A plain concrete defect detection method and system based on multi-source data fusion

Abstract

The application discloses a kind of based on multi-source data fusion's plain concrete defect detection method and system, it is related to construction defect detection technical field.Full-process construction process parameters, surface high-definition image data and infrared thermal imaging data of plain concrete are respectively collected;Surface high-definition image data is sequentially preprocessed and panoramic splicing, and the complete three-dimensional appearance form of plain concrete is reconstructed to extract appearance form features;Infrared thermal imaging data is extracted, and infrared feature extraction is completed by combining acousto-optic thermal multi-physical field coupling analysis, to obtain infrared imaging features;Call defect risk early warning database based on industry standard topology structure, and the processed data features are compared with the standard topology data in database Multidimensional correlation mapping and similarity;Based on fusion comparison result, defect analysis is carried out, plain concrete defect identification result is generated and rectification suggestions are provided.The application can guarantee the quality of concrete construction and structural service safety.

Description

Technical Field

[0001] This invention relates to the field of construction defect detection technology, and more specifically to a method and system for detecting defects in plain concrete based on multi-source data fusion. Background Technology

[0002] Plain concrete, with its advantages of being internally strong and externally beautiful, having a uniform appearance, and being low-carbon and environmentally friendly, has been widely used in underground structural engineering projects such as subway stations. However, during the entire process of pouring, vibration, and curing, plain concrete in underground structures is prone to hidden defects such as voids, looseness, cracks, and leakage due to factors such as mix ratio fluctuations, uneven vibration, temperature stress, and improper treatment of construction joints. Traditional testing methods are difficult to achieve accurate, efficient, and non-destructive identification.

[0003] Current methods for detecting concrete defects often rely on single approaches: surface image processing can only identify superficial problems such as honeycomb, pitting, and color difference, but cannot determine density and damage; infrared thermal imaging can detect defects, but it is easily affected by ambient temperature, curing humidity, and heat dissipation from the formwork, limiting its detection accuracy; at the same time, construction process parameters (such as vibration time, placement temperature, curing cycle, and formwork positioning accuracy) are highly correlated with defect formation, and existing technologies do not integrate multi-source data, making it difficult to establish a quantitative correlation between defects and their causes, resulting in high misjudgment rates and difficulties in tracing the source, and failing to meet the full-cycle quality control requirements for high-quality plain concrete in underground structures.

[0004] To address the aforementioned issues, there is an urgent need for a multi-source data detection method that integrates infrared imaging, image processing, and process parameters to achieve precise defect location, quantitative identification, and rapid tracing of causes, thereby ensuring the safety and appearance quality of plain concrete structures. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for detecting defects in plain concrete based on multi-source data fusion, in order to solve the problems existing in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for detecting defects in plain concrete based on multi-source data fusion includes the following steps: The construction process parameters, high-definition surface image data, and infrared thermal imaging data of the plain concrete were collected separately. The high-definition image data of the surface is preprocessed and stitched together in sequence to reconstruct the complete three-dimensional appearance of the plain concrete and extract its appearance features, thereby completing the identification and quantification of surface defects. The construction process parameters are thresholded and normalized, and the infrared thermal imaging data are subjected to temperature field calculation, abnormal region segmentation and feature extraction. The infrared feature extraction is completed by combining acoustic-optical-thermal multi-physics field coupling analysis to obtain infrared imaging features. The defect risk warning database based on the industry standard topology is retrieved, and the processed construction process parameters, appearance features, and infrared thermal imaging features are correlated and mapped with the standard topology data in the database in multiple dimensions and compared for similarity. Based on the fusion comparison results, defect analysis is performed to generate defect identification results for plain concrete and provide rectification suggestions.

[0007] Optionally, the construction process parameters include concrete mix proportion parameters, pouring parameters, vibration parameters, curing parameters, formwork parameters, and environmental parameters; the construction process parameters are thresholded, normalized, and coded, specifically including: comparing each process parameter with the design threshold to obtain the process deviation value, normalizing the deviation value to a process compliance feature in the range of 0~1; sorting the process compliance features according to the concrete construction sequence and spatial location to generate a process coded data sequence representing the compliance of the entire construction process.

[0008] Optionally, the construction process parameters are used for defect authenticity determination: the Euclidean distance between the process compliance characteristics and the standard process characteristics is calculated, and the normalized reciprocal is used as the process consistency index; the spatial location of the thermal anomaly area obtained by infrared thermal imaging, the surface defect area identified by the appearance morphology model, and the process deviation area are matched; only when the thermal anomaly, surface defect, and process deviation spatially overlap are they determined to be real defects, excluding false defects caused by environmental interference.

[0009] Optionally, the panoramic stitching process involves: using optical flow motion detection to identify dynamic targets within the overlapping areas of the images and generating a dynamic mask; employing Laplacian pyramid multi-band fusion based on image gradients for static areas and a single-source priority strategy to avoid ghosting for dynamic areas; and combining adaptive exposure equalization to adjust the fusion weights, eliminating stitching seams and brightness discontinuities to complete the panoramic image reconstruction.

[0010] Optionally, the temperature field calculation of infrared thermal imaging data includes the following steps: Based on single-frame infrared thermal imaging data and concrete thermal properties, a virtual continuous infrared thermal imaging sequence is generated using the Green's function of the heat conduction control equation. Gaussian filtering, Fourier-Hankel transform, and cylindrical harmonic function inversion are performed on the thermal imaging sequence, and the inverse transform is performed back to the spatiotemporal domain to generate a virtual heat flow image sequence. Binarization and sigmoid discriminant function enhancement are performed on the virtual heat flow image to obtain an enhanced binarized image of thermal anomalies.

[0011] Optional, the acousto-optic-thermal multiphysics coupling analysis specifically includes: Simultaneously apply acoustic and thermal excitation to the defect candidate region segmented by infrared thermal imaging, and collect the pixel temperature change sequence. Discrete Fourier transform and digital phase-locked loop analysis were performed on the temperature change sequence to calculate the thermo-acoustic coupling parameters and resonance intensity parameters; The defect type is determined by combining geometric reference parameters, and the void depth and bonding stiffness are inverted based on the resonance peak characteristics.

[0012] Optionally, surface defect identification and quantification specifically involves: performing histogram equalization, sharpening, and noise reduction enhancement on the panoramic image; dividing the defect detection sub-region according to texture complexity; identifying defects using grayscale abrupt changes and gradient features; extracting defect shape and size features; and determining repairability.

[0013] A defect detection system for plain concrete based on multi-source data fusion includes: Correlation data acquisition module: used to collect construction process parameters, high-definition surface image data and infrared thermal imaging data of plain concrete throughout the entire process; Image processing and recognition module: used to preprocess and stitch together high-definition surface image data sequentially, reconstruct the complete three-dimensional appearance of the plain concrete surface, extract appearance features, and complete surface defect identification and quantification. Process parameters and infrared feature processing module: used to threshold and normalize construction process parameters, calculate temperature field, segment abnormal regions and extract features from infrared thermal imaging data, and complete infrared feature extraction by combining acoustic-optical-thermal multi-physics field coupling analysis to obtain infrared imaging features; Similarity comparison module: retrieves a defect risk warning database based on industry standard topology structure, and performs multi-dimensional association mapping and similarity comparison between the processed construction process parameters, appearance features, and infrared thermal imaging features and the standard topology data in the database; Defect Analysis Module: Used to perform defect analysis based on fusion comparison results, generate defect identification results for plain concrete and provide rectification suggestions.

[0014] As can be seen from the above technical solution, compared with the prior art, the present invention provides a method and system for detecting defects in plain concrete based on multi-source data fusion. By integrating construction process parameters, high-definition surface image data, and infrared thermal imaging multi-source data, combined with image preprocessing panoramic stitching, three-dimensional appearance reconstruction, temperature field calculation, and multi-physics field coupling analysis, it can comprehensively mine concrete defect information from multiple dimensions such as construction process, appearance, and internal temperature anomalies, breaking through the limitations of single detection methods in terms of incomplete coverage and limited identification accuracy. Relying on the industry standard topological structure defect risk warning database, it carries out multi-dimensional correlation mapping and similarity comparison to achieve intelligent and accurate defect identification, quantitative classification, and risk warning. At the same time, it can automatically generate targeted rectification suggestions, improve the comprehensiveness, accuracy, and efficiency of plain concrete defect detection, and ensure the quality of concrete construction and the safety of structural service. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the method flow provided by the present invention; Figure 2 This is a schematic diagram of the system structure provided by the present invention. Detailed Implementation

[0017] 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.

[0018] Example 1

[0019] This invention discloses a method for detecting defects in plain concrete based on multi-source data fusion, such as... Figure 1 As shown, it includes the following steps: The construction process parameters, high-definition surface image data, and infrared thermal imaging data of the plain concrete were collected separately. The high-definition image data of the surface is preprocessed and stitched together in sequence to reconstruct the complete three-dimensional appearance of the plain concrete and extract its appearance features, thereby completing the identification and quantification of surface defects. The construction process parameters are thresholded and normalized, and the infrared thermal imaging data are subjected to temperature field calculation, abnormal region segmentation and feature extraction. The infrared feature extraction is completed by combining acoustic-optical-thermal multi-physics field coupling analysis to obtain infrared imaging features. The defect risk warning database based on the industry standard topology is retrieved, and the processed construction process parameters, appearance features, and infrared thermal imaging features are correlated and mapped with the standard topology data in the database in multiple dimensions and compared for similarity. Based on the fusion comparison results, defect analysis is performed to generate defect identification results for plain concrete and provide rectification suggestions.

[0020] Furthermore, the construction process parameters include concrete mix proportion parameters, pouring parameters, vibration parameters, curing parameters, formwork parameters, and environmental parameters. The construction process parameters are then subjected to thresholding, normalization, and coding processing, specifically including: comparing each process parameter with the design threshold to obtain process deviation values; normalizing the deviation values ​​to process compliance characteristics within the 0-1 range; and sorting the process compliance characteristics according to concrete construction procedures and spatial location to generate a process coding data sequence representing the compliance of the entire construction process.

[0021] Furthermore, the association method between the construction process coding data sequence and the three-dimensional appearance model and infrared thermal imaging data is as follows: using the spatial coordinates of the concrete structure as an index, the process coding data sequence, the surface defect quantitative feature sequence, and the infrared thermal anomaly feature sequence are paired point by point to construct a multi-source data pair at the same spatial location; the scaling, rotation, and translation transformation matrices of the coordinates of the entire construction process and the detection spatial coordinates are registered to eliminate spatial deviations between the construction and detection stages and achieve accurate alignment of multi-source data.

[0022] Furthermore, the construction process parameters are used to determine the authenticity of defects: the Euclidean distance between the process compliance characteristics and the standard process characteristics is calculated, and the normalized reciprocal is used as the process consistency index; the spatial location of the thermal anomaly area obtained by infrared thermal imaging, the surface defect area identified by the appearance morphology model, and the process deviation area are matched; only when the thermal anomaly, surface defect, and process deviation spatially overlap are they determined to be real defects, excluding false defects caused by environmental interference.

[0023] Furthermore, the construction process parameters are used for automatic source tracing of defect causes: a process deviation-defect type mapping library is established, and the dominant cause of the defect is matched according to the difference between the normalized process compliance characteristics and the standard characteristics; based on multi-source data, it is used to trace whether surface defects and internal thermal anomalies originate from potential anomalies in the early construction process, forming a construction-defect closed-loop source tracing.

[0024] Furthermore, the construction process parameters are used for the quantitative assessment of defect levels: the process consistency index, surface defect quantitative index, and infrared thermal anomaly intensity index are weighted and fused to obtain a comprehensive quality score for the corresponding spatial location; areas with a comprehensive quality score lower than a preset score threshold are marked as construction process abnormal areas, and based on this, four levels of defect risk levels—minor, general, severe, and critical—are classified.

[0025] Furthermore, the adaptive distortion correction specifically involves: establishing a distortion correction lookup table based on a pre-calibrated camera intrinsic parameter matrix and distortion coefficients, performing distortion removal processing on the high-definition surface image data, and correcting barrel distortion and edge distortion in the image.

[0026] Furthermore, the multi-band panoramic stitching based on image gradient specifically involves: using optical flow motion detection to identify dynamic targets within the overlapping areas of the images and generating a dynamic mask; employing Laplacian pyramid multi-band fusion based on image gradient in static areas and a single-source priority strategy to avoid ghosting in dynamic areas; and combining adaptive exposure equalization to adjust the fusion weights to eliminate stitching seams and brightness discontinuities, thereby completing the panoramic image reconstruction.

[0027] Furthermore, the infrared thermal imaging data temperature field calculation and virtual heat flow reconstruction specifically involve: generating a virtual continuous infrared thermal imaging sequence based on single-frame infrared thermal imaging data and concrete thermal property parameters through the Green's function of the heat conduction control equation; performing Gaussian filtering, Fourier-Hankel transform, and cylindrical harmonic function inversion on the thermal imaging sequence, and inversely transforming it back to the spatiotemporal domain to generate a virtual heat flow image sequence; and performing binarization and sigmoid discriminant function enhancement on the virtual heat flow image to obtain a thermal anomaly enhanced binarized image.

[0028] Furthermore, this embodiment also includes: using a deep learning object detection network to identify non-concrete structural interference objects in the high-resolution surface image; and removing interference areas in the enhanced binary image of thermal anomalies through dual-light image registration and coordinate mapping to obtain a clean thermal anomaly image.

[0029] and Figure 1 Corresponding to the method shown, this invention also discloses an implementation of a plain concrete defect detection system based on multi-source data fusion, the specific structure of which is as follows: Figure 2 As shown, it includes: Correlation data acquisition module: used to collect construction process parameters, high-definition surface image data and infrared thermal imaging data of plain concrete throughout the entire process; Image processing and recognition module: used to preprocess and stitch together high-definition surface image data sequentially, reconstruct the complete three-dimensional appearance of the plain concrete surface, extract appearance features, and complete surface defect identification and quantification. Process parameters and infrared feature processing module: used to threshold and normalize construction process parameters, calculate temperature field, segment abnormal regions and extract features from infrared thermal imaging data, and complete infrared feature extraction by combining acoustic-optical-thermal multi-physics field coupling analysis to obtain infrared imaging features; Similarity comparison module: retrieves a defect risk warning database based on industry standard topology structure, and performs multi-dimensional association mapping and similarity comparison between the processed construction process parameters, appearance features, and infrared thermal imaging features and the standard topology data in the database; Defect Analysis Module: Used to perform defect analysis based on fusion comparison results, generate defect identification results for plain concrete and provide rectification suggestions.

[0030] Example 2

[0031] This embodiment selects a cast-in-place plain concrete shear wall with a strength grade of C40 as the inspection object. The wall surface is 8.0m × 3.0m in size and has no decorative layer or covering. The plain concrete defect detection method based on multi-source data fusion described in this invention is used for inspection. First, the concrete wall surface is divided into spatial sections of 0.5m × 0.5m. Simultaneously, the full-process construction process parameters, high-definition surface image data, and infrared thermal imaging data corresponding to each spatial section are collected. The construction process parameters include mix proportion parameters, pouring parameters, vibration parameters, curing parameters, and environmental parameters. Each process parameter is compared with a preset threshold to obtain the process deviation value. The deviation value is normalized to a process compliance feature in the range of 0 to 1 and arranged according to the position order of the spatial sections to generate a process code data sequence characterizing the compliance of the entire construction process. High-resolution surface images are acquired by a drone equipped with a 4K visible light camera along a preset path, with an overlap rate of no less than 30% between adjacent images. Infrared thermal imaging data are acquired by an uncooled infrared thermal imager within a determined optimal detection time window, with a temperature measurement accuracy of no less than ±0.5℃ and an image resolution of 640×512.

[0032] The high-definition surface images acquired are sequentially subjected to grayscale correction and noise suppression. A distortion correction lookup table is established based on the pre-calibrated camera intrinsic parameter matrix and distortion coefficients to complete adaptive distortion correction. Edge detection and feature point extraction are performed on the corrected images. The stitching priority is determined based on the number of feature points in the overlapping area, the proportion of feature point types, and the difference in image clarity.

[0033] The Lucas-Kanade optical flow method is used to dynamically identify targets in the overlapping areas of adjacent images, generating dynamic target masks. Static areas are processed using a Laplacian pyramid multi-band fusion method based on image gradients, while dynamic areas are processed using a single-source priority strategy to avoid motion ghosting. At the same time, adaptive exposure equalization is used to adjust the fusion weights to eliminate stitching seams and brightness discontinuities. Finally, panoramic image stitching and reconstruction are completed to obtain a complete three-dimensional appearance model of plain concrete.

[0034] Based on single-frame infrared thermal imaging data and combined with concrete thermal property parameters, a virtual continuous sequence of infrared thermal imaging images is generated through the Green's function of the heat conduction control equation.

[0035] The infrared thermal imaging sequence is sequentially processed by Gaussian filtering, Fourier-Hankel transform, cylindrical harmonic function inversion, and inverse transform to generate a virtual heat flow image sequence representing the heat flow distribution. The reference frame in the virtual heat flow image sequence is selected for binarization and then enhanced by the sigmoid discriminant function to obtain an enhanced binarized image of the thermal anomaly.

[0036] A deep learning object detection network is used to process visible light images to identify non-concrete structure interference objects. Through dual-light image registration and coordinate mapping, the interference object regions are removed from the thermal anomaly enhanced binarized image to obtain a clean thermal anomaly image.

[0037] Simultaneously apply 100~2000Hz linear sweep frequency acoustic excitation and constant thermal excitation to the defect candidate region obtained by segmenting infrared thermal imaging data, and collect the temperature change sequence corresponding to each pixel.

[0038] Discrete Fourier transform and digital phase-locked loop operation are performed on the temperature change sequence to obtain the temperature spectrum amplitude. The thermo-acoustic coupling parameters and resonance intensity parameters are further calculated. Combined with the geometric reference parameters determined by the three-dimensional appearance model, the defect type is determined. For hollow type defects, the hollow depth and bonding stiffness are obtained by inversion based on the resonance peak characteristics.

[0039] Using spatial segmented coordinates as a unified index, the process coding data sequence, surface defect quantitative feature sequence, and infrared thermal anomaly feature sequence are paired point by point. The scaling, rotation, and translation registration between construction coordinates and detection coordinates are completed through coordinate transformation matrix, eliminating spatial deviations between construction and detection stages and constructing multi-source data pairs corresponding to the same spatial location.

[0040] The process consistency index, surface defect severity index, and thermal anomaly severity index are calculated separately. The three types of indexes are weighted and fused to obtain the comprehensive quality score for the corresponding spatial location. Areas with a comprehensive quality score lower than a preset threshold are marked as construction process abnormal areas and judged as real defect areas. Areas with only surface or infrared abnormalities but compliant process parameters are judged as environmental interference and are removed.

[0041] Finally, based on the fusion judgment results, the location, quantification and qualitative analysis of defects are completed, and quantitative data such as the three-dimensional coordinates of the defect, spatial segment number, defect area, depth, volume, hollow rate, and bond stiffness are output. The defect types are determined to include hollow, surface cracks, internal looseness, pitting and cold joints. According to the process deviation, the dominant cause of the defect is matched, and four risk levels are divided into minor, general, serious and critical. Corresponding rectification suggestions are automatically generated, including grouting reinforcement, surface sealing repair, enhanced maintenance and local rework.

[0042] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0043] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting defects in plain concrete based on multi-source data fusion, characterized in that, Includes the following steps: The construction process parameters, high-definition surface image data, and infrared thermal imaging data of the plain concrete were collected separately. The high-definition image data of the surface is preprocessed and stitched together in sequence to reconstruct the complete three-dimensional appearance of the plain concrete and extract its appearance features, thereby completing the identification and quantification of surface defects. The construction process parameters are thresholded and normalized, and the infrared thermal imaging data are subjected to temperature field calculation, abnormal region segmentation and feature extraction. The infrared feature extraction is completed by combining acoustic-optical-thermal multi-physics field coupling analysis to obtain infrared imaging features. The defect risk warning database based on the industry standard topology is retrieved, and the processed construction process parameters, appearance features, and infrared thermal imaging features are correlated and mapped with the standard topology data in the database in multiple dimensions and compared for similarity. Based on the fusion comparison results, defect analysis is performed to generate defect identification results for plain concrete and provide rectification suggestions.

2. The method for detecting defects in plain concrete based on multi-source data fusion according to claim 1, characterized in that, The construction process parameters include concrete mix proportion parameters, pouring parameters, vibration parameters, curing parameters, formwork parameters, and environmental parameters; The construction process parameters are thresholded, normalized, and coded. Specifically, each process parameter is compared with the design threshold to obtain the process deviation value, and the deviation value is normalized to the process compliance characteristics in the range of 0 to 1. The process compliance characteristics are sorted according to the concrete construction sequence and spatial location to generate a process code data sequence that represents the compliance of the entire construction process.

3. The method for detecting defects in plain concrete based on multi-source data fusion according to claim 1, characterized in that, The construction process parameters are used to determine the authenticity of defects: the Euclidean distance between the process compliance characteristics and the standard process characteristics is calculated, and the normalized reciprocal is used as the process consistency index; the spatial location of the thermal anomaly area obtained by infrared thermal imaging, the surface defect area identified by the appearance model, and the process deviation area are matched; only when the thermal anomaly, surface defect, and process deviation spatially overlap are they determined to be real defects, excluding false defects caused by environmental interference.

4. The method for detecting defects in plain concrete based on multi-source data fusion according to claim 1, characterized in that, The panoramic stitching process involves: using optical flow motion detection to identify dynamic targets within overlapping image regions and generating dynamic masks; employing Laplacian pyramid multi-band fusion based on image gradients for static regions and a single-source priority strategy to avoid ghosting for dynamic regions. By combining adaptive exposure equalization to adjust the fusion weights, stitching seams and brightness discontinuities are eliminated, and panoramic image reconstruction is completed.

5. The method for detecting defects in plain concrete based on multi-source data fusion according to claim 1, characterized in that, The specific steps for calculating the temperature field of infrared thermal imaging data are as follows: Based on single-frame infrared thermal imaging data and concrete thermal property parameters, a virtual continuous infrared thermal imaging sequence is generated through the Green's function of the heat conduction control equation. Gaussian filtering, Fourier-Hankel transform, and cylindrical harmonic function inversion are performed on the thermal imaging sequence, and the inverse transform is performed back to the spatiotemporal domain to generate a virtual heat flow image sequence. Binarization and sigmoid discriminant function enhancement are performed on the virtual heat flow image to obtain an enhanced binarized image of thermal anomalies.

6. The method for detecting defects in plain concrete based on multi-source data fusion according to claim 1, characterized in that, The specific details of the acoustic-optical-thermal multiphysics coupling analysis are as follows: Simultaneously apply acoustic and thermal excitation to the defect candidate region segmented by infrared thermal imaging, and collect the temperature change sequence of pixel points; Discrete Fourier transform and digital phase-locked analysis were performed on the temperature change sequence to calculate the thermo-acoustic coupling parameters and resonance intensity parameters; The defect type is determined by combining geometric reference parameters, and the void depth and bonding stiffness are inverted based on the resonance peak characteristics.

7. The method for detecting defects in plain concrete based on multi-source data fusion according to claim 1, characterized in that, The surface defect identification and quantification process involves: performing histogram equalization, sharpening, and noise reduction on the panoramic image; dividing the defect detection sub-region according to texture complexity; identifying defects using grayscale abrupt changes and gradient features; extracting defect shape and size features; and determining repairability.

8. A defect detection system for plain concrete based on multi-source data fusion, characterized in that, include: Correlation data acquisition module: used to collect construction process parameters, high-definition surface image data and infrared thermal imaging data of plain concrete throughout the entire process; Image processing and recognition module: used to preprocess and stitch together high-definition surface image data sequentially, reconstruct the complete three-dimensional appearance of the plain concrete surface, extract appearance features, and complete surface defect identification and quantification. Process parameters and infrared feature processing module: used to threshold and normalize construction process parameters, calculate temperature field, segment abnormal regions and extract features from infrared thermal imaging data, and complete infrared feature extraction by combining acoustic-optical-thermal multi-physics field coupling analysis to obtain infrared imaging features; Similarity comparison module: retrieves a defect risk warning database based on industry standard topology structure, and performs multi-dimensional association mapping and similarity comparison between the processed construction process parameters, appearance features, and infrared thermal imaging features and the standard topology data in the database; Defect Analysis Module: Used to perform defect analysis based on fusion comparison results, generate defect identification results for plain concrete and provide rectification suggestions.

Images