whether there are exterior wall defects; and a defect positioning module, used to determine the distribution position of defects on the exterior wall surface according to the output results of the detection model, and classify the detected defects.

[0042] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned concrete building exterior wall detection method when executing the computer program.

[0043] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned concrete building exterior wall detection method.

[0044] Beneficial effects of the present invention: The concrete building exterior wall detection method provided by the present invention optimizes the exterior wall defect detection process by introducing drone image acquisition and convolutional neural network (CNN) image analysis technology. Drones can efficiently capture large-area exterior walls with full coverage, avoiding the inefficiency of manual detection and the safety risks of high-altitude operations. At the same time, the convolutional neural network can automatically extract multi-level features from the image, accurately identify fine cracks and other surface defects, and has strong anti-interference capabilities, overcoming the accuracy problems of traditional infrared detection under changes in environmental factors. Through automated analysis and classification, the present invention significantly improves detection efficiency and accuracy, reduces errors caused by human factors, and reduces dependence on the technical level of operators, and can be widely used in rapid detection and long-term monitoring of large-scale buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use 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 ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0046] Figure 1 An overall flow chart of a method for detecting exterior walls of concrete buildings provided by one embodiment of the present invention.

[0047] Figure 2An overall structural diagram of a concrete building exterior wall detection system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0049] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0050] Example 1

[0051] Reference Figure 1 , as an embodiment of the present invention, provides a method for detecting exterior walls of concrete buildings, comprising:

[0052] Step S1: Acquire multi-angle image data of the concrete building exterior wall;

[0053] In step S1, the embodiment of the present invention acquires multi-angle image data of the concrete building exterior wall by using the drone, including setting the flight path and flight parameters of the drone according to the structural characteristics of the concrete building exterior wall, the parameters including the flight altitude and the tilt angle;

[0054] Control the drone to fly to different collection locations in sequence along the path to collect image data covering the exterior wall of the concrete building;

[0055] During the image acquisition process, the drone's lens exposure parameters and flight tilt angle are adjusted according to the real-time lighting conditions.

[0056] It should be noted that the drone is equipped with a high-definition camera and related sensors, flying around the concrete building exterior wall to obtain a large amount of images and data. It is ensured that the data is collected under different lighting conditions, angles and distances to obtain comprehensive and accurate information.

[0057] Step S2: preprocessing the image data to extract the target area of ​​the concrete building exterior wall;

[0058] In step S2, the embodiment of the present invention uses a noise suppression algorithm to denoise the image data;

[0059] Enhance the image data using contrast enhancement methods;

[0060] Crop the image to keep the relevant part of the concrete building exterior wall;

[0061] Convert the color space of image data.

[0062] When detecting multiple images, image registration is performed;

[0063] Image registration adjusts image positions based on geometric relationships among multiple images;

[0064] After registration, the images form a unified reference coordinate system.

[0065] It should be noted that the execution order of each processing step can be adjusted according to the detection target and processing requirements.

[0066] The embodiment of the present invention performs preprocessing operations on the collected images to improve the image quality and the accuracy of subsequent analysis. The image preprocessing operation can significantly improve the quality and usability of the image, providing a more reliable basis for subsequent feature extraction and defect detection.

[0067] Furthermore, the collected drone images may be interfered by various noises, such as electronic noise, environmental noise, etc. The purpose of denoising is to reduce the impact of these noises on image quality. Common denoising methods include median filtering, mean filtering, Gaussian filtering, etc.

[0068] The embodiment of the present invention can effectively remove salt and pepper noise through median filtering and maintain edge details of the image; mean filtering and Gaussian filtering have a good suppression effect on Gaussian noise.

[0069] Image enhancement aims to improve the contrast and brightness of the image and make the defect features more prominent. Methods such as histogram equalization and grayscale stretching can be used.

[0070] The embodiment of the present invention enhances the overall contrast of the image by redistributing the grayscale of the image through histogram equalization; the grayscale stretching specifically expands the specific grayscale range to highlight the important areas.

[0071] Since the images collected by drones may contain a large number of irrelevant areas, in order to improve the efficiency and accuracy of subsequent processing, the images need to be cropped to retain only the concrete building exterior wall.

[0072] Depending on the requirements of subsequent processing algorithms, images need to be converted from the RGB color space to other color spaces that are more suitable for analysis, such as HSV (hue, saturation, value) or grayscale space.

[0073] If multiple images are being tested, image registration is required to ensure that the same positions in different images correspond accurately, which facilitates comprehensive analysis and comparison.

[0074] Step S3: extracting feature information representing exterior wall defects from the preprocessed image data;

[0075] In step S3, the characteristic information characterizing the exterior wall defects in the embodiment of the present invention includes crack characteristics, peeling characteristics and hole characteristics;

[0076] Crack characteristics include shape characteristics, texture characteristics and gradient characteristics;

[0077] The peeling characteristics include regional characteristics and color characteristics;

[0078] Hole features include geometric features and grayscale features.

[0079] The shape feature extracts the geometric shape parameters of the crack by performing edge detection and morphological processing on the crack area; the texture feature analyzes the difference in texture between the crack area and the surrounding normal area through image processing methods to extract the texture information of the crack area; the gradient feature detects the edge and gradient changes in the image through the gradient operator to extract the gradient information of the crack area;

[0080] Specifically, shape features include the length, width, direction, curvature, etc. of the crack. By performing edge detection and morphological processing on the image, the outline of the crack is obtained, and then these shape parameters are calculated. Texture features analyze the difference in texture between the crack area and the surrounding normal concrete area. Gray-level co-occurrence matrix, local binary pattern and other methods can be used to describe texture features. Gradient features use gradient operators (such as Sobel operator, Canny operator, etc.) to detect gradient changes in images, because cracks usually cause significant changes in image gradients.

[0081] The regional feature calculates the geometric parameters of the peeling area and evaluates the shape complexity; the color feature calculates the color difference between the peeling area and the normal area to extract the color information of the peeling area;

[0082] Specifically, the regional features determine the area, perimeter, shape complexity, etc. of the peeling region. Color features The color of the peeling region may be different from the surrounding normal area, which is described by color histogram or color mean, variance and other features.

[0083] The geometric features extract the geometric shape parameters of the hole through image processing methods; the grayscale features extract the grayscale difference information inside and outside the hole by analyzing the grayscale distribution of the hole area.

[0084] Specifically, the geometric features include but are not limited to the size, shape, depth, etc. of the hole. The grayscale feature is because the grayscale value distribution inside the hole usually has certain characteristics.

[0085] Furthermore, AI algorithms, such as convolutional neural networks (CNNs) in deep learning, are used to automatically extract features from concrete exterior wall images, such as the shape, size, and direction of cracks, as well as surface peeling and discoloration, in order to extract feature information that can effectively characterize defects from preprocessed concrete building exterior wall images. For exterior wall defects, classification training and extraction can be performed based on the exterior wall defect type.

[0086] Furthermore, in actual operation, multiple feature extraction methods are often used in combination with deep learning algorithms, such as convolutional neural networks (CNNs), to automatically learn and extract more representative and discriminative features. CNNs can automatically learn the deep features of different types of defects from a large amount of image data, thereby improving the accuracy and robustness of defect detection.

[0087] In a preferred embodiment, the texture features are determined by a gray level co-occurrence matrix.

[0088] Specifically, first, the collected concrete building exterior wall image data is converted into a grayscale image. Through the grayscale operation, each pixel is represented as a single grayscale value to facilitate the subsequent analysis of texture features. After grayscale, the grayscale value is quantized from 0 to 255 to fewer grayscale levels, such as 16 or 32, to reduce the computational complexity.

[0089] Then, construct the gray level co-occurrence matrix. Set the distance parameter of the gray level pair and direction parameters ,in Represents the distance between two pixels, such as represents adjacent pixels, represents an interval of one pixel; and Indicates the relative direction between two pixels, such as (level) , (vertical) 、 (diagonal) or (Anti-diagonal). By traversing all pixels in the grayscale image and counting the number of occurrences of grayscale value combinations between each pair of pixels, a grayscale co-occurrence matrix is ​​constructed. The elements in the matrix Represents grayscale value and The number of times a pixel pair appears in the image.

[0090] Finally, each element in the matrix is ​​divided by the total number of all pixel pairs to normalize it so that it represents a probability distribution.

[0091] Based on the normalized gray-level co-occurrence matrix ,The following four texture features can be extracted, including: homogeneity, contrast, energy and entropy;

[0092] The formula for calculating homogeneity is: ; in, represents the homogeneity eigenvalue, Represents grayscale value and The probability value in the gray-level co-occurrence matrix. Homogeneity is used to measure whether the gray-level values ​​are concentrated near the diagonal of the matrix. The higher the value, the more uniform the texture of the image area.

[0093] The contrast ratio is calculated as: ; in, represents the contrast eigenvalue, Represents grayscale value and The probability value in the gray-level co-occurrence matrix. Contrast is used to measure the significance of gray-level value changes. The higher the value, the greater the gray-level difference in the area and the rougher the texture.

[0094] The energy calculation formula is: ; in, represents the energy eigenvalue, Represents grayscale value and The probability value in the gray-level co-occurrence matrix. Energy is used to measure the regularity and repeatability of image texture. The higher the value, the more regular the image texture distribution.

[0095] The entropy calculation formula is: ; in, represents the entropy eigenvalue, Represents grayscale value and The probability value in the gray-level co-occurrence matrix. Entropy is used to describe the complexity of the texture. The higher the value, the greater the randomness and complexity of the texture.

[0096] The above eigenvalues ​​can be used as a representation of the texture difference between the crack area and the surrounding normal area. By quantitatively analyzing the homogeneity, contrast, energy and entropy of the crack area, the texture characteristics of the crack can be effectively extracted. The texture features are input into the artificial intelligence detection model to classify and locate defects, which significantly improves the accuracy and reliability of defect detection.

[0097] In a preferred embodiment, the gradient feature is determined by a Sobel operator.

[0098] Specifically, the Sobel operator is expressed as, ; As above, and They represent the results of horizontal and vertical gradient detection on image A. (Note that it is the horizontal gradient, not the horizontal edge) The sum of the squares of the two can be used to obtain the gradient value of each point on the image, that is, the gradient in the x direction and the y direction is calculated at the same time at this point; ; The direction of the gradient at this point can be determined by taking the inverse tangent of the ratio of these two values. get:

[0099] Step S4: input the feature information into the artificial intelligence detection model to determine whether there are exterior wall defects;

[0100] In step S4, the embodiment of the present invention performs model training on the artificial intelligence detection model; Collect images of concrete building exterior walls containing defects and annotate the defects in each image, including the defect type and its location; A convolutional neural network is selected as the detection model. The output layer of the convolutional neural network has a classification unit for classifying defect types. The training set and validation set are divided, the weight and bias of the model are adjusted through the back-propagation algorithm, the model parameters are updated using the optimization algorithm, and iterative training is performed to obtain the trained artificial intelligence detection model.

[0101] Specifically, a large amount of labeled concrete exterior wall defect image data is used to train the AI ​​model so that it can learn the characteristic patterns of different types of defects:

[0102] Collect a large number of concrete building exterior wall images and accurately annotate the defects in these images. The annotation work is usually done by professionals, who manually mark the location and type of various defects such as cracks, spalling, holes, etc. These annotated data will be used as samples for model learning.

[0103] Common models used for image recognition and defect detection include convolutional neural networks (CNN), such as ResNet, VGG, Inception, etc. You can also use deep learning-based object detection models, such as FasterR-CNN, YOLO, etc.

[0104] During the training process, the model learns the mapping relationship between the input image and the annotation by continuously adjusting internal parameters (such as weights and biases). Training usually uses the back-propagation algorithm to calculate the gradient of the parameters, and uses optimization algorithms (such as stochastic gradient descent, Adagrad, Adadelta, etc.) to update the parameters so that the model's prediction results gradually approach the actual annotations.

[0105] In a preferred embodiment, the artificial intelligence detection model of the present invention is constructed based on a convolutional neural network (CNN).

[0106] Step S5: According to the output results of the detection model, the distribution positions of the defects on the exterior wall surface are determined, and the detected defects are classified.

[0107] In step S5, the embodiment of the present invention determines the distribution position of defects on the exterior wall surface, including setting a defect threshold, comparing it with the output result of the artificial intelligence detection model, and if the output result exceeds the defect threshold, it is judged that a defect exists; determining the position of the defect on the exterior wall surface, and calibrating the specific position of the defect area through coordinates or a bounding box.

[0108] In an optional embodiment of the present invention, the confidence matrix output by the artificial intelligence detection model is first used to screen out areas where defects may exist. Each pixel in the confidence matrix corresponds to a defect probability value determined by the model, and these probability values ​​are screened according to a pre-set defect threshold. When the confidence value of a pixel is greater than the set threshold, the pixel is judged to belong to the range where defects may exist. Subsequently, these screened high-confidence pixels are used as seed points, and the region growing algorithm is used to expand the defect area, thereby identifying the complete boundary of the defect.

[0109] Specifically, the region growing algorithm analyzes the neighborhood pixels of the seed point, includes the pixels in the neighborhood whose confidence values ​​exceed the set threshold into the defect area, and repeats this process until there are no pixels in the neighborhood that meet the conditions. After the expansion is completed, the system uses edge detection algorithms (such as the Canny operator) to further accurately extract the outer contour of the defect area. Subsequently, based on the extracted contour information, the corresponding bounding box is generated, and the geometric parameters such as the area and perimeter of the defect area are calculated. In order to facilitate practical applications, the pixel coordinates of the bounding box will be converted into the physical coordinates of the building's exterior wall in combination with the drone's flight parameters and image resolution, and aligned with the reference coordinate system of the building structure to ensure the accurate positioning of the defect position.

[0110] It should be noted that for multiple images from different perspectives, the system uses image registration technology to accurately match the feature points in each image to form a unified reference coordinate system. In this process, the key feature points in the image are first extracted using a feature point matching algorithm (such as SIFT or ORB), and the coarse registration is completed by calculating the similarity between the feature points. Subsequently, a global optimization algorithm (such as RANSAC) is used to remove incorrectly matched point pairs to ensure registration accuracy. In a unified coordinate system, multiple detection results of the same physical location are fused.

[0111] In order to handle multiple detection results in overlapping areas, the detection results of overlapping areas are weighted fused based on the quality score of each image (such as resolution, illumination uniformity, and viewing angle coverage). Specifically, the confidence value of each bounding box in the detection result is used as a weighting factor, and after fusion, a more stable defect boundary parameter is generated. In addition, the size and position of the bounding box are further optimized by combining the geometric shape and spatial distribution of each bounding box in the multi-view image to avoid redundant annotations or inconsistent detection results.

[0112] The fused detection results are converted into the actual physical coordinates of the exterior wall through geometric correction, and aligned with the reference coordinate system of the building structure, and finally a complete defect distribution calibration map is generated. This calibration map can fully reflect the distribution characteristics of building exterior wall defects and provide a high-precision reference for subsequent classification and repair solutions. Through the above improvements, the system effectively solves the problems of boundary blur, overlapping result conflicts and position deviations that may occur in multi-view images, which significantly improves the accuracy and consistency of the detection results.

[0113] Step S6: Classifying the detected defects includes comparing the detected defect features with the known defect feature patterns learned during the training process to determine the defect type; generating a probability value for each defect type, and selecting the defect category with the highest probability value as the final classification result.

[0114] In an optional embodiment of the present invention, feature information is extracted for each detected defect area. These feature information are obtained through the output of the middle layer of the artificial intelligence detection model, and combined with the image analysis results in the preprocessing link to form a complete defect feature vector. These feature vectors are compared one by one with the feature templates learned during the training process, and the matching degree between each feature vector in the area and the templates of different defect categories is calculated, so as to generate a probability value of the corresponding category for each defect.

[0115] Specifically, the matching process is based on the classification unit of the convolutional neural network (CNN), which comprehensively analyzes the feature vectors and calculates the classification probability in the form of a multi-layer neural network. For each defective area, the system selects the category with the highest probability value as the final classification result of the area, and records the classification confidence. For some areas with low classification confidence, the system will mark them as "suspected areas" and prompt manual review. Furthermore, if the bounding boxes of some defective areas are adjacent and the classification results are consistent, the system will automatically merge these areas into a whole and re-evaluate their classification results and severity.

[0116] It should be noted that in the classification process, not only the characteristic information of a single area is taken into consideration, but also the relationship between areas is comprehensively analyzed. For example, for crack defects, their shapes usually have a certain degree of ductility. When multiple crack areas form a continuous or nearly continuous distribution on the physical coordinates, these areas will be integrated into an overall crack and reclassified and rated according to their total length, width and shape complexity. This correlation analysis enhances the accuracy and robustness of the classification results.

[0117] Step S7: Analyze the inspection results and count the number, location, severity and other information of the defects. Generate a detailed inspection report, including defect images, location annotations, type descriptions and recommended repair measures.

[0118] Furthermore, drones are used regularly for inspections to monitor the development and changes of defects. AI models are updated and optimized based on new data to improve the accuracy and reliability of inspections.

[0119] The concrete building exterior wall detection method proposed in the embodiment of the present invention can quickly cover a large area of ​​building exterior walls through drones, greatly improving the speed and efficiency of detection. Compared with the traditional manual detection method, the detection task can be completed in a shorter time. The risk of inspectors needing to work at high altitudes during manual detection is avoided, and potential safety accidents are reduced. Images can be collected from multiple angles and different distances. Combined with the powerful analysis ability of AI, various types and subtle defects can be detected more comprehensively and accurately, reducing the probability of missed detection. A large number of images and data generated during the detection process can be stored and recorded, which is convenient for subsequent query, analysis and comparison, and is conducive to long-term monitoring and maintenance planning of the building exterior wall condition. There is no direct contact with the building exterior wall, and no damage or interference will be caused to the building structure. The detection results can be obtained in real time, problems can be discovered in time, and emergency repair measures can be taken. It can work under various complex weather and environmental conditions, and is not restricted by factors such as height and terrain. Through the analysis of the detection data, a scientific decision-making basis is provided for building maintenance and repair, and the pertinence and effectiveness of maintenance work are improved.

[0120] Example 2

[0121] Reference Figure 2 , which is an embodiment of the present invention, provides a concrete building exterior wall detection system, including: an image acquisition module 100, a data processing module 200, a feature extraction module 300, a defect detection module 400 and a defect positioning module 500.

[0122] The image acquisition module 100 is used to obtain multi-angle image data of the concrete building exterior wall.

[0123] Specifically, image data can be acquired through drones. The drones cover all areas of the building's exterior walls according to the preset flight path and flight parameters (including flight altitude, tilt angle, etc.). During the flight, the lens exposure parameters are adjusted in real time to adapt to different lighting conditions to ensure the integrity and clarity of image acquisition. When acquisition failure or signal loss is detected, the system can adjust the path or parameters through a feedback mechanism to ensure the continuity and reliability of data acquisition.

[0124] The data processing module 200 is used to pre-process the collected image data and extract the target area of ​​the concrete building exterior wall. Specifically, it includes operations such as image denoising, contrast enhancement, cropping and color space conversion to improve image quality and data availability. When multiple images are involved, the data processing module can also perform image registration, adjust the image position through geometric relationships, and form a unified reference coordinate system, laying the foundation for subsequent feature extraction and analysis.

[0125] The feature extraction module 300 is used to extract feature information representing exterior wall defects from preprocessed image data. By using gray-level co-occurrence matrix to analyze texture features, Sobel operator to extract gradient features, and morphological processing to obtain the shape features of cracks, the module can automatically identify the multi-dimensional features of cracks, spalling and holes. Combined with convolutional neural networks (CNN), the module can learn and extract deep features from a large number of images, further enhancing the ability to express complex defect features.

[0126] The defect detection module 400 is used to input feature information into the artificial intelligence detection model to determine whether there are exterior wall defects. The system combines the labeled training data and optimizes the detection model based on the convolutional neural network, and improves the detection accuracy and robustness of the model through iterative training. The detection module can identify defects in the input image in real time and output the determination result of the defect area based on the preset threshold.

[0127] The defect location module 500 is used to determine the specific location of the defect on the exterior wall surface according to the output results of the detection model and classify the detected defects. This module marks the defect area with a bounding box and automatically classifies the defect type, such as cracks, peeling, holes, etc., according to the output probability of the model. The system further generates coordinate information of the defect location to provide data support for subsequent maintenance.

[0128] Specifically, the pre-processed newly collected drone images are input into the trained artificial intelligence detection model, and the model will automatically detect and classify the types of defects, such as cracks, peeling, holes, etc.

[0129] In a preferred embodiment, the method of the present invention determines that the types of defects in the concrete building exterior wall include cracks, spalling and holes.

[0130] Specifically, the preprocessed and feature-extracted concrete building exterior wall image is input into the trained detection model. The model analyzes each pixel or region in the image to determine whether there is an anomaly. Based on the output of the model, a threshold is usually set to determine whether there is a defect. If the feature value of a certain area exceeds the threshold, it is considered to be a defective area. The location of the defect in the image is determined, usually marked in the form of coordinates or bounding boxes. This helps with the subsequent specific description and evaluation of the defect.

[0131] Specifically, the detected defect features are compared with known defect feature patterns of various types. These feature patterns are learned during the model training process. Consider multiple features, such as shape, size, texture, color, etc., to accurately classify the type of defect. For example, a slender defect may be classified as a crack, and a large area of ​​detachment may be classified as peeling. The model generates a probability value for each possible defect category, indicating the possibility that the defect belongs to that category. The category with the highest probability is usually selected as the final classification result. For some similar or complex defect types, further detailed analysis and judgment may be required, combined with more professional knowledge and experience for accurate classification.

[0132] Based on the limitations of traditional technical methods for detecting defects in the exterior walls of concrete buildings, this embodiment combines the advantages of modern technologies such as drones and AI to build a comprehensive detection system that can better ensure the safety and maintenance of buildings and improve the efficiency, accuracy and safety of detection.

[0133] Specifically, drones can fly quickly and take high-resolution images of large-area building exterior walls, greatly improving the speed of detection. Compared with manual detection, drones can complete large-scale scanning and data collection in a short time, improving the efficiency of large-scale coverage; drones can transmit the collected images to the ground station in real time, and use AI algorithms to process and analyze them instantly. AI algorithms process a large amount of image data at the same time, quickly identify and locate defects, improve the efficiency of data processing, and shorten the detection cycle.

[0134] Specifically, drones can photograph exterior walls from different angles and distances, providing comprehensive and multi-perspective image data to reduce omissions and blind spots. By using deep learning models such as convolutional neural networks (CNNs), algorithms for identifying various types of defects can be trained to accurately detect different defects such as cracks, peeling, weathering, and leakage, thereby reducing the impact of human subjective judgment and improving the accuracy and consistency of detection results.

[0135] Specifically, drones can fly and inspect at high altitudes and in dangerous areas, reducing the need for manual high-altitude operations and significantly reducing the safety risks of operators. Operators can control drones for inspection in safe areas on the ground without approaching dangerous areas, further ensuring personnel safety.

[0136] Specifically, drones can conduct regular inspections according to a set schedule, promptly detect and monitor changes and potential defects in building exterior walls, and provide continuous data recording and analysis.

[0137] Example 3

[0138] An embodiment of the present invention is different from the first two embodiments in that:

[0139] Specifically, if the concrete building exterior wall detection method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0140] The computer program includes several instructions that can cause a computer device (such as a server, a personal computer or other network device) to perform the following steps:

[0141] Control image acquisition equipment (such as drones) to collect multi-angle image data of concrete building exterior walls according to set paths and parameters;

[0142] Preprocess the collected image data, including noise suppression, contrast enhancement, image cropping, and color space conversion;

[0143] Extract feature information of exterior wall defects in preprocessed images, such as using gray-level co-occurrence matrix to analyze texture features, using Sobel operator to extract gradient features, and combining deep learning algorithms (such as convolutional neural networks) to extract multi-level features;

[0144] Input the feature information into the artificial intelligence detection model, use the trained model to determine whether there are exterior wall defects in the image, and mark the specific location of the defects;

[0145] Based on the inspection results, defects are classified, such as cracks, peeling, holes, etc., and an inspection report is generated containing the defect type, location, severity and distribution.

[0146] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0147] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0148] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0149] The computer program of the present invention further supports real-time operation, for example, through the communication interface with the drone, it can realize real-time image acquisition and preprocessing, and directly push the detection results to the control terminal after defect detection. Compared with traditional detection methods, the program can significantly reduce manual intervention and provide more efficient and accurate exterior wall defect detection services.

[0150] Example 4

[0151] As an embodiment of the present invention, a method for detecting the exterior wall of a concrete building is provided. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0152] A comprehensive evaluation was conducted on traditional manual detection methods, infrared detection methods, sound wave and radar detection methods, and the detection method of the present invention. The experimental site is the outer wall area of ​​a simulated concrete building, with a wall height of 30 meters and a width of 50 meters. A variety of defects are preset on the outer wall surface, including cracks, spalling and holes, among which the crack length is 0.1 to 5 cm and the width is 0.05 to 0.5 cm; the spalling area is 5 to 50 square centimeters; and the hole diameter is 0.5 to 2 cm. The defective areas are marked in advance for subsequent comparative analysis.

[0153] The experiment was divided into four groups, using manual detection, infrared detection, sound wave and radar detection, and the method of the present invention for detection. The experimental environment set up a variety of interference conditions, including lighting conditions (alternating strong light and shadow), surface pollutants (dust coverage), and meteorological interference with a wind speed of 5 to 10 meters per second to simulate complex detection scenarios.

[0154] The manual inspection method was completed by three experienced inspectors, who used rope climbing tools and aerial work platforms, as well as crack rulers and magnifying glasses, to inspect the exterior wall surface section by section. During the inspection, the inspectors needed to climb to different heights of the exterior wall and record the length and width of the cracks, as well as the area and location of the peeling and holes. After completing the full wall inspection, the time required was recorded, and the inspection results were compared with the true value, and the number of missed inspections and false inspections was counted.

[0155] Infrared inspection is performed using an infrared thermal imager with a resolution of 640x480. The inspection equipment is fixed on the ground 5 meters away from the exterior wall, and the instrument parameters are adjusted to adapt to different temperature difference conditions. The inspector gradually scans the entire exterior wall, marks the defective area according to the temperature difference in the thermal imaging image, and records the relevant data. After the inspection is completed, the thermal imaging data is analyzed and the deviation between the inspection result and the true value is counted.

[0156] Acoustic and radar detection uses handheld acoustic and radar equipment, and the detection depth is set to 0 to 5 cm. The inspector inspects the wall surface piece by piece, and records the intensity and position of the defect reflection in the radar echo signal. After the inspection is completed, the defect area is marked in combination with the echo image of the radar signal, and the relevant data is recorded. The test results are compared with the true value, and the missed detection and false detection are counted.

[0157] The detection method of the present invention completes image acquisition and detection work through a drone. The drone is equipped with a high-resolution camera and covers the entire exterior wall area according to the set flight path. During the flight, the drone adjusts the exposure parameters of the lens according to the real-time lighting conditions, and ensures image clarity through the built-in anti-shake system. The collected image data is transmitted to the ground control station in real time, and the convolutional neural network (CNN) detection model automatically identifies and classifies the defects in the image. The detection model marks the coordinates and type of the defects and generates a detection report containing the defect distribution and severity. After the detection is completed, the detection time and classification results are recorded and compared with the true value data. The experimental results are shown in Table 1.

[0158] Table 1 Comparison of experimental results Detection Methods Detection time (minutes) Accuracy (%) Recall rate (%) F1 value (%) Number of missed detections False positives Manual inspection 320 70 65 67.4 25 12 Infrared detection 150 78 62 69.2 30 8 Acoustic and radar detection 180 82 75 78.3 15 7 Method of the present invention 75 90 88 89 6 3

[0159] The present invention uses a drone equipped with a high-definition camera to obtain multi-angle image data of the exterior wall of a concrete building, significantly improving the coverage efficiency of exterior wall detection. In the experiment, the traditional manual detection method requires long-term high-altitude operations and checks the exterior wall area block by block, resulting in significantly extended detection time and high safety risks. The automated flight and real-time adjustment functions of the drone achieve rapid coverage of large-area exterior walls, and only require a short time to complete the data collection task, which significantly shortens the total detection time in the experiment.

[0160] In the image processing and feature extraction process, the present invention uses a convolutional neural network (CNN) combined with a grayscale co-occurrence matrix to analyze texture features. Compared with infrared detection methods, CNN can extract multi-level image features and meet the detection needs of fine cracks, peeling and other complex defects. However, due to the limitations of temperature differences, infrared detection shows a high missed detection rate for superficial cracks and tiny peeling areas. In addition, infrared detection is easily affected by changes in ambient light. The present invention uses drones to adjust exposure parameters in real time and combines artificial intelligence detection models to overcome the interference caused by changes in ambient light, and has shown stronger stability and adaptability in experiments.

[0161] Compared with the sonic and radar detection methods, the present invention can not only effectively detect superficial defects, but also accurately locate and classify tiny cracks and peeling areas. The applicability of sonic radar detection to large-scale building detection is relatively poor, and it is mainly limited to local areas or specific defect types. In the experiment, its detection efficiency and coverage were lower than those of the present invention. The artificial intelligence detection model adopted by the present invention can quickly and accurately complete multi-dimensional feature analysis through training and optimization with a large amount of annotated data, effectively improve the recognition rate of defects, and reduce the number of false detections and missed detections. It is significantly better than traditional detection methods in terms of recall rate and accuracy.

[0162] The overall automated processing flow of the present invention integrates collection, processing, analysis and classification, effectively reducing the dependence on the technical level of the operator. Other existing technologies are time-consuming and highly dependent on personnel experience. In experiments, they show high false detection and missed detection rates, and unstable detection results. The present invention, through automated detection technology, reduces human intervention while ensuring the objectivity and consistency of the results. Combined with the analysis of experimental results, the present invention is superior to existing technologies in terms of detection efficiency, accuracy, stability and safety, and meets the needs of large-scale building detection.

[0163] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting the exterior wall of a concrete building, characterized in that: include: Acquire multi-angle image data of concrete building exterior walls; Preprocessing the image data to extract a target area of ​​a concrete building exterior wall; Extracting feature information representing exterior wall defects from preprocessed image data; Inputting the feature information into an artificial intelligence detection model to determine whether there are exterior wall defects; According to the output results of the detection model, the distribution positions of defects on the exterior wall surface are determined and the detected defects are classified; Crack characteristics, spalling characteristics and hole characteristics; The crack characteristics include shape characteristics, texture characteristics and gradient characteristics; The peeling characteristics include regional characteristics and color characteristics; The hole features include geometric features and grayscale features; The shape feature extracts the geometric shape parameters of the crack by performing edge detection and morphological processing on the crack area; The texture feature uses an image processing method to analyze the difference in texture between the crack area and the surrounding normal area, and extracts the texture information of the crack area; The gradient feature detects the edge and gradient changes in the image through a gradient operator, and extracts the gradient information of the crack area; The regional features are obtained by calculating geometric parameters of the peeling region and evaluating the shape complexity; The color feature extracts the color information of the peeling area by calculating the color difference between the peeling area and the normal area; The geometric features extract the geometric shape parameters of the holes by image processing method; The grayscale feature extracts grayscale difference information inside and outside the hole by analyzing the grayscale distribution of the hole area.

2. The method for detecting the exterior wall of a concrete building according to claim 1, characterized in that: The method of acquiring multi-angle image data of the concrete building exterior wall includes: Control the drone to fly to different collection positions in sequence along the path to collect image data covering the exterior wall of the concrete building; During the image acquisition process, the drone's lens exposure parameters and flight tilt angle are adjusted according to the real-time lighting conditions.

3. The method for detecting the exterior wall of a concrete building according to claim 1 or 2, characterized in that: The pre-processing comprises: Use noise suppression algorithm to denoise image data; Enhance the image data using contrast enhancement methods; Crop the image to keep the relevant part of the concrete building exterior wall; Converts the color space of image data.

4. The method for detecting the exterior wall of a concrete building according to claim 3, characterized in that: The preprocessing further includes: When detecting multiple images, image registration is performed; The image registration adjusts the image positions according to the geometric relationships among the multiple images; After registration, the images form a unified reference coordinate system.

5. The method for detecting the exterior wall of a concrete building according to any one of claims 1, 2 or 4, characterized in that: The artificial intelligence detection model is trained, including: Collecting images of concrete building exterior walls containing defects, and annotating the defects in each image, wherein the annotations include the defect type and its location; A convolutional neural network is selected as the detection model, wherein the output layer of the convolutional neural network has a classification unit for classifying defect types; Divide the training set and validation set, adjust the model's weights and biases through the back-propagation algorithm, use the optimization algorithm to update the model parameters, perform iterative training, and obtain the trained artificial intelligence detection model.

6. The method for detecting the exterior wall of a concrete building according to any one of claims 1, 2 or 4, characterized in that: The determining of the distribution position of defects on the exterior wall surface includes: Set a defect threshold and compare it with the output result of the artificial intelligence detection model. If the output result exceeds the defect threshold, it is judged that a defect exists; Determine the location of defects on the exterior wall surface and calibrate the specific location of the defect area through coordinates or bounding boxes.

7. The method for detecting the exterior wall of a concrete building according to any one of claims 1, 2 or 4, characterized in that: The detected defects are classified, including: Compare the detected defect features with the known defect feature patterns learned during the training process to determine the defect type; Generate a probability value for each defect type, and select the defect category with the highest probability value as the final classification result.

8. A concrete building exterior wall detection system, characterized in that: include: An image acquisition module (100) is used to acquire multi-angle image data of the concrete building exterior wall; A data processing module (200) is used to pre-process the image data and extract a target area of ​​a concrete building exterior wall; A feature extraction module (300), used to extract feature information representing exterior wall defects from the pre-processed image data; A defect detection module (400) is used to input the feature information into an artificial intelligence detection model to determine whether there are exterior wall defects; and The defect location module (500) is used to determine the distribution position of defects on the exterior wall surface according to the output results of the detection model, and to classify the detected defects.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the concrete building exterior wall detection method are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the concrete building exterior wall detection method are implemented.

Images