Building crack intelligent identification method and system based on multi-source data fusion

Abstract

The invention relates to the technical field of data processing, and discloses a building crack intelligent identification method and system based on multi-source data fusion, and the method comprises the steps: extracting crack initial contour features from a surface optical enhancement feature map, and collecting the features into crack space distribution features; compensating and registering the deformation quantity based on the thermal expansion characteristic of the material to obtain a fusion base image; performing crack trend adaptive frequency domain filtering processing on the fusion base image to obtain a feature enhanced image; performing multi-physical-field probability fusion on visible light texture, infrared temperature and a depth curvature gradient field in the feature enhanced image based on the brittle fracture parameters of the material to obtain a crack probability distribution diagram; the topological connectivity is optimized through multi-scale morphological iteration, and a continuous crack mask is obtained; analyzing the spatial trend of the building structure based on the stress distribution characteristics of the building structure to obtain a crack risk level; according to the invention, the accuracy of building crack intelligent identification based on multi-source data fusion can be improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for intelligent identification of building cracks based on multi-source data fusion. Background Technology

[0002] Current technologies for building crack identification suffer from inadequate fusion processing of multi-source data, often failing to fully integrate feature information from different physical fields, resulting in insufficient completeness and accuracy of crack feature extraction. Furthermore, the lack of effective compensation mechanisms for environmental factors leads to low registration accuracy of crack images, impacting the reliability of subsequent identification results.

[0003] Furthermore, existing technologies have relatively simple processing methods in areas such as crack feature enhancement, topological connectivity optimization, and risk level assessment. They fail to take into account key factors such as crack direction and material properties for targeted processing, resulting in low efficiency in identifying complex cracks and making it difficult to meet the needs of rapid and accurate crack identification in actual building safety monitoring. Summary of the Invention

[0004] This invention provides a method and system for intelligent identification of building cracks based on multi-source data fusion, the main purpose of which is to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for intelligent identification of building cracks based on multi-source data fusion, comprising: S1. The initial contour features of the cracks extracted from the surface optical enhancement feature map are aggregated into crack spatial distribution features; S2. Based on the thermal expansion characteristics of the material, the deformation of the spatial distribution features of the cracks is compensated and registered to obtain the fused base image of the building cracks; S3. Perform adaptive frequency domain filtering on the fused base image to obtain a feature-enhanced image of the building cracks; S4. Based on the material brittle fracture parameters, perform multi-physics probability fusion on the visible light texture gradient field, infrared temperature gradient field and depth curvature gradient field in the feature enhancement image to obtain the crack probability distribution map of the building crack; S5. The topological connectivity of the crack probability distribution map is optimized through multi-scale morphological iteration to obtain a continuous crack mask for the building cracks; S6. Based on the stress distribution characteristics of the building structure, the spatial orientation of the continuous crack mask is analyzed to obtain the crack risk level of the building crack.

[0006] In a preferred embodiment, the step of summarizing the initial crack contour features extracted from the surface optical enhancement feature map into crack spatial distribution features includes: Edge detection is performed on the surface optical enhancement feature map to obtain the initial edge feature map of the building crack; By connecting the fracture regions in the initial edge feature map based on pixel connectivity constraints, a continuous crack skeleton map of the building crack is obtained. The spatial coordinates and direction vectors of the continuous crack skeleton map are extracted to obtain the spatial distribution characteristics of the building cracks.

[0007] In a preferred embodiment, the deformation compensation registration based on the thermal expansion characteristics of the material to obtain the fused base image of the building cracks includes: Obtain the deformation compensation parameters corresponding to the current ambient temperature from the material thermal expansion property database; The deformation compensation parameters are inversely mapped to the deformation of the spatial distribution characteristics of the cracks to obtain the compensation deformation feature set of the building cracks; The compensated deformation feature set is superimposed onto the original crack image to obtain the fused base image of the building crack.

[0008] In a preferred embodiment, the step of superimposing the compensated deformation feature set onto the original crack image to obtain the fused base image of the building crack includes: The original crack image is calibrated by using the geometric space coordinates of the compensation deformation feature set to obtain the calibrated image of the original crack image. The calibration image is fused with the multidimensional physical features of the compensation deformation feature set through channel fusion to obtain the fused feature tensor of the calibration image; Spatial domain reconstruction is performed on the fused feature tensor to obtain the fused base image of the building crack.

[0009] In a preferred embodiment, the step of performing frequency domain filtering on the fused base image to adaptively filter the crack orientation and obtain the feature-enhanced image of the building crack includes: The frequency domain coefficient matrix of the building cracks is obtained by performing a fast Fourier transform on the fused base image. Based on the directional distribution characteristics of the initial contour features of the crack, a band-stop filter transfer function for the building crack is constructed, wherein the formula for the band-stop filter transfer function is as follows:

[0010] In the formula: Transfer function values ​​to the frequency domain. For spatial frequency variables, It is the square of the stopband center frequency reference value. This is the stopband width adjustment factor. The filter order; The frequency domain coefficient matrix is ​​multiplied by the band-stop filter transfer function to obtain the filtered frequency domain coefficients of the building cracks.

[0011] In a preferred embodiment, constructing the band-stop filter transfer function of the building crack based on the directional distribution characteristics of the initial crack contour features includes: Identify line segment features with a length greater than a threshold in the initial contour features of the crack; The dispersion analysis of the direction angle of the line segment element is performed to obtain the direction dispersion parameter of the line segment element; Based on the directional dispersion parameter, the stopband width of the band-stop filter is configured, and the transfer function of the band-stop filter for the building crack is obtained.

[0012] In a preferred embodiment, the multi-physics probability fusion of the visible light texture gradient field, infrared temperature gradient field, and depth curvature gradient field in the feature-enhanced image based on material brittle fracture parameters to obtain the crack probability distribution map of the building cracks includes: The visible light texture gradient field is normalized by structural tensor to obtain the first field strength distribution of the building cracks. The thermal diffusion equation is inverted on the infrared temperature gradient field to obtain the second field strength distribution of the building cracks; Based on the curvature tensor eigenvalues ​​derived from the deep point cloud data, the third field strength distribution of the building cracks is obtained. The crack probability value of the building crack is obtained by weighted fusion of the first field strength distribution, the second field strength distribution and the third field strength distribution based on energy density. The crack probability value is mapped to the corresponding spatial location of the feature-enhanced image to obtain the crack probability distribution map of the building crack.

[0013] In a preferred embodiment, the step of optimizing the topological connectivity of the crack probability distribution map through multi-scale morphological iteration to obtain a continuous crack mask for the building cracks includes: Spatial connectivity analysis is performed on the fractured regions in the crack probability distribution map to obtain the primary connectivity mask of the building cracks; Based on the regional continuity of the residual fracture gap in the primary connectivity mask, the boundary of the unconnected region of the building crack is determined; The broken sections in the boundary of the unconnected area are filled and connected to form a continuous crack mask for the building cracks.

[0014] In a preferred embodiment, the step of analyzing the spatial orientation of the continuous crack mask based on the stress distribution characteristics of the building structure to obtain the crack risk level of the building crack includes: A set of rules for mapping crack orientation to stress direction is generated based on a library of stress distribution characteristics of building structures. The spatial orientation of the continuous crack mask is decomposed into a sequence of independent directional components; The crack risk level of the building crack is obtained by matching the coupling strength between the independent directional component sequence and the mapping rule set.

[0015] To address the aforementioned problems, the present invention also provides a building crack intelligent identification system based on multi-source data fusion, the system comprising: The contour feature aggregation module is used to aggregate the initial contour features of cracks extracted from the surface optical enhancement feature map into crack spatial distribution features. A thermal deformation compensation registration module is used to compensate and register the deformation of the spatial distribution characteristics of the cracks based on the thermal expansion characteristics of the material, so as to obtain a fused base image of the building cracks. An adaptive frequency domain filtering module is used to perform adaptive frequency domain filtering on the fused base image to obtain a feature-enhanced image of the building cracks. The multi-physics probability fusion module is used to perform multi-physics probability fusion on the visible light texture gradient field, infrared temperature gradient field and depth curvature gradient field in the feature enhancement image based on the material brittle fracture parameters, so as to obtain the crack probability distribution map of the building cracks. The topology connectivity optimization module is used to optimize the topology connectivity of the crack probability distribution map through multi-scale morphological iteration to obtain a continuous crack mask of the building cracks. The structural risk analysis module is used to analyze the spatial orientation of the continuous crack mask based on the stress distribution characteristics of the building structure, and to obtain the crack risk level of the building crack.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention significantly improves the effectiveness of intelligent building crack identification through multi-source data fusion and multi-step refined processing. Based on the thermal expansion characteristics of materials, it performs deformation compensation registration, combined with adaptive frequency domain filtering of crack direction, which can accurately extract and enhance crack features, making the spatial distribution, edge contours, and other information of cracks clearer and more accurate, thus greatly improving the accuracy of identification.

[0017] 2. Simultaneously, by using multi-physics probability fusion and multi-scale morphological iterative optimization, the topological connectivity of the crack probability distribution map is effectively improved. Combined with the stress distribution characteristics of the building structure to analyze the crack risk level, not only can a continuous and complete crack mask be obtained efficiently, but crack risk can also be scientifically assessed, comprehensively improving the efficiency and reliability of building crack identification and providing strong support for building safety monitoring. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a method for intelligent identification of building cracks based on multi-source data fusion, provided in an embodiment of the present invention. Figure 2 A functional module diagram of a building crack intelligent identification system based on multi-source data fusion provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] This application provides a method for intelligent identification of building cracks based on multi-source data fusion. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for intelligent identification of building cracks based on multi-source data fusion can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0021] Reference Figure 1 The diagram shown is a flowchart illustrating a method for intelligent identification of building cracks based on multi-source data fusion, according to an embodiment of the present invention. In this embodiment, the method for intelligent identification of building cracks based on multi-source data fusion includes: S1. The initial contour features of the cracks extracted from the surface optical enhancement feature map are aggregated into crack spatial distribution features; In this embodiment of the invention, the step of collecting the initial contour features of the cracks extracted from the surface optical enhancement feature map into crack spatial distribution features includes: Edge detection is performed on the surface optical enhancement feature map to obtain the initial edge feature map of the building crack; By connecting the fracture regions in the initial edge feature map based on pixel connectivity constraints, a continuous crack skeleton map of the building crack is obtained. The spatial coordinates and direction vectors of the continuous crack skeleton map are extracted to obtain the spatial distribution characteristics of the building cracks.

[0022] Specifically, edge detection is performed on the surface optical enhancement feature map, and an edge detection tool is used to traverse each pixel in the image.

[0023] Furthermore, by comparing the grayscale value difference between a pixel and its surrounding neighboring pixels, pixels with abrupt changes in grayscale value are identified. These pixels are then connected to form the edge lines of the building crack, thus obtaining the initial edge feature map of the building crack.

[0024] Furthermore, based on pixel connectivity constraints, the broken regions in the initial edge feature map are connected. First, the positions of all broken regions in the initial edge feature map are determined, based on the continuity of adjacent pixels in spatial position and the similarity of gray values.

[0025] Furthermore, starting from both ends of the fracture zone, pixels that meet the pixel connectivity conditions are searched and connected step by step to fill in the fractured part, so that the originally discontinuous crack edges form complete lines, thus obtaining a continuous crack skeleton diagram of the building crack.

[0026] Furthermore, the spatial coordinates and direction vectors of the continuous crack skeleton image are extracted, and each pixel on the continuous crack skeleton image is traversed to record the coordinates of each pixel in the image coordinate system, thus forming the spatial coordinates of the crack.

[0027] Furthermore, by analyzing the positional relationship between each pixel and its adjacent pixels, the direction angle of crack extension at that point is calculated to form a direction vector. By integrating all spatial coordinates and direction vectors, the spatial distribution characteristics of building cracks are obtained.

[0028] In summary, edge detection of the surface optical enhancement feature map yields an initial edge feature map, which can accurately capture the edge contour of building cracks and clearly present the basic shape of the cracks. This provides accurate basic data for subsequent crack analysis, ensuring high accuracy in crack identification from the source.

[0029] In summary, by connecting the broken areas in the initial edge feature map based on pixel connectivity constraints to obtain a continuous crack skeleton map, the problem of breakage that may occur in the initial edge detection can be effectively compensated for. This allows the originally discontinuous crack edges to form a complete skeleton structure, fully restore the overall direction of the crack, avoid analysis bias caused by crack breakage, and improve the completeness of crack morphology description.

[0030] In summary, extracting spatial coordinates and direction vectors from a continuous crack skeleton map reveals the spatial distribution characteristics of cracks. This allows for a comprehensive record of the specific location and extension direction of cracks in space, systematically integrating the spatial and directional information of cracks. This provides detailed and comprehensive data support for subsequent crack risk assessment and deformation analysis, enhancing the accuracy and comprehensiveness of understanding crack characteristics.

[0031] S2. Based on the thermal expansion characteristics of the material, the deformation of the spatial distribution features of the cracks is compensated and registered to obtain the fused base image of the building cracks; In this embodiment of the invention, the step of compensating and registering the deformation of the spatial distribution characteristics of the cracks based on the thermal expansion properties of the materials to obtain a fused base image of the building cracks includes: Obtain the deformation compensation parameters corresponding to the current ambient temperature from the material thermal expansion property database; The deformation compensation parameters are inversely mapped to the deformation of the spatial distribution characteristics of the cracks to obtain the compensation deformation feature set of the building cracks; The compensated deformation feature set is superimposed onto the original crack image to obtain the fused base image of the building crack.

[0032] In this embodiment of the invention, the step of superimposing the compensated deformation feature set onto the original crack image to obtain the fused base image of the building crack includes: The original crack image is calibrated by using the geometric space coordinates of the compensation deformation feature set to obtain the calibrated image of the original crack image. The calibration image is fused with the multidimensional physical features of the compensation deformation feature set through channel fusion to obtain the fused feature tensor of the calibration image; Spatial domain reconstruction is performed on the fused feature tensor to obtain the fused base image of the building crack.

[0033] Specifically, the deformation compensation parameters corresponding to the current ambient temperature are obtained from the material thermal expansion characteristics database. First, the real-time temperature value of the building's environment is collected by a temperature sensor, and then the temperature value is entered into the material thermal expansion characteristics database for retrieval.

[0034] Furthermore, the database stores parameters such as the coefficient of thermal expansion and deformation rate of building materials at different temperatures. Parameters that perfectly match the current ambient temperature are retrieved from the database and determined as deformation compensation parameters.

[0035] Furthermore, the deformation compensation parameters are back-mapped to the deformation of the crack spatial distribution characteristics. First, the deformation data caused by temperature changes in the crack spatial distribution characteristics are extracted, including the crack length, width and position offset.

[0036] Furthermore, based on the thermal expansion law recorded in the deformation compensation parameters, a correspondence between deformation and compensation amount is established. The compensation value required to offset the temperature effect is obtained through reverse calculation. These compensation values ​​are arranged according to the spatial distribution law of cracks to obtain the compensation deformation feature set of building cracks.

[0037] Furthermore, the compensation deformation feature set is superimposed onto the original crack image to first determine the pixel position and shape of the crack in the original crack image.

[0038] Furthermore, according to the compensation values ​​in the compensation deformation feature set, the position and size of each pixel in the original crack image are adjusted to correct the deformation caused by temperature changes. After the adjustment is completed, the compensated crack features are fused with the background information of the original crack image to form a complete image, thus obtaining the fused base image of the building crack.

[0039] Specifically, the original crack image is calibrated based on the geometric spatial coordinates of the compensation deformation feature set. First, the geometric spatial coordinates of each feature point recorded in the compensation deformation feature set are extracted, and then the coordinate position of the corresponding crack feature point in the original crack image is determined.

[0040] Furthermore, by comparing the coordinate differences between the two, the specific values ​​that the original crack image needs to be translated, rotated or scaled are calculated. The original crack image is then adjusted according to these values ​​so that the crack feature points in the original crack image correspond completely to the geometric space coordinates in the compensation deformation feature set, thus obtaining the calibration image of the original crack image.

[0041] Furthermore, the calibration image and the multidimensional physical features of the compensation deformation feature set are fused through channels. First, the pixel channel information of the calibration image is determined, including channel data such as brightness and color.

[0042] Furthermore, multidimensional physical features, such as the width variation and depth estimate of the crack, are extracted from the compensation deformation feature set and converted into a data format consistent with the calibration image channel format.

[0043] Furthermore, the converted physical feature data is then superimposed and integrated with the pixel channel data of the calibration image at corresponding positions to form a comprehensive data set containing image information and physical features, thus obtaining the fused feature tensor of the calibration image.

[0044] Furthermore, the spatial domain of the fused feature tensor is reconstructed. First, the spatial location information and feature data contained in the fused feature tensor are analyzed. Then, according to the distribution pattern of building cracks in the actual space, the arrangement order and correlation of each feature data in the spatial domain are determined.

[0045] In detail, the data in the fused feature tensor is then mapped onto a two-dimensional image plane using a spatial reconstruction tool to restore the spatial morphology and distribution of physical features of the cracks, while maintaining the image clarity and feature integrity, ultimately obtaining a fused base image of the building cracks.

[0046] In summary, obtaining the deformation compensation parameters corresponding to the current ambient temperature from the material thermal expansion characteristics database can accurately obtain compensation basis that matches the real-time ambient temperature, ensuring that subsequent deformation compensation has clear and realistic parameter support, and guaranteeing the pertinence and accuracy of compensation from the source.

[0047] In summary, by inversely mapping the deformation compensation parameters to the deformation of the crack spatial distribution characteristics, a compensation deformation feature set can be obtained. This allows for the offsetting of the influence of temperature changes on crack morphology through inverse calculation, accurately quantifying the degree of deformation that needs to be compensated, making the compensated crack features more closely match their true state, and improving the realism of crack features.

[0048] In summary, superimposing the compensated deformation feature set onto the original crack image to obtain the fused base image can organically integrate the crack features after temperature deformation compensation with the information of the original image, forming a base image that takes into account both the details of the original image and the features after accurate compensation. This provides more reliable image data for subsequent crack analysis, feature extraction, etc., and enhances the accuracy and effectiveness of subsequent processing.

[0049] In summary, the calibration image is obtained by calibrating the original crack image based on the geometric spatial coordinates of the compensation deformation feature set. By accurately aligning the spatial positions of the two, the spatial offset between the original image and the compensation features can be eliminated, ensuring that each feature corresponds in the same spatial coordinate system during subsequent fusion. This provides an accurate spatial reference for fusion processing and improves the spatial consistency of the fusion.

[0050] In summary, channel fusion of the calibration image with the multidimensional physical features of the compensation deformation feature set to obtain the fused feature tensor can deeply integrate the visual information of the calibrated image with the physical property data of the crack, expanding the image information of a single channel into comprehensive data containing multidimensional features, enriching the information content of the features, and providing a more comprehensive basis for subsequent analysis.

[0051] In summary, spatial domain reconstruction of the fused feature tensor yields a fused base image, which can transform the integrated multidimensional feature data back into an image format that conforms to human observation and machine processing. This fully preserves the calibrated spatial positional relationships and multidimensional physical features, forming a base image that combines spatial accuracy and feature richness. This provides high-quality data support for subsequent crack feature extraction and analysis, ensuring the accuracy and reliability of subsequent processing.

[0052] S3. Perform adaptive frequency domain filtering on the fused base image to obtain a feature-enhanced image of the building cracks; In this embodiment of the invention, the step of performing frequency domain filtering on the fused base image to adaptively filter the crack direction and obtain the feature-enhanced image of the building crack includes: The frequency domain coefficient matrix of the building cracks is obtained by performing a fast Fourier transform on the fused base image. Based on the directional distribution characteristics of the initial contour features of the crack, a band-stop filter transfer function for the building crack is constructed, wherein the formula for the band-stop filter transfer function is as follows:

[0053] In the formula: Transfer function values ​​to the frequency domain. For spatial frequency variables, It is the square of the stopband center frequency reference value. This is the stopband width adjustment factor. The filter order; The frequency domain coefficient matrix is ​​multiplied by the band-stop filter transfer function to obtain the filtered frequency domain coefficients of the building cracks.

[0054] In this embodiment of the invention, constructing the band-stop filter transfer function of the building crack based on the directional distribution characteristics of the initial contour features of the crack includes: Identify line segment features with a length greater than a threshold in the initial contour features of the crack; The dispersion analysis of the direction angle of the line segment element is performed to obtain the direction dispersion parameter of the line segment element; Based on the directional dispersion parameter, the stopband width of the band-stop filter is configured, and the transfer function of the band-stop filter for the building crack is obtained.

[0055] Specifically, a fast Fourier transform is performed on the fused base image to decompose it into multiple sine and cosine wave components of different frequencies.

[0056] Furthermore, by calculating the amplitude and phase of each pixel in different frequency components, the image is transformed from the spatial domain to the frequency domain, forming a matrix that can reflect the frequency characteristics of the image. Each matrix element corresponds to the signal strength of a specific frequency, thus obtaining the frequency domain coefficient matrix of the building crack.

[0057] Furthermore, based on the directional distribution characteristics of the initial crack profile features, a band-stop filter transfer function for building cracks is constructed. First, the orientation angle of each crack segment in the initial crack profile features is analyzed, and the distribution density and extension length of cracks in different directions are statistically analyzed.

[0058] Furthermore, the directional range corresponding to the noise frequency that needs to be suppressed is determined, and the stopband region of the band-stop filter is set according to this directional range. The filter attenuates the signal in this region while allowing the signal to pass through in other regions, forming a transfer function that can specifically filter noise in a particular direction, thus obtaining the band-stop filter transfer function for building cracks.

[0059] Furthermore, the frequency domain coefficient matrix is ​​multiplied by the band-stop filter transfer function. Each element in the frequency domain coefficient matrix is ​​multiplied by the corresponding element in the band-stop filter transfer function. This attenuates the frequency components in the band-stop region of the frequency domain coefficient matrix while preserving other effective frequency components. After all the elements are processed, the filtered frequency domain coefficients of the building cracks are obtained.

[0060] Specifically, the process involves identifying line segments with lengths greater than a threshold in the initial profile features of cracks. First, a fixed length threshold is set based on the actual needs of building crack detection. This threshold is used to distinguish between valid line segments and noise line segments. Then, all line segments in the initial profile features of cracks are extracted one by one.

[0061] Furthermore, the length of each line segment is precisely calculated using image measurement tools. The length of each line segment is compared with a set threshold, and all line segments whose length exceeds the threshold are retained. These line segments together constitute the line segment element.

[0062] Furthermore, a dispersion analysis is performed on the direction angles of the line segment elements. First, the extension direction of each line segment element is determined, and then the angle between each line segment and the horizontal reference line is measured using image analysis tools to obtain the direction angle of each line segment.

[0063] Furthermore, the direction angles of all line segment elements are sorted according to their size, the number distribution of line segments in different angle intervals is statistically analyzed, the total deviation between each direction angle and the average direction angle is calculated, and the degree of concentration of direction angles is judged by the magnitude of the total deviation. The smaller the total deviation, the more concentrated the direction is, and vice versa. Thus, the direction dispersion parameter reflecting this degree of concentration or dispersion is obtained.

[0064] Furthermore, the stopband width of the stopband filter is configured based on the directional dispersion parameter. When the directional dispersion parameter is small, that is, when the directional angles of the line segment elements are relatively concentrated, it indicates that the crack direction is relatively uniform. At this time, a narrower stopband width is set to suppress noise frequencies only near the concentrated direction.

[0065] Furthermore, when the directional dispersion parameter is large, that is, the directional angles of the line segment elements are relatively dispersed, it indicates that the crack direction is diverse. At this time, a wider stopband width is set to cover more noise frequencies in more directions. Through such configuration, the band-stop filter can accurately adapt to the directional distribution characteristics of the crack, and finally obtain the band-stop filter transfer function of the building crack.

[0066] Furthermore, the frequency domain coefficient matrix is ​​multiplied by the band-stop filter transfer function. Each element in the frequency domain coefficient matrix is ​​located one by one, and the function value corresponding to that element in the band-stop filter transfer function is found. The two values ​​are multiplied to obtain the filtered coefficient value at that position.

[0067] Furthermore, the same operation is performed on all elements of the frequency domain coefficient matrix, and the resulting matrix is ​​the filtered frequency domain coefficient of the building crack.

[0068] Specifically, the frequency domain transfer function value is obtained by processing signals of different spatial frequencies using a band-stop filter, and the spatial frequency variable is the frequency information contained in the frequency domain coefficient matrix obtained after the fused basis image undergoes a fast Fourier transform.

[0069] Furthermore, the square of the stopband center frequency reference value is determined based on the frequency of the main direction in the initial profile features of the crack. By analyzing the directional angle distribution of the line segment elements, the frequency corresponding to the direction with the highest frequency is found as the stopband center frequency reference value and its square is calculated.

[0070] Furthermore, the stopband width adjustment factor is determined by the directional dispersion parameter. A smaller directional dispersion parameter results in a smaller stopband width adjustment factor, while a larger directional dispersion parameter results in a larger stopband width adjustment factor. The filter order is set according to the accuracy requirements of enhancing the building crack features. Higher accuracy requirements result in a higher order, and vice versa.

[0071] Furthermore, this formula describes the attenuation degree of the band-stop filter for signals of different spatial frequencies. When the spatial frequency variable is close to the reference value of the stopband center frequency, the frequency domain transfer function value is small, indicating that the filter attenuates the signal at that frequency strongly and can effectively suppress noise near that frequency. When the spatial frequency variable is far from the reference value of the stopband center frequency, the frequency domain transfer function value is close to 1, indicating that the filter attenuates the signal at that frequency weakly and can retain the effective characteristic signal of the crack, thereby realizing frequency domain filtering of building cracks and enhancing crack characteristics.

[0072] Furthermore, as the spatial frequency variable gradually moves closer to the stopband center frequency reference value from a distance, the frequency domain transfer function value gradually decreases from close to 1, and the rate of decrease accelerates with the increase of the filter order. When the spatial frequency variable equals the stopband center frequency reference value, the frequency domain transfer function value reaches its minimum. As the spatial frequency variable gradually moves away from the stopband center frequency reference value, the frequency domain transfer function value gradually increases from its minimum value and approaches 1, and the rate of increase also accelerates with the increase of the filter order. At the same time, the larger the stopband width adjustment factor, the wider the frequency range in which the frequency domain transfer function value changes significantly, i.e., the wider the stopband width, and vice versa.

[0073] In summary, performing a fast Fourier transform on the fused base image to obtain the frequency domain coefficient matrix can transform the image from the spatial domain to the frequency domain, clearly separating signal components of different frequencies. This makes the crack features and noise exhibit more obvious differences in the frequency domain, providing convenience for subsequent targeted filtering and improving the accuracy of filtering processing.

[0074] In summary, constructing a band-stop filter transfer function based on the directional distribution characteristics of the initial crack profile allows the filter's stopband characteristics to be adapted to the crack's directional distribution, accurately locating and suppressing noise frequencies unrelated to the crack's orientation, while preserving the crack's own effective frequency components, ensuring that the filtering process can specifically enhance crack characteristics.

[0075] In summary, multiplying the frequency domain coefficient matrix with the transfer function of the band-stop filter to obtain the filtered frequency domain coefficients can effectively attenuate noise frequency components through frequency domain operations, highlighting the frequency characteristics of the crack. After subsequent inverse transformation, a clearer crack image can be obtained, significantly enhancing the identification of crack features and providing higher quality image data for subsequent crack analysis.

[0076] In summary, identifying line segments with lengths greater than a threshold in the initial contour features of cracks can filter out the main body of cracks that have practical significance, eliminate false line segments formed by short noise or noise interference, and ensure that subsequent directional analysis is only for valid crack line segments, thus laying a reliable foundation for constructing accurate filter transfer functions.

[0077] In summary, the dispersion parameter obtained by performing dispersion analysis on the direction angle of line segment elements can quantitatively reflect the degree of concentration or dispersion of crack direction, clearly present the overall direction distribution characteristics of cracks, and enable the filter design to closely follow the actual direction law of cracks, avoiding filtering deviations caused by ambiguity in direction judgment.

[0078] In summary, the transfer function of the band-stop filter is obtained by configuring the stopband width based on the directional dispersion parameter. This allows the stopband width to match the dispersion of the crack orientation. When the crack orientation is concentrated, a narrow stopband is used to accurately suppress noise in a specific direction. When the orientation is dispersed, a wide stopband is used to fully cover the noise frequency, ensuring that the filter can efficiently filter interference and completely preserve crack characteristics, thereby improving the adaptability and effectiveness of the filtering process.

[0079] S4. Based on the material brittle fracture parameters, perform multi-physics probability fusion on the visible light texture gradient field, infrared temperature gradient field and depth curvature gradient field in the feature enhancement image to obtain the crack probability distribution map of the building crack; In this embodiment of the invention, the multi-physics probability fusion of the visible light texture gradient field, infrared temperature gradient field, and depth curvature gradient field in the feature enhancement image based on the material brittle fracture parameters to obtain the crack probability distribution map of the building cracks includes: The visible light texture gradient field is normalized by structural tensor to obtain the first field strength distribution of the building cracks. The thermal diffusion equation is inverted on the infrared temperature gradient field to obtain the second field strength distribution of the building cracks; Based on the curvature tensor eigenvalues ​​derived from the deep point cloud data, the third field strength distribution of the building cracks is obtained. The crack probability value of the building crack is obtained by weighted fusion of the first field strength distribution, the second field strength distribution and the third field strength distribution based on energy density. The crack probability value is mapped to the corresponding spatial location of the feature-enhanced image to obtain the crack probability distribution map of the building crack.

[0080] Specifically, the visible light texture gradient field is normalized using a structure tensor. First, the structure tensor of the visible light texture gradient field in the feature-enhanced image is extracted. This tensor contains information about the direction and intensity of the texture.

[0081] Furthermore, by calculating the modulus of the structural tensor, all tensor elements are divided by this modulus, so that the processed tensor element values ​​are between 0 and 1, eliminating the influence of texture intensity differences in different regions, and obtaining the first field intensity distribution of building cracks.

[0082] Furthermore, the thermal diffusion equation is inverted for the infrared temperature gradient field. First, the temperature distribution data of the infrared temperature gradient field in the feature-enhanced image is obtained. Based on the thermal conductivity of the material, the thermal diffusion process is deduced in reverse, and the temperature gradient is converted into field strength data that reflects the intensity of heat transfer. This ensures that the data can reflect the temperature difference characteristics between the crack area and the surrounding area, and the second field strength distribution of the building crack is obtained.

[0083] Furthermore, based on the depth point cloud data, the curvature tensor feature values ​​are derived. First, the depth point cloud data of the building surface corresponding to the feature-enhanced image is collected. These data record the three-dimensional coordinates of each point on the surface. By calculating the spatial relationship between adjacent points, the curvature tensor is constructed. Then, feature values ​​that can reflect the curvature of the surface are extracted from the curvature tensor. These are used as field strength data to obtain the third field strength distribution of the building cracks.

[0084] Furthermore, the first, second, and third field strength distributions are weighted and fused based on energy density. First, the energy density of the first, second, and third field strength distributions at each spatial location is calculated, and the field strength distribution with higher energy density corresponds to a higher weight.

[0085] Furthermore, the field strength values ​​of the three field strengths distributed at the same spatial location are superimposed according to their respective weights to obtain the comprehensive field strength value of each spatial location, which is the crack probability value of the building crack.

[0086] Furthermore, the crack probability values ​​are mapped to the corresponding spatial locations in the feature-enhanced image. First, the spatial coordinates of each pixel in the feature-enhanced image are determined. Then, the calculated crack probability values ​​are mapped one-to-one to the pixels in the image according to the spatial coordinates, so that each pixel has a corresponding crack probability value. Through this spatial correspondence, a complete image is formed, and the crack probability distribution map of the building cracks is obtained.

[0087] In summary, normalizing the visible light texture gradient field using the structural tensor to obtain the first field strength distribution can standardize the intensity and direction information of the texture gradient field, eliminate the interference caused by the difference in texture intensity in different regions, and enable the texture features to be reflected at a unified scale. This accurately reflects the texture change characteristics of cracks under visible light and provides a stable visual feature basis for multi-physics field fusion.

[0088] In summary, the second field strength distribution is obtained by inverting the thermal diffusion equation of the infrared temperature gradient field. The heat transfer process can be derived in reverse, and the temperature gradient can be converted into field strength data that reflects the heat transfer characteristics of the crack region. This effectively captures the temperature anomalies caused by structural discontinuities in cracks, adds a thermal feature dimension to crack identification, and improves the ability to identify hidden cracks.

[0089] In summary, the third field strong distribution is derived from the curvature tensor feature value based on deep point cloud data. This allows for the extraction of surface bending characteristics of the crack region from a three-dimensional spatial perspective. By reflecting the depth and morphological characteristics of the crack through curvature changes, it supplements the information on the spatial morphology of the crack, making the crack feature description more comprehensive and enhancing the understanding of the three-dimensional morphology of the crack.

[0090] In summary, the probability value of cracks is obtained by weighted fusion of the three field strength distributions based on energy density. The weights can be determined according to the energy density of each field strength distribution in different regions, with higher energy density corresponding to higher weights. This allows the fusion results to highlight the regions with obvious crack characteristics in each physical field, improve the accuracy of crack judgment by integrating multi-dimensional information, and reduce the limitations of single physical field analysis.

[0091] In summary, mapping crack probability values ​​to their corresponding spatial locations in feature-enhanced images to obtain crack probability distribution maps can transform abstract probability values ​​into intuitive image forms, clearly showing the likelihood that each location belongs to a crack. This provides a visualized and accurate probability distribution basis for subsequent crack topology optimization and risk assessment, improving the intuitiveness and reliability of crack analysis.

[0092] S5. The topological connectivity of the crack probability distribution map is optimized through multi-scale morphological iteration to obtain a continuous crack mask for the building cracks; In this embodiment of the invention, the step of obtaining a continuous crack mask for the building cracks by iteratively optimizing the topological connectivity of the crack probability distribution map through multi-scale morphological methods includes: Spatial connectivity analysis is performed on the fractured regions in the crack probability distribution map to obtain the primary connectivity mask of the building cracks; Based on the regional continuity of the residual fracture gap in the primary connectivity mask, the boundary of the unconnected region of the building crack is determined; The broken sections in the boundary of the unconnected area are filled and connected to form a continuous crack mask for the building cracks.

[0093] Specifically, spatial connectivity analysis is performed on the fracture region in the crack probability distribution map. First, all pixels in the crack probability distribution map whose probability values ​​match the crack characteristics are traversed.

[0094] Furthermore, by determining whether these pixels are adjacent in spatial location, the interconnected pixels are divided into the same connected region, and the broken parts between each connected region are marked to form a mask that can initially reflect the connection state of the crack, thus obtaining the primary connected mask of the building crack.

[0095] Furthermore, based on the regional continuity of the residual fracture gaps in the primary connected mask, the boundary of the unconnected area of ​​the building crack is determined. First, observe the morphology of each residual fracture gap in the primary connected mask, and analyze whether the direction, width and probability value distribution of the areas on both sides of the gap have continuity.

[0096] Furthermore, if the two sides of the area are consistent in direction and width and have similar probability value characteristics, then the gap is determined to be part of an unconnected area. Then, a clear boundary line is drawn along the edge pixels of the gap to determine the boundary of the unconnected area of ​​the building crack.

[0097] Furthermore, the fracture points in the boundary of the unconnected region are filled and connected to form a continuous crack mask for building cracks. First, the start and end points of the fracture points in the boundary of the unconnected region are determined. Based on the direction and width of the cracks on both sides of the fracture point, transition pixels matching the features of the cracks on both sides are generated in the fracture area.

[0098] Furthermore, the probability values ​​of these transition pixels are set to be similar to those of the surrounding crack pixels. By filling these transition pixels, the two ends of the fracture are connected, so that the entire crack area forms a complete connected shape, thus obtaining a continuous crack mask for the building crack.

[0099] In summary, spatial connectivity analysis of fracture regions in the crack probability distribution map yields a primary connectivity mask, which can preliminarily identify interconnected regions in the crack probability distribution map, clarify the basic connectivity state of cracks, provide a basic framework for subsequent optimization of topological connectivity, and reduce the blindness of subsequent processing.

[0100] In summary, determining the boundaries of unconnected areas based on the regional continuity of residual fracture gaps in the primary connectivity mask can accurately locate fractures caused by probability value fluctuations or noise interference, clarify the boundary range that needs to be filled, ensure that subsequent filling work is targeted, and avoid mishandling of non-crack areas.

[0101] In summary, filling and connecting the fractures at the boundaries of unconnected regions to form a continuous crack mask can eliminate the fracture phenomenon in the crack probability distribution map, form a complete and coherent crack morphology, and truly restore the actual direction and connectivity of the cracks. This provides accurate morphological basis for subsequent risk assessment based on crack morphology, and improves the completeness and reliability of crack analysis.

[0102] S6. Based on the stress distribution characteristics of the building structure, the spatial orientation of the continuous crack mask is analyzed to obtain the crack risk level of the building crack.

[0103] In this embodiment of the invention, the step of analyzing the spatial orientation of the continuous crack mask based on the stress distribution characteristics of the building structure to obtain the crack risk level of the building crack includes: A set of rules for mapping crack orientation to stress direction is generated based on a library of stress distribution characteristics of building structures. The spatial orientation of the continuous crack mask is decomposed into a sequence of independent directional components; The crack risk level of the building crack is obtained by matching the coupling strength between the independent directional component sequence and the mapping rule set.

[0104] Specifically, a set of rules for mapping crack direction to stress direction is generated based on the stress distribution characteristic library of building structures. First, stress direction data of different types of building structures under various stress states and historical records of crack direction in the corresponding structures are collected from the stress distribution characteristic library of building structures.

[0105] Furthermore, by analyzing this data, we can clarify the risk level of crack development when the crack direction and stress direction present different angles. We can then organize these correspondences into specific rules, each of which includes a specific crack direction range, the corresponding stress direction range, and a corresponding risk association description, thus forming a set of rules for mapping the crack direction and stress direction of building cracks.

[0106] Furthermore, the spatial orientation of the continuous crack mask is decomposed into a sequence of independent directional components. First, along the orientation of the continuous crack mask, starting from the starting point of the crack, feature points are selected at fixed intervals, and the tangent direction of the crack at each feature point is recorded. These tangent directions are the orientation of the crack at that location.

[0107] Furthermore, these continuous directional data are divided according to the nodes where the direction changes. The direction within each segment remains consistent. The direction of each segment is treated as an independent directional component. These components are arranged in sequence to obtain the independent directional component sequence of building cracks.

[0108] Furthermore, the coupling strength of the independent directional component sequences is matched with the mapping rule set, and each directional component in the independent directional component sequence is extracted one by one.

[0109] Furthermore, each directional component is compared with each rule in the mapping rule set to determine whether the directional component meets the requirements of the crack orientation range in the rule, and the corresponding stress direction range is determined according to the risk association description set in the rule.

[0110] Furthermore, the coupling strength between the directional component and the corresponding stress direction is calculated. The higher the coupling strength, the greater the risk. By combining the coupling strength of all independent directional components, the overall risk level is determined according to the preset judgment criteria, and the crack risk level of the building crack is obtained.

[0111] In summary, generating a crack orientation-stress direction mapping rule set based on a building structure stress distribution characteristic library can transform professional knowledge of building structure stress distribution into directly applicable matching rules, providing a scientific and systematic basis for determining crack risk levels and ensuring that risk assessment has clear standards to follow.

[0112] In summary, decomposing the spatial orientation of a continuous crack mask into a sequence of independent directional components can break down the complex crack orientation into simple directional segments. The orientation characteristics of each segment are clearer, making it easier to accurately compare with the rules in the mapping rule set, avoiding matching deviations caused by the complexity of the overall orientation, and improving the detail of the analysis.

[0113] In summary, the crack risk level is obtained by coupling the independent directional component sequence with the mapping rule set. By analyzing the correlation between the crack direction and the stress direction, the development trend and potential risks of cracks under stress can be scientifically judged, making the risk level determination more consistent with the actual stress situation of the building structure and providing reliable decision support for the subsequent treatment of cracks.

[0114] like Figure 2 The diagram shown is a functional module diagram of a building crack intelligent identification system based on multi-source data fusion provided in an embodiment of the present invention.

[0115] The intelligent building crack identification system 100 based on multi-source data fusion described in this invention can be installed in an electronic device. Depending on the functions implemented, the intelligent building crack identification system 100 based on multi-source data fusion may include a contour feature aggregation module 101, a thermal deformation compensation registration module 102, an adaptive frequency domain filtering module 103, a multi-physics probability fusion module 104, a topology connectivity optimization module 105, and a structural risk analysis module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0116] In this embodiment, the functions of each module / unit are as follows: The contour feature aggregation module 101 is used to aggregate the initial contour features of cracks extracted from the surface optical enhancement feature map into crack spatial distribution features. The thermal deformation compensation registration module 102 is used to compensate and register the deformation of the spatial distribution characteristics of the crack based on the thermal expansion characteristics of the material, so as to obtain the fused base image of the building crack. The adaptive frequency domain filtering module 103 is used to perform adaptive frequency domain filtering on the fused base image to obtain a feature-enhanced image of the building cracks. The multi-physics probability fusion module 104 is used to perform multi-physics probability fusion on the visible light texture gradient field, infrared temperature gradient field and depth curvature gradient field in the feature enhancement image based on the material brittle fracture parameters, so as to obtain the crack probability distribution map of the building crack. The topology connectivity optimization module 105 is used to optimize the topology connectivity of the crack probability distribution map through multi-scale morphological iteration to obtain a continuous crack mask for the building cracks. The structural risk analysis module 106 is used to analyze the spatial orientation of the continuous crack mask based on the stress distribution characteristics of the building structure, and to obtain the crack risk level of the building crack.

[0117] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0118] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0119] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0120] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0121] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent identification of building cracks based on multi-source data fusion, characterized in that, The method includes: S1. The initial contour features of the cracks extracted from the surface optical enhancement feature map are aggregated into crack spatial distribution features; S2. Based on the thermal expansion characteristics of the material, the deformation of the spatial distribution features of the cracks is compensated and registered to obtain the fused base image of the building cracks; S3. Perform adaptive frequency domain filtering on the fused base image to obtain a feature-enhanced image of the building cracks; S4. Based on the material brittle fracture parameters, perform multi-physics probability fusion on the visible light texture gradient field, infrared temperature gradient field, and depth curvature gradient field in the feature enhancement image to obtain the crack probability distribution map of the building crack; S5. The topological connectivity of the crack probability distribution map is optimized through multi-scale morphological iteration to obtain a continuous crack mask for the building cracks; S6. Based on the stress distribution characteristics of the building structure, the spatial orientation of the continuous crack mask is analyzed to obtain the crack risk level of the building crack.

2. The intelligent building crack identification method based on multi-source data fusion as described in claim 1, characterized in that, The step of summarizing the initial contour features of the cracks extracted from the surface optical enhancement feature map into crack spatial distribution features includes: Edge detection is performed on the surface optical enhancement feature map to obtain the initial edge feature map of the building crack; By connecting the fracture regions in the initial edge feature map based on pixel connectivity constraints, a continuous crack skeleton map of the building crack is obtained. The spatial coordinates and direction vectors of the continuous crack skeleton map are extracted to obtain the spatial distribution characteristics of the building cracks.

3. The intelligent building crack identification method based on multi-source data fusion as described in claim 1, characterized in that, The deformation compensation registration based on the thermal expansion characteristics of the material to obtain the fused base image of the building cracks includes: Obtain the deformation compensation parameters corresponding to the current ambient temperature from the material thermal expansion property database; The deformation compensation parameters are inversely mapped to the deformation of the spatial distribution characteristics of the cracks to obtain the compensation deformation feature set of the building cracks; The compensated deformation feature set is superimposed onto the original crack image to obtain the fused base image of the building crack.

4. The intelligent building crack identification method based on multi-source data fusion as described in claim 3, characterized in that, The step of superimposing the compensated deformation feature set onto the original crack image to obtain the fused base image of the building crack includes: The original crack image is calibrated by using the geometric space coordinates of the compensation deformation feature set to obtain the calibrated image of the original crack image. The calibration image is fused with the multidimensional physical features of the compensation deformation feature set through channel fusion to obtain the fused feature tensor of the calibration image; Spatial domain reconstruction is performed on the fused feature tensor to obtain the fused base image of the building crack.

5. The intelligent building crack identification method based on multi-source data fusion as described in claim 1, characterized in that, The step of performing adaptive frequency domain filtering on the fused base image to obtain a feature-enhanced image of the building cracks includes: The frequency domain coefficient matrix of the building cracks is obtained by performing a fast Fourier transform on the fused base image. Based on the directional distribution characteristics of the initial contour features of the crack, a band-stop filter transfer function for the building crack is constructed, wherein the formula for the band-stop filter transfer function is as follows: ， In the formula: Transfer function values ​​to the frequency domain. For spatial frequency variables, It is the square of the stopband center frequency reference value. This is the stopband width adjustment factor. The filter order; The frequency domain coefficient matrix is ​​multiplied by the band-stop filter transfer function to obtain the filtered frequency domain coefficients of the building cracks.

6. The intelligent building crack identification method based on multi-source data fusion as described in claim 5, characterized in that, The construction of the band-stop filter transfer function for the building crack based on the directional distribution characteristics of the initial contour features of the crack includes: Identify line segment features with a length greater than a threshold in the initial contour features of the crack; The dispersion analysis of the direction angle of the line segment element is performed to obtain the direction dispersion parameter of the line segment element; Based on the directional dispersion parameter, the stopband width of the band-stop filter is configured, and the transfer function of the band-stop filter for the building crack is obtained.

7. The intelligent building crack identification method based on multi-source data fusion as described in claim 1, characterized in that, The method involves multi-physics probability fusion of the visible light texture gradient field, infrared temperature gradient field, and depth curvature gradient field in the feature-enhanced image based on material brittle fracture parameters to obtain a crack probability distribution map of the building cracks, including: The visible light texture gradient field is normalized by structural tensor to obtain the first field strength distribution of the building cracks. The thermal diffusion equation is inverted on the infrared temperature gradient field to obtain the second field strength distribution of the building cracks; Based on the curvature tensor eigenvalues ​​derived from the deep point cloud data, the third field strength distribution of the building cracks is obtained. The crack probability value of the building crack is obtained by weighted fusion of the first field strength distribution, the second field strength distribution and the third field strength distribution based on energy density. The crack probability value is mapped to the corresponding spatial location of the feature-enhanced image to obtain the crack probability distribution map of the building crack.

8. The intelligent building crack identification method based on multi-source data fusion as described in claim 1, characterized in that, The step of optimizing the topological connectivity of the crack probability distribution map through multi-scale morphological iteration to obtain a continuous crack mask for the building cracks includes: Spatial connectivity analysis is performed on the fractured regions in the crack probability distribution map to obtain the primary connectivity mask of the building cracks; Based on the regional continuity of the residual fracture gap in the primary connectivity mask, the boundary of the unconnected region of the building crack is determined; The broken sections in the boundary of the unconnected area are filled and connected to form a continuous crack mask for the building cracks.

9. The intelligent building crack identification method based on multi-source data fusion as described in claim 1, characterized in that, The spatial orientation of the continuous crack mask is analyzed based on the stress distribution characteristics of the building structure to obtain the crack risk level of the building crack, including: A set of rules for mapping crack orientation to stress direction is generated based on a library of stress distribution characteristics of building structures. The spatial orientation of the continuous crack mask is decomposed into a sequence of independent directional components; The crack risk level of the building crack is obtained by matching the coupling strength between the independent directional component sequence and the mapping rule set.

10. A building crack intelligent identification system based on multi-source data fusion, characterized in that, The system includes: The contour feature aggregation module is used to aggregate the initial contour features of cracks extracted from the surface optical enhancement feature map into crack spatial distribution features. A thermal deformation compensation registration module is used to compensate and register the deformation of the spatial distribution characteristics of the cracks based on the thermal expansion characteristics of the material, so as to obtain a fused base image of the building cracks. An adaptive frequency domain filtering module is used to perform adaptive frequency domain filtering on the fused base image to obtain a feature-enhanced image of the building cracks. The multi-physics probability fusion module is used to perform multi-physics probability fusion on the visible light texture gradient field, infrared temperature gradient field and depth curvature gradient field in the feature enhancement image based on the material brittle fracture parameters, so as to obtain the crack probability distribution map of the building cracks. The topology connectivity optimization module is used to optimize the topology connectivity of the crack probability distribution map through multi-scale morphological iteration to obtain a continuous crack mask of the building cracks. The structural risk analysis module is used to analyze the spatial orientation of the continuous crack mask based on the stress distribution characteristics of the building structure, and to obtain the crack risk level of the building crack.

Images