Unmanned aerial vehicle based prefabricated building defect automatic identification system

Abstract

The application belongs to the technical field of intelligent detection of building engineering, and particularly relates to an automatic defect identification system for fabricated buildings based on an unmanned aerial vehicle, which comprises a data acquisition module, a data registration module, a sleeve positioning module, a feature extraction module, a defect identification module and a report output module. The system collects multi-source heterogeneous data through the unmanned aerial vehicle, constructs a time-sequenced three-dimensional data field, eliminates the influence of environmental interference through pretreatment, extracts visible light image features, infrared thermal imaging features and three-dimensional point cloud geometric features in parallel, realizes comprehensive diagnosis of the internal and external filling state of grouting, accurately identifies insufficient hollows of grouting, late shrinkage hollows and defects of insufficient vertical depth of grouting, simultaneously carries out dynamic tracking, accurate quantification and tracing, effectively solves the problems that the existing detection of the filling state of sleeve grouting of fabricated buildings is single in dimension, poor in space-time continuity, weak in environmental adaptability and difficult to accurately quantify and trace defects, and improves the intelligent reliability of quality control.

Description

Technical Field

[0001] This invention belongs to the field of intelligent inspection technology in building engineering, specifically an automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs). Background Technology

[0002] Prefabricated buildings have gained widespread application globally due to their efficient and environmentally friendly construction advantages. Among them, component connection nodes, especially sleeve grouting connections, are the core to ensure the integrity and safety of the structure, and the fullness of the internal grouting directly determines the long-term durability and seismic performance of the building.

[0003] Currently, in the quality inspection of grouting connection nodes in prefabricated buildings, existing technologies are mainly implemented through three paths: manual inspection, in which technicians check the grout overflow and macroscopic defects on the node surface by visual inspection and tapping; fixed-point sensing, in which strain gauges, fiber optic sensors and other devices are pre-embedded in the node or on the surface to monitor its stress, strain or vibration signals to detect anomalies; and automated analysis based on static data, which uses drones equipped with cameras or lidar to collect node images or point clouds in a single time series, and uses algorithms to identify surface cracks, dimensional deviations or geometric deformations.

[0004] The existing technology has the following three main defects: 1. The existing technology has a single detection dimension. Neither manual inspection nor static automated methods can obtain the internal three-dimensional filling state, and fixed-point sensing is difficult to quantify the spatial distribution of defects.

[0005] 2. Existing technologies cannot couple the time-series data of the entire grouting construction and maintenance process, resulting in a break in the spatiotemporal continuity and making it impossible to trace the evolution trajectory and causes of defects.

[0006] 3. Existing technologies do not adequately consider environmental adaptability. Visible light and infrared data are easily affected by on-site conditions such as light and temperature, and lack an effective adaptive correction mechanism, which affects the reliability and comparability of the test results. Summary of the Invention

[0007] To overcome the shortcomings in the background art, embodiments of the present invention provide an automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs), which can effectively solve the problems involved in the background art.

[0008] The objective of this invention can be achieved through the following technical solution: an automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs), comprising: a data acquisition module, a data registration module, a sleeve positioning module, a feature extraction module, a defect identification module, and a report output module.

[0009] The data acquisition module is connected to the data registration module, the data registration module is connected to the sleeve positioning module, the sleeve positioning module is connected to the feature extraction module, the feature extraction module is connected to the defect identification module, and the defect identification module is connected to the report output module.

[0010] The data acquisition module controls a drone equipped with multimodal sensors to perform synchronous data acquisition along a preset path at key grouting timing nodes at the connection points of prefabricated building components, thereby acquiring multi-source heterogeneous data with time-series markers.

[0011] The data registration module performs fusion registration on the multi-source heterogeneous data based on a spatiotemporal synchronization strategy to construct a temporally sequenced three-dimensional data field.

[0012] The sleeve positioning module identifies the three-dimensional coordinates of the grouting hole and the grout outlet hole of each sleeve based on the time-series three-dimensional data field, and generates the corresponding three-dimensional bounding box of the sleeve with the line connecting the centers of the two holes as the axis, combined with the sleeve design parameters and the measured spatial posture.

[0013] The feature extraction module extracts visible light image features, infrared thermal imaging features, and three-dimensional point cloud geometric features related to grout fullness in parallel within the three-dimensional bounding box of the sleeve, forming a multi-physics evidence chain.

[0014] The defect identification module performs logical reasoning on the multi-physics evidence chain to identify the type of grouting defect and associate its time sequence nodes to trace the cause of the defect.

[0015] The report output module performs risk assessment based on the type and timing of grouting defects and outputs a structured inspection report.

[0016] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention breaks through the limitations of existing single-dimensional detection by multi-physics evidence chain fusion analysis. By extracting visible light image features, infrared thermal imaging features and three-dimensional point cloud geometric features in parallel, it realizes the comprehensive diagnosis of the grouting full state, and can accurately identify hidden problems that existing methods cannot detect, such as insufficient grouting voids, late shrinkage voids, and insufficient vertical grouting depth.

[0017] (2) This invention achieves dynamic monitoring of the entire grouting process by creating a time-series three-dimensional data field construction mechanism, continuously collects and registers data at key grouting time nodes, establishes spatiotemporal evolution analysis of grouting status, solves the process blind zone limitation of static detection methods, and supports accurate tracing of defect formation mechanism and cause determination.

[0018] (3) The present invention establishes a multi-source data adaptive correction system, which significantly improves the detection reliability under complex working conditions. Through technologies such as ambient temperature compensation, emissivity correction and adaptive adjustment of illumination, it effectively eliminates the interference of environmental factors on the detection data and ensures the consistency and accuracy of the detection results under different construction conditions. Attached Figure Description

[0019] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the module connection of the present invention.

[0021] Figure 2 This is a schematic diagram illustrating the construction logic of the temporal three-dimensional data field in the data registration module of this invention.

[0022] Figure 3 This is a schematic diagram illustrating the logical structure of the multiphysics evidence chain in the feature extraction module of this invention. Detailed Implementation

[0023] 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.

[0024] Reference Figure 1 As shown, the present invention provides an automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs), including: a data acquisition module, a data registration module, a sleeve positioning module, a feature extraction module, a defect identification module, and a report output module.

[0025] The data acquisition module is connected to the data registration module, the data registration module is connected to the sleeve positioning module, the sleeve positioning module is connected to the feature extraction module, the feature extraction module is connected to the defect identification module, and the defect identification module is connected to the report output module.

[0026] The data acquisition module controls a drone equipped with multimodal sensors to perform synchronous data acquisition along a preset path at key grouting timing nodes at the connection points of prefabricated building components, thereby acquiring multi-source heterogeneous data with time-series markers.

[0027] It should be noted that the aforementioned multimodal sensor includes at least a visible light camera component, an infrared thermal imaging component, a lidar scanning component, and an environmental parameter acquisition component.

[0028] The key grouting timing nodes include at least the baseline time node before grouting, the time node during grouting, the immediate time node after grouting ends, and at least one preset cycle node during the curing stage.

[0029] The implementation of time-stamped multi-source heterogeneous data is based on the following: during data acquisition, all sensors achieve time synchronization through a hardware triggering mechanism and use network time protocols for system-level timestamp calibration. Each acquired multimodal data packet, including visible light image sequences, infrared thermal image sequences, lidar point cloud data, synchronous ambient light intensity readings, ambient temperature readings, and component surface temperature readings, is bound to the corresponding grouting time sequence node, forming time-stamped multi-source heterogeneous data.

[0030] The data registration module performs fusion registration on the multi-source heterogeneous data based on a spatiotemporal synchronization strategy to construct a temporally sequenced three-dimensional data field.

[0031] Reference Figure 2 As shown, in a preferred embodiment of the present invention, the construction of the temporal three-dimensional data field includes: using the lidar point cloud of each temporal node as a spatial reference, projecting the visible light and infrared image pixel coordinates at the corresponding time onto the point cloud coordinate system through the camera intrinsic and extrinsic parameters and the fixed transformation relationship with lidar, and assigning RGB color value and infrared temperature value to each three-dimensional point.

[0032] In the projected multi-period point cloud, the previous time-series node point cloud is taken as the target point cloud and the current time-series node point cloud is taken as the source point cloud. For each point in the source point cloud, the corresponding point pair is determined by the nearest neighbor search in the target point cloud.

[0033] It should be noted that the specific process of the nearest neighbor search is as follows: for each point to be registered in the source point cloud, within the entire space of the target point cloud, the single point with the closest spatial Euclidean distance is found through a tree data structure of preset dimensions as a candidate corresponding point.

[0034] Calculate the Euclidean distance between the point to be registered and its candidate corresponding point. If the distance is less than the preset matching distance threshold, and the candidate corresponding point also finds its nearest neighbor in the reverse search, then the two points are accepted as a valid corresponding point pair; otherwise, the candidate point pair is discarded.

[0035] The preset matching distance threshold can be set to 2 to 3 times the average point spacing of the point cloud.

[0036] The rigid transformation matrix that minimizes the mean square error is calculated based on all corresponding point pairs through singular value decomposition.

[0037] The source point cloud is aligned to the target point cloud by applying the transformation matrix. The nearest neighbor search, transformation calculation and point cloud alignment are performed iteratively until the mean square error change is lower than the preset convergence threshold, forming a temporal three-dimensional data field with consistent spatial location and fused RGB color values, infrared temperature values ​​and three-dimensional information.

[0038] In a preferred embodiment of the present invention, before assigning RGB color values ​​and infrared temperature values ​​to each three-dimensional point, the image is further preprocessed in relation to environmental factors: based on the synchronously acquired ambient temperature and the measured temperature of the reference point at the component connection, the infrared thermal image is emissivity corrected and the ambient background radiation compensated to obtain infrared data that reflects the true temperature distribution.

[0039] It should be noted that the specific process of the above emissivity correction is as follows: at the preset reference point of the component connection, the actual temperature of the point is collected synchronously using a contact temperature probe.

[0040] The apparent temperature of the pixel region corresponding to the reference point is extracted from the infrared thermal image.

[0041] Based on the measured temperature and apparent temperature, and according to the relative positional relationship between the infrared thermal imager and the environment, the actual emissivity of the component surface material is calculated by using the Stefan-Boltzmann law.

[0042] The calculated actual emissivity is used as a uniform emissivity parameter for the component surface. Each pixel in the infrared thermal image is recalculated, and the apparent temperature value of each point is corrected to a temperature value based on the true emissivity.

[0043] The specific process of environmental background radiation compensation is as follows: based on the synchronously collected ambient temperature, the radiation intensity of the environmental background is calculated according to the Stefan-Boltzmann law.

[0044] Calculate the ambient radiation component reflected from the surface of the component, and subtract the ambient radiation component from the total radiation signal received by the infrared thermal imager to obtain the thermal radiation intensity of the component itself.

[0045] The separated thermal radiation intensity of the component itself is finally converted into infrared temperature data that characterizes the true temperature distribution of the component, eliminating the influence of environmental reflection.

[0046] Based on the brightness value provided by the ambient light intensity sensor, adaptive brightness correction is performed on the visible light image to obtain RGB color data with consistent illumination.

[0047] It should be noted that the specific process of the above adaptive brightness correction is as follows: analyze the overall brightness level of the visible light image and calculate its average pixel intensity value.

[0048] The average pixel intensity value is compared with the standard brightness range determined based on the brightness value of the ambient light intensity sensor to calculate the required global brightness compensation coefficient.

[0049] Based on the global brightness compensation coefficient, the RGB channel values ​​of all pixels in the visible light image are synchronously scaled and adjusted.

[0050] In the corrected image, for overexposed or shadowed areas due to increased brightness, detail enhancement based on local contrast constraints is performed to restore texture information.

[0051] Output RGB color data with consistent lighting after global adjustment and local optimization.

[0052] This invention establishes a multi-source data adaptive correction system, which significantly improves the reliability of detection under complex working conditions. Through technologies such as ambient temperature compensation, emissivity correction, and adaptive illumination adjustment, it effectively eliminates the interference of environmental factors on the detection data, ensuring the consistency and accuracy of detection results under different construction conditions.

[0053] The sleeve positioning module identifies the three-dimensional coordinates of the grouting hole and the grout outlet hole of each sleeve based on the time-series three-dimensional data field, and generates the corresponding three-dimensional bounding box of the sleeve with the line connecting the centers of the two holes as the axis, combined with the sleeve design parameters and the measured spatial posture.

[0054] In a preferred embodiment of the present invention, the generation of the sleeve three-dimensional bounding box includes: based on the sleeve design parameters, defining a cylindrical initial analysis domain coaxial with the design axis for each sleeve in a time-seriesd three-dimensional data field, wherein the cross-sectional diameter is increased with a dimensional margin based on the design diameter, and the length covers the entire design length of the sleeve.

[0055] Within the initial analysis domain, surface regions are segmented based on the distribution of point cloud normal vectors, and end planes whose angle between the normal vector and the design axis is within a preset threshold range are identified.

[0056] The circular edge of the hole is detected in the end plane and the three-dimensional coordinates of the center are calculated. Based on the vertical height difference of the center, the upper center is determined to be the center of the grout outlet hole and the lower center is determined to be the center of the grouting hole.

[0057] Using the line connecting the centers of the two holes as the actual spatial axis of the sleeve, a three-dimensional bounding box is generated with a cross-sectional dimension based on the designed outer diameter plus a fixed allowance, a height equal to the actual vertical distance between the two holes, and a fixed length extending bidirectionally along the axis.

[0058] Reference Figure 3 As shown, the feature extraction module extracts visible light image features, infrared thermal imaging features, and three-dimensional point cloud geometric features related to grout fullness in parallel within the three-dimensional bounding box of the sleeve, forming a multi-physics evidence chain.

[0059] In a preferred embodiment of the present invention, the visible light image feature extraction includes: dividing an annular region with a preset radius around the center of the slurry outlet hole; using the Euclidean distance between the pixel color and the color of the components around the hole in RGB space as the growth criterion; growing from the seed pixel at the edge of the hole; and aggregating connected pixels with similar color features but different from the component colors as potential slurry overflow areas.

[0060] It should be noted that the above-mentioned connected pixels with similar color features but different from the component colors specifically include: during the region growing process, calculating the Euclidean distance between the average color of the current pixel to be examined and the seed pixel region in the RGB color space, as the first color difference.

[0061] Calculate the Euclidean distance between the current pixel under investigation and the average color of the material of the predefined surrounding components of the aperture, and use it as the second color difference.

[0062] The merging condition for region growth is set as follows: the first color difference is significantly less than the second color difference, so as to ensure that the aggregated pixels are closer to the grown slurry area in the color space and farther away from the surface of the component body. The condition of being significantly less can be defined by example as follows: the first color difference must be less than 1.5 times the standard deviation of the color difference inside the slurry area, and the second color difference must be greater than 1.2 times the average color difference between the component body and the boundary of the slurry area.

[0063] Based on the distance between the centroid of the potential overflow area and the center of the slurry outlet, determine whether it is located in the orifice area.

[0064] It should be noted that the logic for determining whether a region is located in the orifice area is as follows: calculate the two-dimensional plane distance between the centroid coordinates of the pixel set of the potential overflow area and the center coordinates of the slurry outlet.

[0065] If the distance in the two-dimensional plane is less than or equal to the preset radius, the potential overflow area is determined to be located in the orifice area; otherwise, it is determined to be a non-orifice area interference and is excluded.

[0066] For potential grout overflow areas located in the orifice region, perform standard grout sample color histogram matching and texture index quantification to determine the grout overflow status and generate feature labels.

[0067] It should be further clarified that the aforementioned texture metrics specifically refer to the contrast and homogeneity parameters extracted based on the gray-level co-occurrence matrix. The determination of the overflow state employs a fuzzy rule method, achieved by matching different ranges of similarity, contrast, and homogeneity parameters using predefined color histograms.

[0068] The specific discrimination rules are as follows: when the color histogram matching similarity is in a high range, the contrast index is in a medium range, and the homogeneity index is in a high range, the overflow state is judged as normal solidification overflow.

[0069] When the color histogram matching similarity or homogeneity parameter is in an extremely low range, or the contrast index is in an extremely high or extremely low range, the overflow state is judged as an abnormal condition.

[0070] The core basis for the above judgment is as follows: the color histogram matching similarity is used to characterize the consistency of the grout material. The color of normally solidified grout should be highly consistent with the standard sample. A high level range means that the material is pure, the hydration reaction is sufficient, and it is not seriously contaminated or has undergone abnormal chemical changes. If this parameter is significantly low, it may be due to mud contamination, abnormal moisture content, or mixing with foreign matter such as oil.

[0071] Texture contrast is used to characterize the degree of drastic change in grayscale in local areas of an image, i.e., surface roughness. A medium level means that the surface has a natural, uniform, fine texture and graininess after the grout has solidified. The surface is generally smooth but has a healthy microstructure. Extremely high values ​​usually mean that there are drastic changes in brightness on the surface, corresponding to macroscopic cracks, honeycomb pitting, or the edges of holes formed by severe segregation and bleeding. Extremely low values ​​mean that the surface is too smooth and slurry-like, which may be due to a water film surface formed by severe bleeding or improper solidification of the grout.

[0072] Texture homogeneity is used to characterize the uniformity of image texture and the density of local structure. A high level indicates that the grout has a dense and uniform structure after solidification, with no significant voids or loose areas inside. A significantly low level indicates that the texture is messy and uneven, with obvious local variations. This is usually due to internal looseness, shrinkage microcracks, or severely uneven aggregate distribution.

[0073] In a preferred embodiment of the present invention, the infrared thermal imaging feature extraction includes: dividing the time sequence nodes into the grouting implementation stage and the curing stage, with the end of grouting as the boundary.

[0074] During the grouting implementation phase, sampling points were set along the actual axis of the sleeve, and the radial temperature change rate was calculated based on the finite difference method to verify whether the heat conduction gradient conforms to the rule of decreasing from the inside to the outside.

[0075] It should be noted that the data condition for the above heat conduction gradient to conform to the rule of decreasing from the inside to the outside is that the direction of the radial temperature gradient vector is consistent from the inside of the sleeve to the outside, and its value is negative.

[0076] During the grouting and curing stage, the infrared temperature distribution inside the three-dimensional enclosure of the sleeve is compared with the temperature of adjacent ungrouted component areas. By setting temperature difference thresholds and analyzing connected regions, it is verified whether there are continuous local low temperature anomalies.

[0077] It should be noted that the verification of whether there is a continuous local low temperature anomaly zone is achieved through the following process: at multiple consecutive time nodes in the grouting and curing stage, the difference between the infrared temperature value of each voxel in the three-dimensional bounding box of the sleeve and the average temperature of the adjacent ungrouted component area is calculated to obtain the temperature difference distribution of each node.

[0078] For each time-series node, regions with temperature differences below a preset negative threshold are marked as candidate low-temperature regions.

[0079] For each candidate low-temperature region, perform connected component analysis to identify spatially continuous low-temperature blocks and record their spatial location, area, and average temperature difference constants.

[0080] The spatial location and area of ​​the cryogenic block are tracked across multiple consecutive time nodes.

[0081] If low-temperature blocks exist at the same spatial location for a consecutive preset number of time-series nodes, and the standard deviation of their area and temperature difference constant value meets the stability standard, then the region is determined to be a continuous local low-temperature anomaly region.

[0082] The verification results of the grouting implementation stage and the grouting curing stage were used as infrared thermal imaging features.

[0083] It should be added that the above-mentioned heat conduction gradient verification logic is based on the following explanation: Under normal grouting conditions, the heat generated by the hydration of the grout is conducted from the high-temperature zone inside the sleeve to the low-temperature zone of the external concrete, inevitably forming a temperature gradient that decreases from the inside to the outside. By verifying the consistency of the gradient direction and the negative value characteristic, the normality of the heat conduction process can be directly determined. If this rule is violated, it indicates an abnormal internal heat source or a defect in insulation.

[0084] The verification of the continuous low temperature anomaly zone adopts a dual temporal and spatial verification logic. By comparing with the normal area and setting a threshold, the low temperature area with statistical significance is identified. Through continuous tracking of multiple time-series nodes, the influence of instantaneous and random temperature fluctuations is eliminated. Areas with internal voids or insufficient filling will exhibit continuous and stable low temperature characteristics during the curing stage because their heat capacity and thermal conductivity are lower than those of normal grouting material and concrete. This spatiotemporal continuity is the key criterion that distinguishes it from temporary temperature unevenness.

[0085] In a preferred embodiment of the present invention, the three-dimensional point cloud geometric feature extraction includes: during the grouting and curing stage, dividing the three-dimensional bounding box of the sleeve into a voxel grid, counting the number of point clouds contained in each voxel unit to analyze the local density, and identifying and marking low-density areas where the point cloud density is lower than a preset density threshold.

[0086] Geometric morphology analysis was performed on the three-dimensional point cloud of the slurry outlet region. The reference plane was fitted by the least squares method, the distance from each three-dimensional point to the reference plane was calculated, the average negative deviation and the maximum negative deviation were statistically analyzed, and significant depressions were identified and marked.

[0087] It should be noted that the above-mentioned significant indentation meets the requirement that the average negative deviation exceeds the preset percentage of the sleeve's designed outer diameter, and the maximum negative deviation exceeds the allowable orifice flatness tolerance in the sleeve's design drawings.

[0088] Low-density regions and significant depressions are used as geometric features of the 3D point cloud.

[0089] The defect identification module performs logical reasoning on the multi-physics evidence chain to identify the type of grouting defect and associate its time sequence nodes to trace the cause of the defect.

[0090] In a preferred embodiment of the present invention, the grouting defect type identification includes: if the following conditions are met simultaneously: abnormal grout overflow at the orifice, the heat conduction gradient during the grouting implementation stage does not conform to the rule of decreasing from the inside to the outside, and the grout outlet area is low density or has significant depressions, then it is identified as a grouting insufficiency void defect.

[0091] If the following conditions are met simultaneously: normal grout overflow at the orifice, a continuous local low temperature abnormality zone during the grouting and curing stage, and settlement on the surface of the corresponding component, then it is identified as a late-stage shrinkage void defect.

[0092] During the grouting implementation phase, the point cloud is scanned upwards from the grouting hole to retrieve the lowest horizontal plane with continuous local density higher than the preset threshold. Its vertical coordinates are used as the actual filling height. If it is lower than the minimum full depth specified in the design, it is identified as a defect of insufficient vertical grouting depth.

[0093] This invention overcomes the limitations of existing single-dimensional detection by fusing and analyzing evidence chains from multiple physics fields. By extracting visible light image features, infrared thermal imaging features, and three-dimensional point cloud geometric features in parallel, it achieves a comprehensive internal and external diagnosis of the grouting fullness state. It can accurately identify hidden problems that existing methods cannot detect, such as insufficient grouting cavities, late-stage shrinkage cavities, and insufficient vertical grouting depth.

[0094] It should be noted that the above grouting defect type identification logic is based on: (1) the grouting insufficiency void defect identification logic, which revolves around the core mechanism of insufficient grouting process leading to incomplete internal filling. Its logic chain is as follows: During normal grouting, the grout should fill the inside of the sleeve under pressure and overflow naturally from the grout outlet hole, forming an overflow area with color and texture that meets the standard, indicating that the grout is fully filled and the pressure transmission is normal. If the overflow state at the hole opening is abnormal, the essence is that the internal pressure has not been sufficient to push the grout to the hole opening position, which directly points to the preliminary judgment of insufficient grout filling.

[0095] Meanwhile, the heat released by the hydration reaction of the grout forms a stable heat conduction gradient that decreases from the inside to the outside of the sleeve under normal filling conditions. When there are cavities due to insufficient grouting, the cavity area lacks an effective hydration heat source, and the thermal conductivity of air is much lower than that of dense grout, causing heat to be unable to conduct normally along the preset path, resulting in disorder or abnormality of the heat conduction gradient. This provides indirect thermal evidence for insufficient internal filling.

[0096] Furthermore, the grout outlet, as the final verification point of grout filling, should, under normal conditions, be completely filled and solidified by the grout, maintaining the same density as the surrounding grout and having a smooth surface. If the point cloud density in this area is lower than a preset threshold or there are significant depressions, it constitutes direct morphological evidence of insufficient grouting. These three abnormal features corroborate each other, jointly confirming the definitive conclusion that there are unfilled cavities inside.

[0097] (2) The logic for identifying shrinkage voids in the later stage focuses on the formation mechanism of internal voids caused by material shrinkage during the curing stage when the grouting process is normal: the normal overflow state of the grout at the orifice indicates that the grout has completely filled the interior of the component and the pressure has been transmitted in place during the grouting implementation stage, which eliminates the defects caused by insufficient construction process. This establishes the premise that the defects are attributed to changes in the later stage, that is, the defects are caused by shrinkage after filling rather than not filling.

[0098] During the curing phase, the grout continues to hydrate and the structure gradually densifies. Under normal conditions, the temperature distribution in different areas of the component is relatively uniform. If a persistent localized low-temperature zone appears, the core cause is the formation of voids in that area. Because the thermal conductivity of air is much lower than that of the cured grout, it cannot effectively transfer the heat generated by the surrounding hydration, causing the temperature in that area to remain lower than that of the normal grout area, forming a characteristic low-temperature anomaly zone. This is a typical thermal characteristic of voids formed by later shrinkage.

[0099] Meanwhile, the volume shrinkage during the curing process of the grout is controllable within the normal range and will not cause surface settlement. If surface settlement is detected, it is essentially due to the internal grout shrinking beyond the allowable range, forming voids, and the surface material losing internal support, resulting in indentation deformation under the component's own weight or external micro-pressure. This characteristic provides external morphological evidence for internal shrinkage voids, and together with the aforementioned characteristics, constitutes a complete chain of evidence for defects caused by later shrinkage.

[0100] (3) The logic for identifying insufficient vertical depth of grouting is based on the core quantitative indicator of grout fullness: the vertical filling height. As a dense fluid, the grout maintains a stable density that is significantly higher than that of air after solidification. When scanning the point cloud upward from the grouting hole, the continuous areas with a density higher than the preset threshold correspond to spaces filled with dense grout, and the lowest horizontal plane of this density interface accurately marks the filling height of the grout in the vertical direction.

[0101] The minimum filling depth specified in the design is the lowest effective filling standard determined based on the structural performance and waterproofing requirements of the component. Only when the actual filling height reaches this critical value can the grout perform its expected bonding, anchoring, and force transmission functions. If the actual filling height obtained through point cloud retrieval is lower than this design value, it clearly indicates insufficient filling in the vertical direction, directly constituting a defect judgment of insufficient vertical grouting depth. This judgment is based on the quantitative detection of the physical density of the point cloud, forming a direct and accurate logical closed loop.

[0102] In a preferred embodiment of the present invention, the association of time-series nodes to trace the cause of defects includes: establishing a mapping relationship between defect types and time-series nodes, and binding the defect to the time-series node where it was first detected or continues to exist.

[0103] Based on the temporal three-dimensional data field, the evolution process of the multi-physics evidence chain of defects at the corresponding nodes is traced back.

[0104] Based on the type of defect and the sequence of its occurrence, the cause category is determined in conjunction with the construction process: if it first appears during the grouting implementation stage, it is attributed to improper operation.

[0105] If the phenomenon occurs during the grouting and curing stage and its characteristics evolve, it is attributed to material shrinkage or environmental factors.

[0106] The report output module performs risk assessment based on the type and timing of grouting defects and outputs a structured inspection report.

[0107] In a preferred embodiment of the present invention, the risk assessment includes: predefined grouting defect types, occurrence time nodes, and multi-dimensional weights of defect spatial distribution range.

[0108] It should be added that the above multidimensional weights can be determined based on expert experience or the analytic hierarchy process. For example, the basic weight of defect type accounts for 50%, the weight of temporal development accounts for 30%, and the weight of spatial regionality accounts for 20%.

[0109] Based on the identified grouting defect types and their weights, the development of defects is judged by their time sequence and the regional impact is judged by their spatial distribution range, and a multi-dimensional comprehensive score is obtained.

[0110] It should be noted that the temporal judgment of defect development is specifically achieved through the following process: based on the temporalized three-dimensional data field, the feature evolution sequence of the same defect at different temporal nodes is extracted.

[0111] Time series analysis is performed on the feature evolution sequence to calculate its trend index: I. The slope of the defect feature change is calculated using linear regression. If the slope is positive and greater than the preset critical slope, it is determined to be a rapidly developing defect.

[0112] Ii. Calculate the relative rate of change of defect features between adjacent time-series nodes. If the rate of change of multiple consecutive nodes exceeds a fixed percentage of the preset critical slope, it is determined to be a continuously deteriorating defect.

[0113] II. When a defect persists for three or more consecutive time nodes and its characteristic intensity increases monotonically, it is determined to be a stable development defect.

[0114] The three levels of defect temporal development are assigned different temporal development coefficient values, with the values ​​assigned in ascending order as stable development defects, rapid development defects, and continuously deteriorating defects.

[0115] The determination of the spatial distribution range affects the region, which is achieved through the following process: Based on the three-dimensional bounding box of the sleeve, the spatial distribution characteristics of the defect area are calculated: the percentage of the defect area to the total volume of the sleeve is calculated, and the influence range level is divided according to the preset percentage range. The influence range level includes local influence, regional influence and overall influence level. The percentage of the defect area to the total volume of the sleeve is defined as follows: local influence can be defined as less than 5%, regional influence as 5%-15%, and overall influence as greater than 15%.

[0116] Analyze the spatial distribution of defects and their positional relationship with the critical stress areas of the sleeve. If a defect is located on the main stress path, it should be classified as at least a regional impact level.

[0117] Different spatial regional coefficient values ​​are assigned to the three levels of regional impact of defects, with the spatial regional coefficient values ​​arranged in ascending order as local impact, regional impact, and overall impact levels.

[0118] The multi-dimensional comprehensive score can be obtained by weighting and fusing the spatial regional coefficient value and the temporal development coefficient value with their corresponding weights, and then summing the fusion result with the weight matching the grouting defect type.

[0119] Based on a multi-dimensional comprehensive score, the defect risk level is classified into urgent, important, or general levels.

[0120] It should be noted that the above-mentioned defect risk level classification can be specifically divided using a fixed threshold method: the multi-dimensional comprehensive score is normalized, and when the multi-dimensional comprehensive score is greater than or equal to 0.8, the defect risk level is determined to be urgent; when it is greater than or equal to 0.5 but less than 0.8, the defect risk level is determined to be important; and when it is less than 0.5, the defect risk level is determined to be moderate.

[0121] The structured inspection report outputs complete assessment information including defect type, risk level, occurrence sequence, spatial location, and cause analysis.

[0122] It should also be added that, based on the determined risk level, the system automatically executes the corresponding handling strategy: for emergency level defects, an alarm is pushed to the construction monitoring center in real time and the emergency response process is initiated.

[0123] For critical defects, a specific handling plan is generated and rectification is required within a specified timeframe.

[0124] For defects of a general level, incorporate them into the routine maintenance plan and monitor them continuously.

[0125] This invention enables dynamic monitoring of the entire grouting process by creating a time-series three-dimensional data field construction mechanism. It continuously collects and registers data at key grouting time nodes, establishes a spatiotemporal evolution analysis of the grouting state, solves the limitations of the process blind zone of static detection methods, and supports accurate tracing of defect formation mechanisms and determination of causes.

[0126] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. An automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs), characterized in that, include: The data acquisition module controls a drone equipped with multimodal sensors to perform synchronous data acquisition along a preset path at key grouting timing nodes at the connection points of prefabricated building components, thereby acquiring multi-source heterogeneous data with time-series markers. The data registration module performs fusion registration on the multi-source heterogeneous data based on a spatiotemporal synchronization strategy to construct a temporal three-dimensional data field; The sleeve positioning module identifies the three-dimensional coordinates of the grouting hole and the grout outlet hole of each sleeve based on the time-series three-dimensional data field, and generates the corresponding three-dimensional bounding box of the sleeve with the line connecting the centers of the two holes as the axis, combined with the sleeve design parameters and the measured spatial posture. The feature extraction module extracts visible light image features, infrared thermal imaging features, and three-dimensional point cloud geometric features related to grout fullness in parallel within the three-dimensional bounding box of the sleeve, forming a multi-physics evidence chain. The defect identification module performs logical reasoning on the multi-physics evidence chain to identify the type of grouting defect and associate its time sequence nodes to trace the cause of the defect. The report output module performs risk assessment based on the type and timing of grouting defects and outputs a structured inspection report. The construction of the temporal three-dimensional data field includes: using the lidar point cloud of each temporal node as a spatial reference, projecting the visible light and infrared image pixel coordinates at the corresponding time onto the point cloud coordinate system through camera intrinsic and extrinsic parameters and a fixed transformation relationship with lidar, and assigning RGB color value and infrared temperature value to each three-dimensional point; taking the point cloud of the previous temporal node as the target point cloud and the point cloud of the current temporal node as the source point cloud from the projected multi-period point cloud, determining the corresponding point pair for each point in the source point cloud through nearest neighbor search in the target point cloud; calculating a rigid transformation matrix that minimizes the mean square error based on all corresponding point pairs through singular value decomposition; applying the transformation matrix to align the source point cloud to the target point cloud, iteratively performing nearest neighbor search, transformation calculation, and point cloud alignment until the mean square error change is lower than a preset convergence threshold, forming a temporal three-dimensional data field with consistent spatial location and fused RGB color value, infrared temperature value, and three-dimensional information; The generation of the three-dimensional bounding box for the sleeve includes: based on the sleeve design parameters, defining a cylindrical initial analysis domain coaxial with the design axis for each sleeve in a time-seriesd three-dimensional data field, with its cross-sectional diameter increased by a dimensional allowance based on the design diameter, and its length covering the entire design length of the sleeve; within the initial analysis domain, performing surface region segmentation based on the distribution of point cloud normal vectors, and identifying end planes where the angle between the normal vector and the design axis is within a preset threshold range; detecting the circular edge of the hole in the end plane and calculating the three-dimensional coordinates of the center, determining the upper center as the center of the grout outlet hole and the lower center as the center of the grouting hole based on the vertical height difference between the centers; using the line connecting the centers of the two holes as the actual spatial axis of the sleeve, generating a three-dimensional bounding box with a cross-sectional size increased by a fixed allowance based on the design outer diameter, a height equal to the actual vertical distance between the two holes, and extending a fixed length in both directions along the axis.

2. The automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, Before assigning RGB color and infrared temperature values ​​to each 3D point, the image undergoes environmental preprocessing: Based on the synchronously acquired ambient temperature and the measured temperature at the reference point of the component connection, the infrared thermal image is subjected to emissivity correction and ambient background radiation compensation to obtain infrared data that reflects the true temperature distribution. Based on the brightness value provided by the ambient light intensity sensor, adaptive brightness correction is performed on the visible light image to obtain RGB color data with consistent illumination.

3. The automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The visible light image feature extraction includes: A ring-shaped area with a preset radius is divided with the center of the slurry outlet as the center. The growth criterion is the Euclidean distance between the pixel color and the color of the components around the outlet in RGB space. The growth starts from the seed pixel at the edge of the outlet and aggregates connected pixels with similar color features but different from the component colors as potential overflow areas. Based on the distance between the centroid of the potential overflow area and the center of the slurry outlet, determine whether it is located in the orifice area; For potential grout overflow areas located in the orifice region, perform standard grout sample color histogram matching and texture index quantification to determine the grout overflow status and generate feature labels.

4. The automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs) according to claim 3, characterized in that, The infrared thermal imaging feature extraction includes: The timeline is divided into the grouting implementation stage and the curing stage, with the completion of grouting as the dividing line. During the grouting implementation phase, sampling points were set along the actual axis of the sleeve, and the radial temperature change rate was calculated based on the finite difference method to verify whether the heat conduction gradient conforms to the rule of decreasing from the inside to the outside. During the grouting and curing stage, the infrared temperature distribution inside the three-dimensional enclosure of the sleeve is compared with the temperature of adjacent ungrouted component areas. By setting temperature difference thresholds and analyzing connected components, it is verified whether there are continuous local low temperature anomaly areas. The verification results of the grouting implementation stage and the grouting curing stage were used as infrared thermal imaging features.

5. The automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs) according to claim 4, characterized in that, The extraction of geometric features from the three-dimensional point cloud includes: During the grouting and curing stage, the three-dimensional bounding box of the sleeve is divided into a voxel mesh. The number of point clouds contained in each voxel unit is counted to analyze the local density, and low-density areas with point cloud density below a preset density threshold are identified and marked. Geometric morphology analysis was performed on the three-dimensional point cloud of the slurry outlet region. The reference plane was fitted by the least squares method, the distance from each three-dimensional point to the reference plane was calculated, the average negative deviation and the maximum negative deviation were statistically analyzed, and significant depressions were identified and marked. Low-density regions and significant depressions are used as geometric features of the 3D point cloud.

6. The automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs) according to claim 5, characterized in that, The identification of grouting defect types includes: If the following conditions are met simultaneously: abnormal grout overflow at the orifice, the heat conduction gradient during the grouting implementation stage does not conform to the rule of decreasing from the inside to the outside, and the grout outlet area is low-density or has significant depressions, then it is identified as a grouting insufficiency void defect. If the following conditions are met simultaneously: the grout overflow at the orifice is normal, there is a continuous local low temperature abnormality zone during the grouting and curing stage, and there is settlement on the surface of the corresponding component, then it is identified as a late-stage shrinkage void defect. During the grouting implementation phase, the point cloud is scanned upwards from the grouting hole to retrieve the lowest horizontal plane with continuous local density higher than the preset threshold. Its vertical coordinates are used as the actual filling height. If it is lower than the minimum full depth specified in the design, it is identified as a defect of insufficient vertical grouting depth.

7. The automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The association of its time-series nodes to trace the cause of defects includes: Establish a mapping relationship between defect types and timing nodes, and bind defects to their first detection or persistent timing nodes; Based on the temporal three-dimensional data field, the evolution process of the multi-physics evidence chain of defects at the corresponding nodes is traced back; The cause category is determined based on the defect type and its occurrence sequence, combined with the construction process: If the problem first occurs during the grouting process, it is attributed to improper operation. If the phenomenon occurs during the grouting and curing stage and its characteristics evolve, it is attributed to material shrinkage or environmental factors.

8. The automatic defect identification system for prefabricated buildings based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The risk assessment includes: Predefined grouting defect types, occurrence time nodes, spatial distribution range of defects, and multi-dimensional weights of defect severity; Based on the identified grouting defect types, weights are matched, and the temporal development and spatial distribution range of the defects are combined to determine the regional impact, thereby obtaining a multi-dimensional comprehensive score. Based on a multi-dimensional comprehensive score, the defect risk level is divided into urgent, important, or general levels; The structured inspection report outputs complete assessment information including defect type, risk level, occurrence sequence, spatial location, and cause analysis.

Images