A multispectral imaging detection method and system for leakage disease of external wall insulation layer

Abstract

This invention belongs to the field of wall leakage detection technology. It discloses a multispectral imaging detection method and system for external wall insulation layer leakage, comprising: simultaneously acquiring multi-band images of the external wall facade using a multispectral imaging device mounted on a UAV, and simultaneously recording environmental perception parameters and the current ground sampling distance; the multi-band images include a first-band image, a second-band image, and a third-band image; performing standardization processing on the multi-band images according to the environmental perception parameters, and constructing a water penetration trajectory inside the wall based on the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction in the second-band image; calculating the interlayer attenuation difference and water content progression consistency value of each pixel at different penetration depths based on the water penetration trajectory, and distinguishing interference areas, normal areas, and leakage areas by combining with a preset leakage spectral feature library; thus improving the identification accuracy of external wall insulation layer leakage.

Description

Technical Field

[0001] This invention relates to the field of wall leakage detection technology, and more specifically, to a multispectral imaging detection method and system for leakage defects in external wall insulation layers. Background Technology

[0002] Existing methods for detecting external wall leaks include manual inspection, infrared thermal imaging, and multispectral imaging. Manual inspection relies on personnel experience, is inefficient, and struggles to detect hidden leaks. Infrared thermal imaging can reflect localized water content, but it is susceptible to interference from various factors and has limited ability to identify deep leaks and complex wall structures. In recent years, multispectral and UAV remote sensing technologies have been applied to external wall leak detection due to their advantages of non-contact, wide-area, and rapid detection.

[0003] However, existing multispectral or infrared exterior wall detection technologies only identify leaks based on fixed flight altitude or empirical resolution, lacking a quantitative description of actual detection capabilities. This makes it impossible to determine the smallest leak scale that can be reliably identified, and easily leads to problems such as leak signals being submerged, small-scale defects being missed, and insufficient stability of detection results.

[0004] Existing technologies mostly employ fixed flight parameters and uniform detection modes, lacking dynamic detection strategies based on actual imaging capabilities. When spatial resolution is insufficient, the system struggles to identify detection capability boundaries. Traditional supplementary measurements rely on manual experience and lack adaptive mechanisms, easily leading to large supplementary measurement ranges, low efficiency, or omissions of critical areas. Furthermore, while terahertz and other deep-layer detection methods possess strong penetrating power and moisture sensitivity, they employ full-area scanning or independent detection, resulting in long detection times and high resource consumption. The lack of collaborative fusion mechanisms for multi-source detection results makes it difficult to balance detection accuracy, deep leakage identification capabilities, and detection efficiency.

[0005] In view of this, the present invention proposes a multispectral imaging detection method for leakage defects in external wall insulation layers to solve the above problems. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a multispectral imaging detection method for leakage defects in external wall insulation layers, comprising:

[0007] S1. Simultaneously acquire multi-band images of the exterior facade using a multispectral imaging device mounted on a drone, and simultaneously record environmental perception parameters and the current ground sampling distance; the multi-band images include a first-band image, a second-band image, and a third-band image;

[0008] S2. Perform standardization processing on multi-band images based on environmental perception parameters, and construct the water penetration trajectory inside the wall based on the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction in the second band image.

[0009] S3. Based on the water penetration trajectory, calculate the interlayer attenuation difference and water content progression consistency value of each pixel at different penetration depths, and combine the preset leakage spectral feature library to distinguish the interference area, normal area and leakage area, and generate the corresponding uncertainty index and multispectral detection results.

[0010] S4. Calculate the minimum detectable leakage area based on the current ground sampling distance. When the minimum detectable leakage area exceeds the preset detection requirements, automatically generate a supplementary detection area.

[0011] S5. The spatial location in the supplementary detection area that satisfies the uncertainty index being greater than the preset uncertainty index threshold is determined as the target area for triggering terahertz supplementary detection, and terahertz imaging detection is performed on the target area. For non-supplementary detection areas, the multispectral detection result is directly used as the final detection result, and a corresponding structured detection report is generated.

[0012] Preferably, the method for acquiring multi-band images of the exterior wall facade includes:

[0013] The drone carries a multispectral imaging device to perform inspection flights along a preset route on the exterior wall of the building. During the flight, the optical axis of the multispectral imaging device is kept facing the exterior wall facade so as to perform continuous multi-band synchronous imaging acquisition of the area to be inspected.

[0014] The multispectral imaging device includes several band imaging units covering the visible light band, near-infrared band, and short-wave infrared band. Each band imaging unit synchronously exposes the outer wall area at the same time to obtain multi-band image data of the corresponding area to be detected. The multi-band image includes a first band image, a second band image, and a third band image.

[0015] During the data acquisition process, environmental sensing parameters of the area to be detected are acquired simultaneously. These parameters include ambient temperature, ambient humidity, solar irradiance, shooting distance, and shooting angle. The current ground sampling distance of the corresponding area to be detected is calculated by combining the UAV's flight altitude, the lens focal length of the multispectral imaging device, and the image resolution.

[0016] Preferably, the method for performing normalization processing on multi-band images includes:

[0017] Read the environmental perception parameters recorded synchronously with the multi-band images, and correlate the environmental perception parameters with the corresponding multi-band images in time; normalize and correct the brightness response of each band image according to the solar irradiance, and perform environmental compensation according to the spectral response of each band image according to the ambient temperature and humidity.

[0018] Perform environmental compensation: A table of environmental compensation parameters is established in advance based on standard wall samples collected under different temperature and humidity conditions to obtain the correspondence between environmental temperature and humidity and the response offset of each band; based on the real-time acquired environmental temperature and humidity parameters, the corresponding compensation coefficients are determined in the environmental compensation parameter table, and the offset of the reflection response of each band image is corrected using the compensation coefficients.

[0019] Based on the shooting distance and shooting angle, establish the corresponding imaging geometric transformation relationship, map the images of each band to the preset standard observation plane, perform perspective correction on images with tilted shooting, and resample the spatial scale differences caused by changes in shooting distance.

[0020] Using a preset reference area as a benchmark, the response ranges of the first-band image, the second-band image, and the third-band image are uniformly mapped to form a standardized multi-band image with a unified radiation response benchmark and spatial scale.

[0021] The preset reference area is obtained by pre-collecting standard wall samples under different environmental conditions. The multi-band response values ​​of each spatial area in the standard wall samples under different environmental temperatures and humidity are statistically analyzed, and the response fluctuation amplitude of each spatial area in several acquisition processes is calculated.

[0022] When the response fluctuation amplitude of the corresponding spatial region is less than the preset response fluctuation amplitude threshold, the corresponding spatial region is determined as the preset reference region; the response fluctuation amplitude is obtained by calculating the standard deviation of the multi-band response value of the corresponding spatial region relative to the average multi-band response value during several acquisitions.

[0023] Preferably, the method for constructing the water penetration trajectory inside the wall includes:

[0024] After completing the multi-band image standardization process, the target wall region in the second-band image is extracted, and the pixel response in the second-band image is analyzed in layers along the wall thickness direction to obtain the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction.

[0025] Based on the response change gradient and continuous change characteristics, interlayer progressive response chains with continuous change patterns are identified. When the gradient change direction between adjacent regions is consistent and the gradient difference is within a preset gradient difference threshold, they are identified as the same interlayer progressive response chain. The interlayer progressive response chains are then spatially connected to form candidate water penetration paths.

[0026] Connectivity and stability analysis is performed on candidate water penetration paths to eliminate abnormal discrete responses and isolated regions, retain paths with continuous response changes, and sort and connect the retained paths according to the direction of response change to form the water penetration trajectory inside the wall that characterizes the direction of water migration and penetration depth.

[0027] Connectivity analysis is performed on candidate water penetration paths: the length of the shared boundary and the number of consecutive connected pixels between adjacent response regions are statistically analyzed, and Min-Max normalization and weighted fusion are performed to construct a connectivity strength index between regions; when the connectivity strength index is higher than the preset connectivity strength index threshold, it is determined that there is a stable spatial connection between the corresponding response regions; otherwise, it is regarded as a weak connection or non-connected region.

[0028] Stability analysis is performed on candidate moisture penetration paths: the sequence of response value changes is statistically analyzed along the extension direction of each path. When the response value in the extension direction of the path shows a continuous increase or decrease, and the change in response value between adjacent response areas is less than the preset threshold for the change in response value, the path is determined to have stable response change characteristics. The change in response value between adjacent response areas is obtained by calculating the absolute value of the difference in response value between the path nodes corresponding to the adjacent response areas.

[0029] Preferably, the method for obtaining the interlayer attenuation difference and the consistency value of water content progression includes:

[0030] Using the extension direction of the water penetration trajectory as the direction of water propagation, and the starting position of the water penetration trajectory as the depth reference point, combined with the current ground sampling distance of the corresponding area to be detected, the depth position of the pixels in the area covered by the water penetration trajectory is mapped to obtain the relative penetration depth of each pixel.

[0031] Pixel response sequences are established according to relative penetration depth, and the response values ​​of each pixel in the second band image are extracted; the response change is calculated based on the response values ​​corresponding to adjacent relative penetration depths, and the interlayer attenuation difference of each pixel at different penetration depths is obtained.

[0032] Using a preset depth range as the analysis window, the direction and fluctuation of the interlayer attenuation difference within the window are statistically analyzed, and the water cut progression consistency value is calculated based on the direction consistency and fluctuation of the interlayer attenuation difference in adjacent depth ranges.

[0033] Consistent values ​​for progressive water content: ;in, This indicates a consistent value for progressive water content; Indicates the directional consistency coefficient; Represents the normalized volatility coefficient; and This indicates the preset weighting coefficients, and and The sum is 1.

[0034] Preferably, the method for generating the corresponding uncertainty index and multispectral detection results includes:

[0035] The system calls a preset leakage spectral feature library, which stores the multispectral response features corresponding to the interference area, normal area, and leakage area. The multispectral response features include features of the interference area, normal area, and leakage area. Taking the area to be detected as the analysis object, the system extracts the multi-band response features of the area to be detected in the first band image, the second band image, and the third band image.

[0036] By combining the interlayer attenuation difference and water content progression consistency values ​​of the corresponding regions, a comprehensive spectral feature of the region to be detected is constructed; the degree of matching between the comprehensive spectral feature of the region to be detected and the feature of each category in the multispectral response feature is calculated, and the region category corresponding to the region to be detected is determined according to the matching results; the degree of matching between the feature of each category is obtained by Euclidean distance, correlation coefficient or cosine similarity.

[0037] Specifically, when the area to be detected matches the leakage area with the highest degree of feature matching and the corresponding water content progressive consistency value reaches the preset consistency value threshold, it is determined to be a leakage area; when the area to be detected matches the normal area with the highest degree of feature matching, it is determined to be a normal area; when the area to be detected matches the interference area with the highest degree of feature matching or the leakage area with the highest degree of feature matching, but the corresponding water content progressive consistency value does not reach the preset consistency value threshold, it is determined to be an interference area.

[0038] The uncertainty index is calculated based on the degree of matching between the comprehensive spectral characteristics of the area to be detected and the characteristics of each category. The area category, uncertainty index and corresponding spatial location information are then associated to generate multispectral detection results.

[0039] Preferably, the method for calculating the minimum detectable leakage area includes:

[0040] Obtain the current ground sampling distance corresponding to the current detection area and the noise equivalent reflectance difference of the multispectral imaging device under multi-band images; determine the standard leakage reflectance response of a single pixel based on the preset leakage spectral feature library.

[0041] The standard leakage reflectance response is obtained by statistically analyzing the average response difference between normal area samples and leakage area samples of the same material in multi-band images, and using a single pixel as the corresponding ground area as the statistical unit. Based on the noise equivalent reflectance difference, the current ground sampling distance, and the standard leakage reflectance response, the minimum detectable leakage area of ​​the multi-band image is calculated.

[0042] Preferably, the method for automatically generating supplementary detection areas includes:

[0043] The minimum detectable leakage area of ​​the multi-band image is compared with the preset minimum detectable leakage area threshold. When the minimum detectable leakage area is greater than the preset minimum detectable leakage area threshold, it is determined that the current detection resolution cannot meet the preset detection requirements, and the generation of supplementary detection areas is triggered.

[0044] The target ground sampling distance for supplementary testing is calculated based on the current ground sampling distance, the minimum detectable leakage area, and the preset minimum detectable leakage area threshold. After obtaining the target ground sampling distance, the candidate areas for supplementary testing are screened based on the multispectral detection results and the corresponding uncertainty index.

[0045] Candidate regions for supplementary testing include regions where the uncertainty index is greater than the preset uncertainty index threshold, regions where the water penetration trajectory extends along the direction of water migration and intersects with the boundary of the identified leakage region, and regions where the area is smaller than the minimum detectable leakage area and where there are abnormal multi-band response characteristics.

[0046] Based on the target ground sampling distance, the corresponding UAV flight altitude and the lens focal length of the multispectral imaging device are calculated. The supplementary acquisition range of the candidate supplementary measurement area that meets the preset target resolution requirements is determined in the area to be detected, and the corresponding supplementary detection area is generated. The area in the area to be detected other than the supplementary detection area is defined as the non-supplementary detection area.

[0047] Preferably, the method for generating the corresponding structured detection report includes:

[0048] After the supplementary detection area is generated, the spatial location in the supplementary detection area that satisfies the uncertainty index being greater than the preset uncertainty index threshold is determined as the target area for triggering terahertz supplementary detection, and terahertz imaging detection is performed on the target area to obtain the corresponding terahertz detection results.

[0049] For non-supplementary detection areas, the corresponding multispectral detection results are directly used as the final detection results; for target areas that trigger terahertz supplementary detection, the terahertz detection results and multispectral detection results are fused to generate the final detection results. Based on the final detection results and the features of each category in the multispectral response characteristics, combined with spatial location information, a structured detection report is generated.

[0050] A multispectral imaging detection system for leakage defects in external wall insulation layers includes:

[0051] The band acquisition module is used to simultaneously acquire multi-band images of the exterior facade using a multispectral imaging device mounted on a UAV, and simultaneously record environmental perception parameters and the current ground sampling distance; the multi-band images include a first band image, a second band image, and a third band image;

[0052] The standardization analysis module is used to perform standardization processing on multi-band images based on environmental perception parameters, and to construct the water penetration trajectory inside the wall based on the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction in the second band image.

[0053] The leakage identification module is used to calculate the interlayer attenuation difference and water content progression consistency value of each pixel at different penetration depths based on the water penetration trajectory, and to distinguish the interference area, normal area and leakage area by combining the preset leakage spectral feature library, and generate the corresponding uncertainty index and multispectral detection results.

[0054] The sampling correction module is used to calculate the minimum detectable leakage area based on the current ground sampling distance. When the minimum detectable leakage area exceeds the preset detection requirements, it automatically generates a supplementary detection area.

[0055] The report generation module is used to identify the spatial locations in the supplementary detection area that meet the condition that the uncertainty index is greater than the preset uncertainty index threshold as the target areas for triggering terahertz supplementary detection, and to perform terahertz imaging detection on the target areas. For non-supplementary detection areas, the multispectral detection results are directly used as the final detection results, and a corresponding structured detection report is generated.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] This invention, through adaptive quantitative evaluation of multispectral detection capabilities, clarifies the minimum leak scale that can be reliably identified under current detection conditions, thereby improving the interpretability and controllability of detection. By incorporating the noise equivalent reflectance difference and ground sampling distance into the minimum detectable leak area model, it synergistically constrains sensor sensitivity and spatial resolution, avoiding false alarms and missed detections, and improving the stability and reliability of detection results.

[0058] A standard leakage reflectance response is established using a pre-defined leakage spectral feature library, allowing the detection threshold to be based on statistical differences in real samples, thus improving the ability to identify small-scale early-stage leakage defects. A linkage mechanism enables closed-loop adjustment of detection capabilities and sampling parameters, adaptively adjusting spatial resolution to avoid missed detections of small-scale leaks and improve detection accuracy and adaptability. Uncertainty index and other parameters are used as criteria for selecting candidate regions for supplementary testing, reducing the problem of blindly expanding the scanning range and improving the targeting and resource utilization efficiency of supplementary testing. A hierarchical collaborative detection mechanism is constructed, triggering terahertz detection only in high-uncertainty regions, reducing detection time and resource consumption, and improving overall detection efficiency. The fusion of terahertz and multispectral detection results dynamically adjusts the contribution of the two types of detection information, enhancing detection reliability and reducing the risk of false positives and false negatives. Attached Figure Description

[0059] Figure 1This is a schematic diagram of a multispectral imaging detection method for leakage defects in external wall insulation layers according to the present invention.

[0060] Figure 2 This is a schematic diagram of the structure of a multispectral imaging detection system for leakage defects in external wall insulation layers according to the present invention. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] Example 1, please refer to Figure 1 As shown, this embodiment provides a multispectral imaging detection method for leakage defects in external wall insulation layers, specifically including the following steps:

[0063] S1. Simultaneously acquire multi-band images of the exterior facade using a multispectral imaging device mounted on a drone, and simultaneously record environmental perception parameters and the current ground sampling distance; the multi-band images include a first-band image, a second-band image, and a third-band image;

[0064] S2. Perform standardization processing on multi-band images based on environmental perception parameters, and construct the water penetration trajectory inside the wall based on the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction in the second band image.

[0065] S3. Based on the water penetration trajectory, calculate the interlayer attenuation difference and water content progression consistency value of each pixel at different penetration depths, and combine the preset leakage spectral feature library to distinguish the interference area, normal area and leakage area, and generate the corresponding uncertainty index and multispectral detection results.

[0066] S4. Calculate the minimum detectable leakage area based on the current ground sampling distance. When the minimum detectable leakage area exceeds the preset detection requirements, automatically generate a supplementary detection area.

[0067] S5. The spatial location in the supplementary detection area that satisfies the uncertainty index being greater than the preset uncertainty index threshold is determined as the target area for triggering terahertz supplementary detection, and terahertz imaging detection is performed on the target area. For non-supplementary detection areas, the multispectral detection result is directly used as the final detection result, and a corresponding structured detection report is generated.

[0068] Methods for acquiring multi-band images of exterior facades include:

[0069] The drone carries a multispectral imaging device to perform inspection flights along a preset route on the exterior wall of the building. During the flight, the optical axis of the multispectral imaging device is kept facing the exterior wall facade so as to perform continuous multi-band synchronous imaging acquisition of the area to be inspected.

[0070] The multispectral imaging device includes several band imaging units covering the visible light band, near-infrared band and short-wave infrared band. Each band imaging unit synchronously exposes the outer wall area at the same time to obtain multi-band image data of the corresponding area to be detected.

[0071] In this embodiment, it should be noted that the multispectral imaging device can be implemented using a split-type multi-channel multispectral camera or a filter-wheel-switching multispectral camera as described in existing technologies. Each imaging unit corresponds to a narrowband filter channel within a different wavelength range, and each imaging unit includes an imaging lens, a narrowband filter, and an area array image sensor. The visible light imaging unit uses a silicon-based CMOS or CCD image sensor to acquire information on wall texture, cracks, and stains; the near-infrared imaging unit is used to obtain information on changes in the shallow moisture content of the wall; and the short-wave infrared imaging unit can use an InGaAs short-wave infrared detector to obtain moisture absorption response information within and deep within the insulation layer.

[0072] The multi-band image includes a first-band image, a second-band image, and a third-band image. The first-band image preferably includes the 450nm blue light band and the 550nm green light band, used to characterize the texture, cracks, and shadow interference information of the exterior wall cladding layer. The second-band image preferably includes the 850nm near-infrared band, the 1200nm short-wave infrared band 1, and the 1450nm short-wave infrared band 2, used to characterize the water content changes at different penetration depths. The third-band image preferably includes the 1600nm short-wave infrared band 3, used to correct the influence of differences in wall material on the multispectral response.

[0073] During the data acquisition process, environmental sensing parameters of the area to be detected are acquired simultaneously. These parameters include ambient temperature, ambient humidity, solar irradiance, shooting distance, and shooting angle. The current ground sampling distance of the corresponding area to be detected is calculated by combining the UAV's flight altitude, the lens focal length of the multispectral imaging device, and the image resolution.

[0074] In this embodiment, the current ground sampling distance is used to characterize the physical size of a single pixel in the image on the actual exterior wall surface, reflecting the spatial resolution of the multispectral image. A smaller current ground sampling distance indicates a smaller actual area corresponding to a single pixel, and a stronger ability to identify minor leaks; conversely, a larger current ground sampling distance means decreased detection accuracy. Therefore, during multispectral inspection, it is necessary to calculate the current ground sampling distance of the corresponding area in real time, taking into account the UAV's flight altitude, imaging focal length, and image resolution.

[0075] Based on the correspondence between sensor physical size and image resolution, the physical size of a single pixel on the sensor is calculated. Then, combining the flight altitude and lens focal length, and according to the imaging geometric proportions, the single pixel size on the sensor is mapped to the actual surface of the outer wall, thereby obtaining the current ground sampling distance for the corresponding area. The calculation relationship can be expressed as: ;in, Indicates the current ground sampling distance; Indicates the drone's flight altitude; Indicates the size of a single pixel in the sensor; This indicates the focal length of the lens in a multispectral imaging device.

[0076] Methods for performing normalization processing on multi-band images include:

[0077] The system reads environmental sensing parameters that are synchronously recorded with multi-band images and correlates these parameters with the corresponding multi-band images in time. It normalizes and corrects the brightness response of each band image based on solar irradiance and performs environmental compensation based on the spectral response of each band image based on ambient temperature and humidity to reduce the impact of changes in environmental conditions on the band reflection characteristics.

[0078] In this embodiment, it should be noted that radiometric normalization processing is performed on the images of each band. Specifically, solar irradiance, ambient temperature, and ambient humidity parameters recorded synchronously with the multi-band images are read, and a temporal correlation between the environmental parameters and the corresponding images is established. Subsequently, using preset standard irradiance conditions as a reference, the deviation ratio between the current solar irradiance and the standard irradiance conditions is calculated, and this deviation ratio is used to normalize and correct the pixel brightness of each band image to eliminate brightness differences caused by different lighting conditions.

[0079] After brightness normalization, environmental compensation correction is performed on the reflection response of each band image based on real-time ambient temperature and humidity parameters. Specifically, an environmental compensation parameter table is pre-established based on standard wall samples collected under different temperature and humidity conditions to obtain the correspondence between ambient temperature and humidity and the response offset of each band. Subsequently, based on the real-time acquired ambient temperature and humidity parameters, the corresponding compensation coefficient is determined in the environmental compensation parameter table, and this compensation coefficient is used to correct the offset of the reflection response of each band image to eliminate the influence of air moisture absorption and scattering changes on the near-infrared and short-wave infrared bands, as well as the influence of ambient humidity changes on the brightness response of the visible light band, thereby obtaining environmentally compensated multi-band images.

[0080] For example, during a certain detection process, the real-time recorded ambient temperature was 32℃ and the ambient humidity was 78%. Based on the pre-established environmental compensation parameter table, it was determined that the 1450nm shortwave infrared band had an approximately 8% water vapor absorption response attenuation under these temperature and humidity conditions. Therefore, an 8% compensation correction was performed on the reflection response of the corresponding band image. At the same time, the 850nm near-infrared band had an approximately 3% scattering shift, so a corresponding shift correction was performed simultaneously. Finally, a standardized multi-band image after environmental compensation was obtained.

[0081] Based on the shooting distance and shooting angle, establish the corresponding imaging geometric transformation relationship, map the images of each band to the preset standard observation plane, perform perspective correction on images with tilted shooting, and resample the spatial scale differences caused by changes in shooting distance.

[0082] Specifically, an imaging geometric transformation relationship is established based on the synchronously recorded shooting distance and shooting angle. Using a preset standard observation plane as the mapping reference, images of each band are mapped to a unified spatial coordinate system. When the shooting angle deviates from the normal direction of the outer wall, perspective correction is performed on the tilted shooting image through a perspective transformation matrix. At the same time, combined with the current ground sampling distance of the corresponding area, spatial resampling processing is performed on the scale differences caused by the change in shooting distance, so that images acquired under different shooting conditions have a consistent spatial scale.

[0083] Using a preset reference area as a benchmark, the response ranges of the first-band image, the second-band image, and the third-band image are uniformly mapped to form a standardized multi-band image with a unified radiation response benchmark and spatial scale.

[0084] The preset reference area is obtained by pre-collecting standard wall samples under different environmental conditions. The multi-band response values ​​of each spatial area in the standard wall samples under different environmental temperatures and humidity are statistically analyzed, and the response fluctuation amplitude of each spatial area in several acquisition processes is calculated.

[0085] When the response fluctuation amplitude of a corresponding spatial region is less than a preset response fluctuation amplitude threshold, the corresponding spatial region is determined as a preset reference region. The response fluctuation amplitude is obtained by calculating the standard deviation of the multi-band response value of the corresponding spatial region relative to the average multi-band response value during several acquisitions. Subsequently, the response distribution range of the preset reference region in each band image is statistically analyzed, and the response interval of the reference region is used as a unified mapping benchmark to uniformly map the gray values ​​of the first band image, the second band image, and the third band image, making the response values ​​between different bands comparable, thereby forming a standardized multi-band image with a unified radiometric response benchmark and spatial scale.

[0086] Methods for constructing the water penetration trajectory inside a wall include:

[0087] After completing the multi-band image standardization process, the target wall region in the second-band image is extracted, and the pixel response in the second-band image is analyzed in layers along the wall thickness direction to obtain the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction.

[0088] In this embodiment, it should be noted that after completing the multi-band image standardization process, the target wall region is first extracted based on the second-band image to avoid interference from non-wall targets such as windows, air conditioner units, pipelines, and building shadows in subsequent moisture analysis. Specifically, based on the edge contour and regional continuity features of the building's exterior wall in the second-band image, region segmentation processing is performed on the image to identify the main body of the exterior wall. Non-wall regions are then eliminated by combining the building facade boundary information, thereby obtaining the target wall region containing only wall information.

[0089] Taking the target wall region as the analysis object, the pixel response in the second-band image is analyzed in layers along the wall thickness direction. Since moisture infiltration causes continuous changes in the material's water content, the response values ​​of different spatial regions in the second band will exhibit gradual characteristics. Therefore, the response change amplitude and trend along the interlayer direction are calculated on a local pixel neighborhood basis to obtain the response change gradient and continuous change characteristics corresponding to different spatial regions.

[0090] Based on the response change gradient and continuous change characteristics, interlayer progressive response chains with continuous change patterns are identified, and the interlayer progressive response chains are spatially connected to form candidate water penetration paths;

[0091] After acquiring the response change gradient and continuous change characteristics, inter-layer progressive pattern identification is performed. Specifically, spatially adjacent regions are used as the analysis objects, and the consistency of the direction and the degree of continuity of the response change gradients in adjacent regions are compared. When the gradient change directions between adjacent regions are consistent and the gradient difference is within a preset gradient difference threshold, they are identified as the same inter-layer progressive response chain. Subsequently, based on the spatial adjacency relationship and extension direction between each inter-layer progressive response chain, the response chains that meet the continuity condition are spatially connected to form candidate water penetration paths.

[0092] For example, in a scenario involving external wall leakage detection, regions A, B, and C are vertically adjacent, with second-band response gradients of -14, -15, and -13, respectively. The gradients all decrease in direction, and the gradient difference between adjacent regions is less than 3. Based on this continuous variation pattern, regions A, B, and C are identified as a progressive response chain within the same layer, and their spatial relationships are used to connect them, forming a candidate moisture penetration path approximately 0.85m in length.

[0093] Connectivity and stability analysis is performed on candidate water penetration paths to eliminate abnormal discrete responses and isolated regions, retain paths with continuous response changes, and sort and connect the retained paths according to the direction of response change to form the water penetration trajectory inside the wall that characterizes the direction of water migration and penetration depth.

[0094] After forming candidate moisture penetration paths, each response region constituting the candidate path is first marked as a connected component, and the spatial location, boundary range, and center coordinates of each response region are recorded. Then, using each response region as a reference region, its spatial adjacency with adjacent response regions is calculated. Spatial adjacency can be characterized by boundary contact, center point distance, or pixel connectivity; when two response regions have boundary contact, continuous pixel connection, or a center point distance less than a preset spatial interval threshold, they are determined to have spatial connectivity.

[0095] To avoid false connections caused by local noise, connectivity strength analysis is performed on response regions with established spatial connectivity. Specifically, the shared boundary length and the number of continuously connected pixels between adjacent response regions are statistically analyzed, and Min-Max normalization and weighted fusion are performed to construct a connectivity strength index between regions. When the connectivity strength index is higher than a preset threshold, the corresponding response regions are determined to have stable spatial connectivity; otherwise, they are considered weakly connected or disconnected regions. Based on the connectivity relationships and connectivity strengths between each response region, a spatial connectivity network corresponding to the candidate path is constructed, and the overall continuity of the network is analyzed. When there are obvious breaks, isolated nodes, or multiple response regions that cannot form a continuous connection chain within the candidate path, the corresponding region is identified as an abnormal discrete response or an isolated region and is removed. Response regions that can form a continuous connection network and have stable connectivity strength are retained as effective moisture propagation paths.

[0096] After completing the connectivity analysis, a stability analysis is performed on the retained paths. Specifically, the response value change sequence is statistically analyzed along the extension direction of each path. When the response value in the path extension direction shows a continuous increase or decrease, and the change in response value between adjacent response areas is less than a preset threshold, the path is determined to have stable response change characteristics and is retained as a path with continuous response change. The change in response value between adjacent response areas is obtained by calculating the absolute value of the difference in response value between the path nodes corresponding to adjacent response areas. Based on the direction of response value change and spatial extension direction in the retained paths, the paths are sorted and connected. Paths with consistent response value change directions and continuous spatial positions are merged to determine the dominant direction and extension range of moisture propagation, thereby forming an internal moisture penetration trajectory that characterizes the direction of moisture migration and penetration depth.

[0097] Methods for obtaining interlayer attenuation difference and water content progression consistency values ​​include:

[0098] Using the extension direction of the water penetration trajectory as the direction of water propagation, and the starting position of the water penetration trajectory as the depth reference point, combined with the current ground sampling distance of the corresponding area to be detected, the depth position of the pixels in the area covered by the water penetration trajectory is mapped to obtain the relative penetration depth of each pixel.

[0099] Specifically, the spatial distance between each pixel and its starting position is calculated pixel by pixel along the water penetration trajectory. This pixel distance is then converted into an actual spatial distance using the current ground sampling distance, thus obtaining the relative penetration depth for each pixel. This mapping establishes a correspondence between the two-dimensional pixel positions on the water penetration trajectory and the relative penetration depth inside the wall, providing a depth coordinate basis for subsequent interlayer attenuation analysis. For example, in an exterior wall inspection scenario, the water penetration trajectory is 1.5m long, and the current ground sampling distance for the corresponding area is 2mm / pixel. Using the trajectory's starting position P0 as the depth reference point, when pixels P1, P2, and P3 on the trajectory are 20 pixels, 45 pixels, and 80 pixels away from P0, respectively, the corresponding relative penetration depths are 40mm, 90mm, and 160mm, thus completing the depth position mapping of the water penetration trajectory pixels.

[0100] Pixel response sequences are established according to relative penetration depth, and the response values ​​of each pixel in the second band image are extracted; the response change is calculated based on the response values ​​corresponding to adjacent relative penetration depths, and the interlayer attenuation difference of each pixel at different penetration depths is obtained.

[0101] After completing the depth location mapping, a pixel response sequence is established in order of relative penetration depth from shallow to deep, and the response values ​​of the corresponding pixels are extracted from the second-band image. Specifically, the pixels on the water penetration trajectory are sorted according to their relative penetration depth, and the grayscale value of the corresponding pixel is read according to the spatial coordinates of each pixel in the second-band image as the response value of the corresponding pixel, forming a pixel response sequence arranged along the direction of water propagation to reflect the water response variation pattern at different depth locations.

[0102] The response change is calculated based on the response values ​​corresponding to adjacent relative penetration depths. Specifically, the difference between the pixel response values ​​at adjacent depth positions is used as the response change, and this response change is used as the interlayer attenuation difference between corresponding depth positions to characterize the degree of response attenuation at adjacent depth positions during the propagation of moisture along the penetration direction.

[0103] After obtaining the pixel response sequence, the response change is calculated based on the response values ​​corresponding to adjacent relative penetration depths to obtain the interlayer attenuation difference at different penetration depths. Specifically, the response value difference at adjacent depths is used as the interlayer attenuation characterization parameter, and the degree of response change in the direction of water propagation is calculated segment by segment to form the corresponding interlayer attenuation difference sequence. The interlayer attenuation difference is used to reflect the attenuation characteristics and local resistance changes generated during water propagation at different depths.

[0104] Using a preset depth range as the analysis window, the direction and fluctuation of the interlayer attenuation difference within the window are statistically analyzed, and the water cut progression consistency value is calculated based on the directional consistency and fluctuation of the interlayer attenuation difference in adjacent depth ranges.

[0105] After obtaining the interlayer attenuation difference, a preset depth interval is used as the analysis window to quantitatively analyze the progressive characteristics of the moisture propagation process. Specifically, a continuous depth analysis window is set along the direction of moisture propagation, and the direction and fluctuation of the interlayer attenuation difference within each analysis window are statistically analyzed. The direction of change is used to characterize whether the interlayer attenuation difference shows an increasing, decreasing, or stable trend within the current depth interval; the fluctuation level is used to characterize the discrete variation amplitude of the interlayer attenuation difference within the same depth interval.

[0106] The directional consistency between adjacent depth intervals is then assessed. When the interlayer attenuation difference changes in the same direction between adjacent depth intervals, a directional consistency marker is assigned; when the directions of change are inconsistent or reversed, a directional inconsistency marker is assigned. The fluctuation amplitude of the corresponding interlayer attenuation difference between adjacent depth intervals is further statistically analyzed, and the directional consistency marker value is combined with the fluctuation amplitude to calculate the corresponding water cut progression consistency value. Specifically, the higher the degree of directional consistency and the smaller the fluctuation amplitude, the higher the corresponding water cut progression consistency value; conversely, the lower the degree of directional consistency or the larger the fluctuation amplitude, the lower the corresponding water cut progression consistency value.

[0107] For example, a weighted combination method can be used to calculate the progressive consistency value of water content: ;in, This indicates a consistent value for progressive water content; The directional consistency coefficient is used to characterize the degree of consistency in the direction of change of interlayer attenuation difference between adjacent depth intervals. This represents the normalized fluctuation coefficient, used to characterize the degree of fluctuation of the interlayer attenuation difference within adjacent depth intervals; and This indicates the preset weighting coefficients, and and The sum is 1;

[0108] In this embodiment, it should be noted that the directional consistency coefficient is used to characterize the degree of consistency in the direction of change of interlayer attenuation difference between adjacent depth intervals. Specifically, a preset depth interval is used as the analysis window. The direction of change of interlayer attenuation difference within each analysis window is statistically analyzed along the water penetration trajectory, and the direction of change between adjacent analysis windows is compared. When the direction of change of adjacent analysis windows is consistent, it is determined that the directions are consistent; otherwise, it is determined that the directions are inconsistent. Subsequently, the number of adjacent depth intervals with consistent directions and the total number of adjacent depth intervals participating in the comparison are counted, and the directional consistency coefficient is obtained by the ratio of the number of intervals with consistent directions to the total number of intervals compared.

[0109] The normalized fluctuation coefficient is used to characterize the degree of fluctuation of interlayer attenuation difference within adjacent depth intervals. Specifically, firstly, the dispersion of interlayer attenuation difference within a preset depth interval is statistically analyzed, and the maximum and minimum values ​​of interlayer attenuation difference involved in the statistics are obtained; then, the dispersion is normalized using the range of variation of interlayer attenuation difference, thereby obtaining the normalized fluctuation coefficient to eliminate the influence of differences in numerical ranges in different detection areas on the results.

[0110] The preset weighting coefficients are used to adjust the influence ratio of the directional consistency coefficient and the normalized fluctuation coefficient in the numerical calculation of water content progression consistency. Specifically, they can be preset according to the requirements of the detection task, or based on the detection results of historical leakage samples and non-leakage samples, the classification effect of different weight combinations can be compared, and the weight combination that meets the preset recognition requirements can be selected as the preset weighting coefficients.

[0111] Methods for generating corresponding uncertainty indices and multispectral detection results include:

[0112] The system calls a preset leakage spectral feature library, which stores the multispectral response features corresponding to the interference area, normal area, and leakage area. The multispectral response features include features of the interference area, normal area, and leakage area. Taking the area to be detected as the analysis object, the system extracts the multi-band response features of the area to be detected in the first band image, the second band image, and the third band image.

[0113] In this embodiment, it should be noted that, to achieve accurate identification of leakage defects in the external wall insulation layer, a pre-defined leakage spectral feature library is first constructed. Specifically, multispectral sample data of external walls in known states are collected, and the sample areas are categorized based on on-site detection results or manual annotation results, forming interference area samples, normal area samples, and leakage area samples. Subsequently, response feature statistics are performed on the first band image, second band image, and third band image corresponding to each sample area to obtain the response value range of each band, the response ratio relationship between bands, and the typical response distribution law. Based on the average response change law of samples of the same category in each band, corresponding band response reference curves are established, and they are classified and stored according to the area category, thereby forming the pre-defined leakage spectral feature library.

[0114] It should be noted that the band response reference curve is used to characterize the typical spectral response variation of a certain category of region under different bands. The response values ​​of samples of the same category in each band are statistically averaged to obtain the average response level of the corresponding band. Specifically, taking a certain band as the statistical object, the average response value of all samples of that category in that band is calculated; after completing the statistics for all bands, the average band response sequence corresponding to that category can be obtained. Further, by connecting the average band response sequences with the band center wavelength as the abscissa and the corresponding average response value as the ordinate, the band response reference curve for the corresponding category is established. A preset leakage spectral feature library stores features of interference areas, normal areas, and leakage areas, used to characterize the multispectral response modes corresponding to different wall conditions.

[0115] After constructing the pre-defined leakage spectral feature library, multi-band response features are extracted using the area to be detected as the analysis object. Specifically, based on the spatial location of the area to be detected in the first, second, and third band images, the band response values ​​of the corresponding pixels within the area are read, and the average value, range of variation, or distribution characteristics of the responses of each band within the area are statistically analyzed to obtain the multi-band response features corresponding to the area to be detected. For example, in a certain area to be detected, the corresponding pixel response values ​​are read from the three types of band images based on its spatial coordinates. The average response value of the first band is 138, the average response value of the second band is 109, and the average response value of the third band is 92, thus forming the multi-band response features corresponding to this area.

[0116] By combining the interlayer attenuation difference and the water content consistency value of the corresponding regions, a comprehensive spectral feature of the area to be tested is constructed. After obtaining the multi-band response features, the comprehensive spectral feature of the area to be tested is constructed by combining the interlayer attenuation difference and the water content consistency value of the corresponding regions. Specifically, the response features of the area to be tested in each band are used as the basis for spectral characterization, and the interlayer attenuation difference is introduced to reflect the attenuation law of water along the penetration direction. At the same time, the water content consistency value is used to characterize the continuity of water propagation. By combining these features, a comprehensive spectral feature including spectral response features and water propagation features is formed to enhance the ability to distinguish between real leakage and interference anomalies. For example, in the above-mentioned area to be tested, the overall distribution of the interlayer attenuation difference is statistically obtained to be 10~13, and the water content consistency value is 0.88. After combining the above parameters with the corresponding response features, the comprehensive spectral feature of the area is formed.

[0117] Calculate the degree of matching between the comprehensive spectral features of the region to be detected and the category features of each category in the multispectral response features, and determine the region category corresponding to the region to be detected based on the matching results;

[0118] Existing feature matching methods such as Euclidean distance, correlation coefficient, or cosine similarity can be used to calculate the degree of matching. The smaller the feature distance between the comprehensive spectral features of the region to be detected and the corresponding category features, the higher the similarity between them. Further normalization of the feature distance yields a matching degree value within the range of 0 to 1. For example, in a certain region to be detected, the normalized matching degrees between its comprehensive spectral features and those of the leakage area, normal area, and interference area are 0.86, 0.34, and 0.29, respectively, indicating that this region is most similar to the leakage area features; therefore, it is determined to have the highest matching degree with the leakage area features.

[0119] Specifically, when the area to be detected matches the leakage area with the highest degree of feature matching and the corresponding water content progression consistency value reaches the preset consistency value threshold, it is determined to be a leakage area; when the area to be detected matches the normal area with the highest degree of feature matching, it is determined to be a normal area; when the area to be detected matches the interference area with the highest degree of feature matching or the leakage area with the highest degree of feature matching, but the corresponding water content progression consistency value does not reach the preset consistency value threshold, it is determined to be an interference area.

[0120] The uncertainty index is calculated based on the degree of matching between the comprehensive spectral characteristics of the area to be detected and the characteristics of each category. The area category, uncertainty index and corresponding spatial location information are then associated to generate multispectral detection results.

[0121] Specifically, the matching difference between the region to be detected and features of different categories is statistically analyzed. When the matching degree between the region to be detected and a feature of one category is significantly higher than that of other categories, it indicates that the classification result is stable, and a lower uncertainty index is generated accordingly. Conversely, when the matching degree between different categories is close, a higher uncertainty index is generated to reflect the uncertainty of the classification boundary. Subsequently, the region category, uncertainty index, and spatial location information of the region to be detected are correlated and output in a spatial labeling manner to generate the final multispectral detection result. For example, in a certain detection scenario, region A is identified as a leakage region, with a feature matching degree of 0.89 with the leakage region and a maximum matching difference of 0.52 with other categories, thus the uncertainty index is calculated to be 0.11. Region B has matching degrees of 0.63 with the leakage region and 0.58 with the interference region, with small matching differences, thus generating an uncertainty index of 0.43. Finally, in the multispectral detection result image, region A is marked as a low uncertainty leakage region, and region B is marked as a high uncertainty anomaly region.

[0122] Methods for calculating the minimum detectable leakage area include:

[0123] Obtain the current ground sampling distance corresponding to the current detection area and the noise equivalent reflectance difference of the multispectral imaging device under multi-band images; determine the standard leakage reflectance response of a single pixel based on the preset leakage spectral feature library.

[0124] It should be noted that the noise equivalent reflectance difference is an existing sensor calibration parameter used to characterize the minimum reflectance difference that the imaging system can reliably distinguish. Its physical meaning is the minimum detectable reflectance change when the target reflectance change is just higher than the system noise floor. It can be obtained from the sensor's factory calibration data or the device's response calibration results.

[0125] Subsequently, the standard leakage reflectance response of a single pixel is determined based on a pre-defined leakage spectral feature library. Specifically, during the construction of the pre-defined leakage spectral feature library, the response characteristics of normal area samples and leakage area samples in each band are statistically analyzed in advance, and corresponding band response reference curves are established. Using normal areas of the same material as a benchmark, the average response difference between leakage samples and normal samples in the water-sensitive band is calculated. Using a single pixel as the statistical unit, the average response change of leakage samples of the same level is statistically analyzed to obtain the standard leakage reflectance response of a single pixel.

[0126] The standard leakage reflectance response is obtained by statistically analyzing the average response difference between normal area samples and leakage area samples of the same material in multi-band images, and using a single pixel as the corresponding ground area as the statistical unit. Based on the noise equivalent reflectance difference, the current ground sampling distance, and the standard leakage reflectance response, the minimum detectable leakage area of ​​the multi-band image is calculated.

[0127] Minimum detectable leakage area: ;in, This represents the minimum detectable leakage area in a multi-band image. It represents the noise equivalent reflectance difference, used to characterize the smallest reflectance change that a multispectral imaging device can reliably distinguish; This represents the actual area scale corresponding to a single pixel on the wall surface. This represents the standard leakage reflectance response of a single pixel within the ground area. This indicates the cell index identifier.

[0128] Methods for automatically generating supplementary detection areas include:

[0129] The minimum detectable leakage area of ​​the multi-band image is compared with the preset minimum detectable leakage area threshold. When the minimum detectable leakage area is greater than the preset minimum detectable leakage area threshold, it is determined that the current detection resolution cannot meet the preset detection requirements, and the generation of supplementary detection areas is triggered.

[0130] The target ground sampling distance corresponding to the supplementary detection is calculated based on the current ground sampling distance, the minimum detectable leakage area, and the preset minimum detectable leakage area threshold. The target ground sampling distance is obtained by scaling the resolution based on the ratio between the preset minimum detectable leakage area threshold and the current minimum detectable leakage area.

[0131] Target ground sampling distance: ;in, This indicates the target ground sampling distance corresponding to the supplementary detection; This indicates the preset minimum detectable leakage area threshold; This indicates the resolution scaling ratio between the preset minimum detectable leakage area threshold and the current minimum detectable leakage area;

[0132] When the preset minimum detectable leakage area threshold is less than the current minimum detectable leakage area, it indicates that the current detection capability cannot meet the preset minimum detectable leakage area threshold requirement. At this time, the scaling factor is less than 1, which corresponds to reducing the ground sampling distance, i.e., increasing the spatial resolution, thereby improving the detection capability for small-scale leakage areas.

[0133] After obtaining the target ground sampling distance, the candidate areas for supplementary testing are selected by combining the multispectral detection results and the corresponding uncertainty index. The candidate areas for supplementary testing include areas where the uncertainty index is greater than the preset uncertainty index threshold, areas where the water penetration trajectory extends along the water migration direction and intersects with the boundary of the identified leakage area, and areas where the area is smaller than the minimum detectable leakage area and there are abnormal multi-band response characteristics.

[0134] Based on the target ground sampling distance, the corresponding UAV flight altitude and the lens focal length of the multispectral imaging device are calculated. The supplementary acquisition range of the candidate supplementary measurement area that meets the preset target resolution requirements is determined in the area to be detected, and the corresponding supplementary detection area is generated. The area in the area to be detected other than the supplementary detection area is defined as the non-supplementary detection area.

[0135] In this embodiment, the preset target resolution is used to characterize the minimum spatial accuracy that the detection system can reliably resolve at the ground scale. It is comprehensively set based on parameters such as the minimum identifiable leakage size of the target, the spatial resolution capability of the imaging system, and the difference in noise equivalent reflectivity, and is used as a spatial constraint condition for the division of the supplementary detection area. Spatial resolution constraints are further introduced for the supplementary candidate areas. The actual resolution capability of each candidate area at the ground scale is compared with the preset target resolution. Areas that cannot meet the preset target resolution requirements are retained and marked. Furthermore, spatially continuous candidate areas are merged and their boundaries are expanded to determine their ground spatial range, thereby generating a continuous supplementary acquisition area that meets the preset target resolution requirements, ultimately forming the corresponding supplementary detection area.

[0136] For example, in a certain embodiment of external wall leakage detection, the preset target resolution is set to 0.05m. This resolution is determined by a combination of the minimum identifiable leakage size of the target, the spatial resolution of the imaging system, and the difference in noise equivalent reflectivity, and is used to limit the spatial precision of supplementary detection.

[0137] After completing the screening of candidate areas for supplementary testing, for example, three types of candidate areas were identified in the building's exterior walls: one type is a local area with an uncertainty index of 0.82 due to shadowing; another type is an area where the water penetration trajectory extends along the crack and intersects with the leakage boundary; and the third type is an area of ​​approximately 0.03m². 2 Furthermore, there are local areas with multi-band response anomalies. Subsequently, the ground resolution corresponding to the above candidate areas is compared with the preset target resolution of 0.05m. Areas that cannot meet this resolution constraint are retained, and spatially adjacent candidate units are merged. For example, adjacent crack extension areas are merged to form a continuous area block of about 2m×3m, thereby generating a supplementary detection area that meets the preset target resolution requirements.

[0138] Methods for generating corresponding structured inspection reports include:

[0139] After the supplementary detection area is generated, the spatial location in the supplementary detection area that satisfies the uncertainty index being greater than the preset uncertainty index threshold is determined as the target area for triggering terahertz supplementary detection, and terahertz imaging detection is performed on the target area to obtain the corresponding terahertz detection results.

[0140] Target area that triggers terahertz supplemental detection: ;in, This indicates the target area that triggers terahertz supplemental detection; This represents the spatial coordinates of any pixel within the area to be detected. Indicates spatial location The corresponding uncertainty index; This indicates the preset uncertainty index threshold;

[0141] For non-supplementary detection areas, the corresponding multispectral detection results are directly used as the final detection results; for target areas that trigger terahertz supplementary detection, the terahertz detection results and multispectral detection results are fused to generate the final detection results. Based on the final detection results and the features of each category in the multispectral response characteristics, combined with spatial location information, a structured detection report is generated.

[0142] The terahertz detection results are fused with the multispectral detection results: ;in, Indicates spatial location The corresponding fused moisture content result, i.e. the final detection result, is the final output value after fusing the multispectral detection result and the terahertz detection result; This indicates the corresponding multispectral moisture content detection result, i.e., the multispectral detection result; Indicates spatial location The corresponding terahertz detection results; This represents the weighting coefficient of the terahertz detection results in the fusion process, used to characterize the degree of contribution of terahertz detection information to the final result; This represents the weighting coefficient of the multispectral detection results in the fusion process, used to characterize the degree of contribution of multispectral detection information to the final result.

[0143] For example, within a certain supplementary detection area, the location obtained by multispectral detection Corresponding moisture content results The water content at this location was 18%; the deep water content was obtained through terahertz supplementary detection. The final moisture content at this location is 26%. Since this area is located deep within the insulation layer and the uncertainty index for multispectral detection is high, the contribution of terahertz detection is increased during the fusion process. The weighting coefficient for terahertz detection is set to 0.7, and the weighting coefficient for multispectral detection is set to 0.3. Therefore, the final fused moisture content at this location is: This example demonstrates that when terahertz detection has higher reliability or stronger deep water characterization capabilities, it can be improved by increasing... Enhance the influence of terahertz results on the final detection results; conversely, when the stability of multispectral detection is high, it can be appropriately increased. To maintain the dominant role of routine testing results.

[0144] It should be noted that spatial location information, region category information, and the corresponding final detection results are structurally correlated to form a data mapping relationship based on spatial coordinates. This data is then aggregated according to a preset spatial grid to generate structured detection data units. Each structured detection data unit is then summarized and processed to generate a structured detection report.

[0145] The preset consistency value threshold is set by staff based on historical data analysis results. This historical analysis process includes the system collecting multiple consistency values ​​and calculating their average value as a reference to obtain the preset consistency value threshold. Similarly, the preset minimum detectable leakage area threshold, preset uncertainty index threshold, preset gradient difference threshold, and preset connectivity strength index threshold are also set by staff based on the system's historical operating data and specific application scenario requirements. During system operation, these thresholds are adjusted by staff according to the actual situation.

[0146] Example 2, please refer to Figure 2As shown, for parts not described in detail in this embodiment, please refer to the description in Embodiment 1. A multispectral imaging detection system for leakage defects in external wall insulation layers is provided, including:

[0147] The band acquisition module is used to simultaneously acquire multi-band images of the exterior facade using a multispectral imaging device mounted on a UAV, and simultaneously record environmental perception parameters and the current ground sampling distance; the multi-band images include a first band image, a second band image, and a third band image;

[0148] The standardization analysis module is used to perform standardization processing on multi-band images based on environmental perception parameters, and to construct the water penetration trajectory inside the wall based on the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction in the second band image.

[0149] The leakage identification module is used to calculate the interlayer attenuation difference and water content progression consistency value of each pixel at different penetration depths based on the water penetration trajectory, and to distinguish the interference area, normal area and leakage area by combining the preset leakage spectral feature library, and generate the corresponding uncertainty index and multispectral detection results.

[0150] The sampling correction module is used to calculate the minimum detectable leakage area based on the current ground sampling distance. When the minimum detectable leakage area exceeds the preset detection requirements, it automatically generates a supplementary detection area.

[0151] The report generation module identifies spatial locations within the supplementary detection area where the uncertainty index exceeds a preset uncertainty index threshold as target areas for triggering terahertz supplementary detection. Terahertz imaging detection is then performed on these target areas. For non-supplementary detection areas, multispectral detection results are directly used as the final detection results, and a corresponding structured detection report is generated. All modules are connected via wired and / or wireless means to enable data transmission between them.

[0152] Example 3

[0153] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the operation mode of the multispectral imaging detection method for leakage defects in external wall insulation layers provided above.

[0154] Since the electronic device described in this embodiment is the electronic device used in implementing the multispectral imaging detection method and system for external wall insulation layer leakage in this application embodiment, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the multispectral imaging detection method and system for external wall insulation layer leakage described in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. Any electronic device used by those skilled in the art in implementing the multispectral imaging detection method and system for external wall insulation layer leakage in this application embodiment falls within the scope of protection of this application.

[0155] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0156] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A multispectral imaging detection method for leakage defects in external wall insulation layers, characterized in that, include: S1. Simultaneously acquire multi-band images of the exterior facade using a multispectral imaging device mounted on a drone, and simultaneously record environmental perception parameters and the current ground sampling distance; the multi-band images include a first-band image, a second-band image, and a third-band image; S2. Perform standardization processing on multi-band images based on environmental perception parameters, and construct the water penetration trajectory inside the wall based on the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction in the second band image. S3. Based on the water penetration trajectory, calculate the interlayer attenuation difference and water content progression consistency value of each pixel at different penetration depths, and combine the preset leakage spectral feature library to distinguish the interference area, normal area and leakage area, and generate the corresponding uncertainty index and multispectral detection results. S4. Calculate the minimum detectable leakage area based on the current ground sampling distance. When the minimum detectable leakage area exceeds the preset detection requirements, automatically generate a supplementary detection area. S5. The spatial location in the supplementary detection area that satisfies the uncertainty index being greater than the preset uncertainty index threshold is determined as the target area for triggering terahertz supplementary detection, and terahertz imaging detection is performed on the target area. For non-supplementary detection areas, the multispectral detection result is directly used as the final detection result, and a corresponding structured detection report is generated.

2. The multispectral imaging detection method for leakage defects in external wall insulation layers according to claim 1, characterized in that, The method for acquiring multi-band images of the exterior wall facade includes: The drone carries a multispectral imaging device to perform inspection flights along a preset route on the exterior wall of the building. During the flight, the optical axis of the multispectral imaging device is kept facing the exterior wall facade so as to perform continuous multi-band synchronous imaging acquisition of the area to be inspected. The multispectral imaging device includes several band imaging units covering the visible light band, near-infrared band, and short-wave infrared band. Each band imaging unit synchronously exposes the outer wall area at the same time to obtain multi-band image data of the corresponding area to be detected. The multi-band image includes a first band image, a second band image, and a third band image. During the data acquisition process, environmental sensing parameters of the area to be detected are acquired simultaneously. These parameters include ambient temperature, ambient humidity, solar irradiance, shooting distance, and shooting angle. The current ground sampling distance of the corresponding area to be detected is calculated by combining the UAV's flight altitude, the lens focal length of the multispectral imaging device, and the image resolution.

3. The multispectral imaging detection method for leakage defects in external wall insulation layers according to claim 2, characterized in that, The method for performing normalization processing on multi-band images includes: Read the environmental perception parameters recorded synchronously with the multi-band images, and correlate the environmental perception parameters with the corresponding multi-band images in time; normalize and correct the brightness response of each band image according to the solar irradiance, and perform environmental compensation according to the spectral response of each band image according to the ambient temperature and humidity. Perform environmental compensation: A table of environmental compensation parameters is established in advance based on standard wall samples collected under different temperature and humidity conditions to obtain the correspondence between environmental temperature and humidity and the response offset of each band; based on the real-time acquired environmental temperature and humidity parameters, the corresponding compensation coefficients are determined in the environmental compensation parameter table, and the offset of the reflection response of each band image is corrected using the compensation coefficients. Based on the shooting distance and shooting angle, establish the corresponding imaging geometric transformation relationship, map the images of each band to the preset standard observation plane, perform perspective correction on images with tilted shooting, and resample the spatial scale differences caused by changes in shooting distance. Using a preset reference area as a benchmark, the response ranges of the first-band image, the second-band image, and the third-band image are uniformly mapped to form a standardized multi-band image with a unified radiation response benchmark and spatial scale. The preset reference area is obtained by pre-collecting standard wall samples under different environmental conditions. The multi-band response values ​​of each spatial area in the standard wall samples under different environmental temperatures and humidity are statistically analyzed, and the response fluctuation amplitude of each spatial area in several acquisition processes is calculated. When the response fluctuation amplitude of the corresponding spatial region is less than the preset response fluctuation amplitude threshold, the corresponding spatial region is determined as the preset reference region; the response fluctuation amplitude is obtained by calculating the standard deviation of the multi-band response value of the corresponding spatial region relative to the average multi-band response value during several acquisitions.

4. The multispectral imaging detection method for leakage defects in external wall insulation layers according to claim 3, characterized in that, The method for constructing the water penetration trajectory inside the wall includes: After completing the multi-band image standardization process, the target wall region in the second-band image is extracted, and the pixel response in the second-band image is analyzed in layers along the wall thickness direction to obtain the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction. Based on the response change gradient and continuous change characteristics, interlayer progressive response chains with continuous change patterns are identified. When the gradient change direction between adjacent regions is consistent and the gradient difference is within a preset gradient difference threshold, they are identified as the same interlayer progressive response chain. The interlayer progressive response chains are then spatially connected to form candidate water penetration paths. Connectivity and stability analysis is performed on candidate water penetration paths to eliminate abnormal discrete responses and isolated regions, retain paths with continuous response changes, and sort and connect the retained paths according to the direction of response change to form the water penetration trajectory inside the wall that characterizes the direction of water migration and penetration depth. Connectivity analysis is performed on candidate water penetration paths: the length of the shared boundary and the number of consecutive connected pixels between adjacent response regions are statistically analyzed, and Min-Max normalization and weighted fusion are performed to construct a connectivity strength index between regions; when the connectivity strength index is higher than the preset connectivity strength index threshold, it is determined that there is a stable spatial connection between the corresponding response regions; otherwise, it is regarded as a weak connection or non-connected region. Stability analysis is performed on candidate moisture penetration paths: the sequence of response value changes is statistically analyzed along the extension direction of each path. When the response value in the extension direction of the path shows a continuous increase or decrease, and the change in response value between adjacent response areas is less than the preset threshold for the change in response value, the path is determined to have stable response change characteristics. The change in response value between adjacent response areas is obtained by calculating the absolute value of the difference in response value between the path nodes corresponding to the adjacent response areas.

5. The multispectral imaging detection method for leakage defects in external wall insulation layers according to claim 4, characterized in that, The methods for obtaining the interlayer attenuation difference and the water content progression consistency value include: Using the extension direction of the water penetration trajectory as the direction of water propagation, and the starting position of the water penetration trajectory as the depth reference point, combined with the current ground sampling distance of the corresponding area to be detected, the depth position of the pixels in the area covered by the water penetration trajectory is mapped to obtain the relative penetration depth of each pixel. Pixel response sequences are established according to relative penetration depth, and the response values ​​of each pixel in the second band image are extracted; the response change is calculated based on the response values ​​corresponding to adjacent relative penetration depths, and the interlayer attenuation difference of each pixel at different penetration depths is obtained. Using a preset depth range as the analysis window, the direction and fluctuation of the interlayer attenuation difference within the window are statistically analyzed, and the water cut progression consistency value is calculated based on the direction consistency and fluctuation of the interlayer attenuation difference in adjacent depth ranges. Consistent values ​​for progressive water content: ;in, This indicates a consistent value for progressive water content; Indicates the directional consistency coefficient; Represents the normalized volatility coefficient; and This indicates the preset weighting coefficients, and and The sum is 1.

6. The multispectral imaging detection method for leakage defects in external wall insulation layers according to claim 5, characterized in that, The method for generating the corresponding uncertainty index and multispectral detection results includes: The system calls a preset leakage spectral feature library, which stores the multispectral response features corresponding to the interference area, normal area, and leakage area. The multispectral response features include features of the interference area, normal area, and leakage area. Taking the area to be detected as the analysis object, the system extracts the multi-band response features of the area to be detected in the first band image, the second band image, and the third band image. By combining the interlayer attenuation difference and water content progression consistency values ​​of the corresponding regions, a comprehensive spectral feature of the region to be detected is constructed; the degree of matching between the comprehensive spectral feature of the region to be detected and the feature of each category in the multispectral response feature is calculated, and the region category corresponding to the region to be detected is determined according to the matching results; the degree of matching between the feature of each category is obtained by Euclidean distance, correlation coefficient or cosine similarity. Specifically, when the area to be detected matches the leakage area with the highest degree of feature matching and the corresponding water content progressive consistency value reaches the preset consistency value threshold, it is determined to be a leakage area; when the area to be detected matches the normal area with the highest degree of feature matching, it is determined to be a normal area; when the area to be detected matches the interference area with the highest degree of feature matching or the leakage area with the highest degree of feature matching, but the corresponding water content progressive consistency value does not reach the preset consistency value threshold, it is determined to be an interference area. The uncertainty index is calculated based on the degree of matching between the comprehensive spectral characteristics of the area to be detected and the characteristics of each category. The area category, uncertainty index and corresponding spatial location information are then associated to generate multispectral detection results.

7. The multispectral imaging detection method for leakage defects in external wall insulation layers according to claim 6, characterized in that, The method for calculating the minimum detectable leakage area includes: Obtain the current ground sampling distance corresponding to the current detection area and the noise equivalent reflectance difference of the multispectral imaging device under multi-band images; determine the standard leakage reflectance response of a single pixel based on the preset leakage spectral feature library. The standard leakage reflectance response is obtained by statistically analyzing the average response difference between normal area samples and leakage area samples of the same material in multi-band images, and using a single pixel as the corresponding ground area as the statistical unit. Based on the noise equivalent reflectance difference, the current ground sampling distance, and the standard leakage reflectance response, the minimum detectable leakage area of ​​the multi-band image is calculated.

8. The multispectral imaging detection method for leakage defects in external wall insulation layers according to claim 7, characterized in that, The method for automatically generating supplementary detection areas includes: The minimum detectable leakage area of ​​the multi-band image is compared with the preset minimum detectable leakage area threshold. When the minimum detectable leakage area is greater than the preset minimum detectable leakage area threshold, it is determined that the current detection resolution cannot meet the preset detection requirements, and the generation of supplementary detection areas is triggered. The target ground sampling distance for supplementary testing is calculated based on the current ground sampling distance, the minimum detectable leakage area, and the preset minimum detectable leakage area threshold. After obtaining the target ground sampling distance, the candidate areas for supplementary testing are screened based on the multispectral detection results and the corresponding uncertainty index. Candidate regions for supplementary testing include regions where the uncertainty index is greater than the preset uncertainty index threshold, regions where the water penetration trajectory extends along the direction of water migration and intersects with the boundary of the identified leakage region, and regions where the area is smaller than the minimum detectable leakage area and where there are abnormal multi-band response characteristics. Based on the target ground sampling distance, the corresponding UAV flight altitude and the lens focal length of the multispectral imaging device are calculated. The supplementary acquisition range of the candidate supplementary measurement area that meets the preset target resolution requirements is determined in the area to be detected, and the corresponding supplementary detection area is generated. The area in the area to be detected other than the supplementary detection area is defined as the non-supplementary detection area.

9. A multispectral imaging detection method for leakage defects in external wall insulation layers according to claim 8, characterized in that, The method for generating the corresponding structured detection report includes: After the supplementary detection area is generated, the spatial location in the supplementary detection area that satisfies the uncertainty index being greater than the preset uncertainty index threshold is determined as the target area for triggering terahertz supplementary detection, and terahertz imaging detection is performed on the target area to obtain the corresponding terahertz detection results. For non-supplementary detection areas, the corresponding multispectral detection results are directly used as the final detection results; for target areas that trigger terahertz supplementary detection, the terahertz detection results and multispectral detection results are fused to generate the final detection results. Based on the final detection results and the features of each category in the multispectral response characteristics, combined with spatial location information, a structured detection report is generated.

10. A multispectral imaging detection system for leakage defects in external wall insulation layers, used to implement the multispectral imaging detection method for leakage defects in external wall insulation layers as described in any one of claims 1 to 9, characterized in that, include: The band acquisition module is used to simultaneously acquire multi-band images of the exterior facade using a multispectral imaging device mounted on a UAV, and simultaneously record environmental perception parameters and the current ground sampling distance; the multi-band images include a first band image, a second band image, and a third band image; The standardization analysis module is used to perform standardization processing on multi-band images based on environmental perception parameters, and to construct the water penetration trajectory inside the wall based on the response change gradient and continuous change characteristics of different spatial regions in the interlayer direction in the second band image. The leakage identification module is used to calculate the interlayer attenuation difference and water content progression consistency value of each pixel at different penetration depths based on the water penetration trajectory, and to distinguish the interference area, normal area and leakage area by combining the preset leakage spectral feature library, and generate the corresponding uncertainty index and multispectral detection results. The sampling correction module is used to calculate the minimum detectable leakage area based on the current ground sampling distance. When the minimum detectable leakage area exceeds the preset detection requirements, it automatically generates a supplementary detection area. The report generation module is used to identify the spatial locations in the supplementary detection area that meet the condition that the uncertainty index is greater than the preset uncertainty index threshold as the target areas for triggering terahertz supplementary detection, and to perform terahertz imaging detection on the target areas. For non-supplementary detection areas, the multispectral detection results are directly used as the final detection results, and a corresponding structured detection report is generated.

Images