Prefabricated Concrete T-Beam Bubble Detection Method and System

Abstract

The present application relates to the field of image processing technology, and discloses a method and system for detecting air bubbles in precast concrete T-beams. The method includes: collecting and calibrating multi-band reflection data on the surface of the T-beam, and obtaining optimized data through light intensity compensation and band screening; fusing the data into an image after spatial registration and noise reduction processing; performing edge detection and image segmentation to extract the morphological features of the air pores; calculating the depth data and performing spatial reconstruction to obtain the air pore distribution data; performing statistical analysis on the distribution data to obtain the defect evaluation result; and finally generating a detection report. Through multi-source data fusion and in-depth analysis, the present application realizes the precise positioning and quantitative evaluation of air bubble defects, and establishes a complete early warning and evaluation mechanism, improving the reliability and practicality of the detection results.

Description

Technical Field

[0001] This application relates to the field of image processing, and particularly to a method and system for detecting air bubbles in precast concrete T-beams. Background Art

[0002] Precast concrete T-beams are widely used in building and bridge engineering, and their quality control directly affects the safety and durability of structures. At present, the detection of air bubbles in precast concrete T-beams mainly relies on traditional methods such as manual visual inspection and local tapping, or single detection means such as ultrasonic waves and X-rays. These detection methods have accumulated rich experience in practical applications and provide a basic basis for the identification and evaluation of air bubble defects. At the same time, with the development of computer vision and multi-source data fusion technology, methods such as image processing and deep learning have also begun to be applied to air bubble detection, improving the automation level of detection.

[0003] However, the existing detection methods have some obvious deficiencies: First, single detection means are difficult to comprehensively reflect the spatial distribution characteristics of air bubble defects, especially the detection effect on deep air bubbles is not ideal; Second, the detection results lack a unified quantitative evaluation standard, and the judgment of the harm degree of air bubble defects mainly relies on empirical judgment and lacks scientific basis; Third, the interference of environmental factors (such as light changes, surface contamination, etc.) during the detection process will affect the accuracy of the detection results, and the existing methods lack an effective compensation mechanism; Finally, the processing and analysis of detection data lack systematicness and it is difficult to form a complete quality evaluation system. Summary of the Invention

[0004] This application provides a method and system for detecting air bubbles in precast concrete T-beams, which are used to achieve precise positioning and quantitative evaluation of air bubble defects through multi-source data fusion and in-depth analysis, and establish a complete early warning and evaluation mechanism, improving the reliability and practicality of the detection results.

[0005] In a first aspect, this application provides a method for detecting air bubbles in precast concrete T-beams, and the method for detecting air bubbles in precast concrete T-beams includes:

[0006] Collect multi-band reflection data on the surface of the precast concrete T-beam, perform position mapping calibration on the collected reflection data according to the calibration plate parameters, perform light intensity compensation processing and signal-to-noise ratio optimization on the calibrated reflection data of each band, and screen the optimal band combination through cross-validation to obtain the compensated multi-band reflection data;

[0007] Perform spatial registration according to the compensated multi-band reflection data, perform noise reduction and smoothing processing on the registered data, perform principal component calculation and data integration on the smoothed multi-band data to obtain a fused image of the reflection data;

[0008] Perform edge detection and image segmentation on the fused image based on the reflection data, numerically transform the segmented image regions according to the contour features, jointly analyze the transformed numerical values with the texture features, and use an adaptive threshold to refine the extraction of the pore boundaries to obtain the morphological parameter data of the pores;

[0009] Perform depth calculation based on the morphological parameter data of the pores, perform spatial coordinate transformation on the calculated depth data, perform double-constraint reconstruction on the transformed spatial coordinate data and the surface reflection intensity, and correct it in combination with the concrete density parameter to obtain the spatial distribution data of the pores;

[0010] Perform numerical statistics based on the spatial distribution data of the pores, classify and divide the statistical results according to a preset threshold, perform correlation operations on the divided results and the pore characteristic parameters to obtain the quantitative evaluation data of the pore defects;

[0011] Determine the warning level based on the quantitative evaluation data of the pore defects, perform automatic annotation and visualization processing on the determination results, and perform comparative analysis on the processing results and the precast component quality parameters to obtain the test report data.

[0012] In a second aspect, the present application provides a precast concrete T-beam bubble detection system, and the precast concrete T-beam bubble detection system includes:

[0013] An acquisition module, configured to acquire according to the multi-band reflection data on the surface of the precast concrete T-beam, perform position mapping calibration on the acquired reflection data according to the calibration plate parameters, perform light intensity compensation processing and signal-to-noise ratio optimization on the calibrated multi-band reflection data, and screen the optimal band combination through cross-validation to obtain the compensated multi-band reflection data;

[0014] A registration module, configured to perform spatial registration according to the compensated multi-band reflection data, perform noise reduction and smoothing processing on the registered data, perform principal component calculation and data integration on the smoothed multi-band data to obtain a fused image of the reflection data;

[0015] A segmentation module, configured to perform edge detection and image segmentation on the fused image based on the reflection data, numerically transform the segmented image regions according to the contour features, jointly analyze the transformed numerical values with the texture features, and use an adaptive threshold to refine the extraction of the pore boundaries to obtain the morphological parameter data of the pores;

[0016] A calculation module, configured to perform depth calculation based on the morphological parameter data of the pores, perform spatial coordinate transformation on the calculated depth data, perform double-constraint reconstruction on the transformed spatial coordinate data and the surface reflection intensity, and correct it in combination with the concrete density parameter to obtain the spatial distribution data of the pores;

[0017] A statistical module, configured to perform numerical statistics based on the spatial distribution data of pores, classify the statistical results according to a preset threshold, perform an association operation between the classification results and the pore characteristic parameters, and obtain quantitative evaluation data of pore defects;

[0018] A determination module, configured to determine the warning level according to the quantitative evaluation data of pore defects, perform automatic annotation and visualization processing on the determination results, compare and analyze the processing results with the quality parameters of precast components, and obtain inspection report data.

[0019] In the technical solution provided by this application, through the collection of multi-band reflection data and the position mapping calibration of calibration plate parameters, the precise spatial positioning and light intensity compensation in the bubble detection process are realized, and the accuracy of data collection is improved; through the spatial registration and noise reduction and smoothing processing of the compensated multi-band reflection data, combined with the principal component calculation and data integration technology, the quality and reliability of the fused image of the reflection data are effectively improved; the edge detection and image segmentation methods are adopted, combined with the numerical conversion of contour features and the joint analysis of texture features, and the pore boundaries are refined by an adaptive threshold, realizing the accurate acquisition of pore morphology parameters; through the depth calculation and spatial coordinate conversion of the pore morphology parameter data, combined with the double-constraint reconstruction of the surface reflection intensity and the correction of the concrete density parameters, the accuracy of the pore spatial distribution data is improved; the numerical statistics and classification technology are used, combined with the association operation of pore characteristic parameters, to realize the quantitative evaluation of pore defects; based on the quantitative evaluation data of pore defects, the warning level is determined, and through automatic annotation and visualization processing, combined with the comparative analysis of the quality parameters of precast components, complete inspection report data are formed, providing a scientific basis for quality control. The entire detection process realizes the full-process automation and standardization from data collection to result output, significantly improves the efficiency and accuracy of bubble detection in precast concrete T-beams, and provides strong technical support for quality control in engineering practice. Through the fusion analysis of multi-source data and the multi-level evaluation system, the limitations of traditional single detection methods are overcome, the accurate identification and quantitative evaluation of bubble defects are realized, and the reliability and practical value of the detection results are improved. Description of the Drawings

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0021] Figure 1 It is a schematic diagram of an embodiment of the method for detecting bubbles in a precast concrete T-beam in an embodiment of this application;

[0022] Figure 2 This is a schematic diagram of an embodiment of the precast concrete T-beam bubble detection system in the embodiments of the present application. Detailed implementation manners

[0023] The embodiments of the present application provide a method and a system for detecting bubbles in precast concrete T-beams. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" or "having" and any deformation thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0024] For ease of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 An embodiment of the method for detecting bubbles in precast concrete T-beams in the embodiments of the present application includes:

[0025] Step S101: Collect multi-band reflection data on the surface of the precast concrete T-beam, perform position mapping calibration on the collected reflection data according to the calibration plate parameters, perform light intensity compensation processing and signal-to-noise ratio optimization on the calibrated reflection data of each band, and screen the optimal band combination through cross-validation to obtain the compensated multi-band reflection data;

[0026] Step S102: Perform spatial registration according to the compensated multi-band reflection data, perform noise reduction and smoothing processing on the registered data, perform principal component calculation and data integration on the smoothed multi-band data to obtain a fused image of the reflection data;

[0027] Step S103: Perform edge detection and image segmentation according to the fused image of the reflection data, perform numerical conversion on the segmented image regions according to the contour features, perform joint analysis on the converted numerical values and the texture features, and use an adaptive threshold to refine the extraction of the pore boundaries to obtain the morphological parameter data of the pores;

[0028] Step S104: Perform depth calculation according to the morphological parameter data of the pores, perform spatial coordinate conversion on the calculated depth data, perform double-constraint reconstruction on the converted spatial coordinate data and the surface reflection intensity, and correct it in combination with the concrete density parameter to obtain the spatial distribution data of the pores;

[0029] Step S105: Perform numerical statistics based on the spatial distribution data of the pores, classify the statistical results according to a preset threshold, and perform an association operation between the classification results and the pore characteristic parameters to obtain quantitative evaluation data of the pore defects.

[0030] Step S106: Determine the warning level based on the quantitative evaluation data of the pore defects, perform automatic annotation and visualization processing on the determination results, and perform a comparative analysis between the processing results and the quality parameters of the precast components to obtain the detection report data.

[0031] It can be understood that the execution subject of this application can be a precast concrete T-beam bubble detection system, or a terminal or a server. Specifically, it is not limited here. In this embodiment of the application, the server is taken as the execution subject for illustration.

[0032] Specifically, in the data acquisition stage, a multi-spectral imaging system is used to perform an omnidirectional scan on the surface of the T-beam. The acquisition process of the multi-band reflection data is carried out through a sensor array equipped with different wavelengths, including visible light, near-infrared, and far-infrared bands. Each band collects data according to the different reflection characteristics of the concrete surface. In the actual acquisition process, the determination of the calibration plate parameters is crucial. A number of calibration points with known spacings are set on the calibration plate, and these calibration points have clear reflection characteristics in different bands. The calibration plate parameters include the spatial position information and reflection intensity information of the calibration points, and are used to establish the mapping relationship between the image space and the actual space. The position mapping calibration calculates the corresponding relationship between the calibration points in the image space and the actual space, and establishes a coordinate transformation matrix to ensure the spatial consistency of the images in different bands.

[0033] The light intensity compensation processing corrects for environmental light changes and sensor response differences. By real-time monitoring the environmental light intensity, a compensation model is established to dynamically adjust the reflection data of each band. The signal-to-noise ratio optimization analyzes the signal intensity and noise level of the data in each band, and uses a filtering algorithm to improve the data quality. When cross-validating and screening the optimal band combination, the data set is divided into a training set and a validation set. By iteratively evaluating the detection effects of different band combinations multiple times, the band combination with the highest detection accuracy is selected. In the spatial registration link, first, feature point matching is performed on the multi-band data. The key points of each band image are extracted using the SIFT feature descriptor, and a point-to-point correspondence is established. The noise reduction and smoothing processing uses the wavelet transform method to decompose and reconstruct the noise at different scales, and retains the effective signals. The principal component analysis calculates the covariance matrix of the multi-band data, extracts the main feature directions, and realizes data dimensionality reduction and information concentration.

[0034] In the image segmentation and edge detection stage, the Canny operator and region growing algorithm are combined to identify the boundaries of the bubble regions. The numerical conversion of contour features includes calculating geometric parameters such as area, perimeter, and circularity. The texture feature analysis uses the gray-level co-occurrence matrix method to extract statistical features such as energy, contrast, and entropy. The determination of the adaptive threshold is based on the gray-level distribution characteristics of the local region, and the segmentation parameters are dynamically adjusted. In the depth calculation and spatial reconstruction stage, the difference in multi-band reflection intensity is used to estimate the bubble depth. The spatial coordinate transformation converts the image coordinate system into the actual physical coordinate system. The dual-constraint reconstruction optimizes the three-dimensional shape of the bubbles by combining the surface reflection intensity and spatial position information. The concrete density parameter is used to correct the depth estimation results, considering the influence of material properties on the detection. In the numerical statistics and classification stage, the spatial distribution characteristics of the bubbles are quantitatively analyzed, including density, depth distribution, size distribution, etc. The preset threshold is determined based on engineering specifications and empirical data for the grading of bubble defects. The association operation comprehensively considers the morphological characteristics and spatial distribution characteristics of the bubbles to form a comprehensive evaluation index. In the early warning determination and result visualization stage, the defect level is determined according to the quantitative evaluation data, and different colors and markings are used for visual annotation. The detection report data includes content such as bubble distribution maps, statistical data, evaluation results, and treatment suggestions, providing a basis for quality control.

[0035] For example: On a production line of precast concrete T-beams, bubble detection is carried out on a batch of newly produced T-beams. During multi-band data acquisition, sampling and analysis are performed on typical bubble regions to obtain reflection intensity data. Through data processing, it is found that the near-infrared band makes the most significant contribution to bubble detection, and the detection effect is the best when used in combination with the visible light band. After spatial registration, the bubble boundaries are clearer, facilitating subsequent feature extraction. The depth calculation results show that the bubble depth distribution is mainly concentrated in the range of 2-5 mm from the surface. After correction according to the concrete density parameter, the actual spatial distribution of the bubbles is determined. The finally generated detection report clearly marks the bubble positions, depths, and hazard levels.

[0036] In the embodiments of the present application, by collecting multi-band reflection data and calibrating the position mapping of calibration plate parameters, accurate spatial positioning and light intensity compensation in the bubble detection process are achieved, improving the accuracy of data collection; by performing spatial registration and noise reduction and smoothing processing on the compensated multi-band reflection data, combining principal component calculation and data integration technology, the quality and reliability of the fused image of the reflection data are effectively improved; by using edge detection and image segmentation methods, combining the numerical conversion of contour features and the joint analysis of texture features, the pore boundaries are refined by an adaptive threshold, achieving the accurate acquisition of pore morphology parameters; by performing depth calculation and spatial coordinate conversion on the pore morphology parameter data, combining the dual-constraint reconstruction of surface reflection intensity and the correction of concrete density parameters, the accuracy of the pore spatial distribution data is improved; by using numerical statistics and classification and division technology, combining the correlation operation of pore characteristic parameters, the quantitative evaluation of pore defects is realized; based on the quantitative evaluation data of pore defects, the warning level is determined, and through automatic annotation and visualization processing, combining the comparative analysis of precast component quality parameters, complete detection report data are formed, providing a scientific basis for quality control. The entire detection process realizes the full-process automation and standardization from data collection to result output, significantly improving the efficiency and accuracy of bubble detection in precast concrete T-beams, and providing strong technical support for quality control in engineering practice. Through the fusion analysis of multi-source data and the multi-level evaluation system, the limitations of traditional single detection methods are overcome, the accurate identification and quantitative evaluation of bubble defects are realized, and the reliability and practical value of the detection results are improved.

[0037] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0038] (1) Collect multi-band reflection data on the surface of the precast concrete T-beam, continuously scan the surface of the precast concrete T-beam through a multi-band sensor, and obtain the original multi-band reflection data;

[0039] (2) Make the original multi-band reflection data spatially correspond to the position reference points of the calibration plate parameters, establish a position mapping relationship through spatial mapping calculation, and obtain the initial position mapping data;

[0040] (3) Calculate the light intensity deviation of each band reflection data in the initial position mapping data, compare the calculated deviation value with the real-time ambient light intensity, and obtain the light intensity correction factor;

[0041] (4) Numerically correct the light intensity correction factor and the initial position mapping data, calculate the signal-to-noise ratio of the corrected multi-band reflection data through data operation, and obtain the band quality evaluation value;

[0042] (5) The band quality assessment values ​​are grouped according to the numerical range, and each group of multi-band reflectance data is cross-validated through data analysis to obtain the band combination score;

[0043] (6) The band combination scores are arranged in descending order, and the band combination with the highest score is selected by numerical comparison. The multi-band reflection data of the band combination is normalized to obtain the compensated multi-band reflection data.

[0044] Specifically, a multi-band sensor array is used for data collection. The multi-band sensor array includes detectors in three bands: visible light, near infrared, and far infrared. Each detector responds to electromagnetic waves in a specific wavelength range. This multi-band collaborative collection method makes full use of the differences in the reflection characteristics of the concrete surface in different bands, providing rich feature information for subsequent bubble detection. During the continuous scanning process, the multi-band sensor array moves at a uniform speed along the surface of the T-beam. The scanning speed and sampling frequency need to be set to ensure that the spatial resolution meets the detection requirements. The original multi-band reflection data records the reflection intensity value of each band at each sampling point, forming a three-dimensional data matrix containing spatial position and spectral information.

[0045] The position reference point of the calibration board parameters plays a key role in the spatial mapping process. A regularly distributed array of calibration points is set on the calibration board, and each calibration point has precise physical coordinates and known reflection characteristics. The spatial transformation matrix is ​​calculated by establishing a correspondence between the calibration point image in the original multi-band reflection data and the actual position on the calibration board. This process is optimized using the least squares method to ensure the accuracy of the mapping transformation. The initial position mapping data contains the corrected spatial coordinate information, providing a unified spatial reference for subsequent processing. The calculation of light intensity deviation is an important step in correcting for changes in ambient illumination and differences in sensor response. For the reflection data of each band, the difference between the theoretical reflection intensity and the measured value at the calibration point is first calculated to establish a light intensity correction model. At the same time, the changes in ambient light intensity are monitored in real time, and the ambient light intensity is used as a reference for dynamic compensation. By comparing the calculated light intensity correction factor, accurate correction of the reflection data is achieved.

[0046] Signal-to-noise ratio calculation is the core indicator for evaluating data quality, and its calculation formula is:

[0047] ;

[0048] in: is the signal-to-noise ratio, is the target signal strength matrix; is the reference signal strength matrix; is the noise component; , is the signal weight coefficient; is the gain factor; is the noise modulation function; is the background noise term; M and N are the dimensions of the signal matrix; L is the number of noise components.

[0049] The calculation of the band quality assessment value comprehensively considers multiple indicators such as signal-to-noise ratio, contrast, and spatial resolution, and obtains the comprehensive score of each band through weighted average. The band grouping process divides the evaluation values according to the preset interval boundaries to form different band combinations. Each band combination has to go through cross-validation to evaluate its effectiveness in bubble detection. The cross-validation operation adopts the K-fold cross-validation method, which divides the data set into K subsets. Each time, one subset is selected as the validation set, and the rest are used as the training set. After cycling K times, the average performance index is obtained. This method effectively avoids the overfitting problem and improves the reliability of the evaluation results.

[0050] The descending order of the band combination scores helps to screen out the optimal band combination. The multi-band reflection data of the optimal band combination is normalized to unify the data ranges of different bands into the interval [0, 1], which is convenient for subsequent processing and analysis. The compensated multi-band reflection data not only retains the effective information of the original data but also eliminates the influence of various interference factors.

[0051] For example: In the bubble detection of a certain precast concrete T-beam, first, a multi-band sensor is used to scan the surface of the T-beam. After the acquisition of the original data is completed, spatial mapping is performed through nine position reference points on the calibration plate. Through the calculation of the light intensity deviation, it is found that the data in the near-infrared band is greatly affected by ambient light and requires a large correction factor. The calculation result of the signal-to-noise ratio shows that the combination of the visible light and near-infrared bands has the highest band quality assessment value. After cross-validation, it is confirmed that this band combination performs optimally in bubble detection. The obtained compensated multi-band reflection data clearly shows the position and morphological characteristics of the bubbles.

[0052] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0053] (1) Extract feature points from the compensated multi-band reflection data, mark the control points for the multi-band reflection data of each band through spatial analysis, calculate the coordinate transformation matrix according to the spatial distribution of the marked points, and obtain the registration reference data;

[0054] (2) Calculate the error of the marked points in the registration reference data according to the least squares criterion, and perform iterative optimization operations on the calculation results to obtain the spatial transformation parameters;

[0055] (3) Perform coordinate mapping transformation on the compensated multi-band reflection data using the spatial transformation parameters, and obtain the initial registration data through resampling calculation;

[0056] (4) Perform wavelet decomposition operation on the initial registration data, sort the decomposition coefficients according to the energy magnitude through data analysis, select the main coefficients for reconstruction calculation, and obtain the denoised and smoothed data of the registered data;

[0057] (5) Perform covariance operation on each band component in the denoised and smoothed data, determine the principal component weights based on eigenvalue decomposition through feature analysis, and obtain the data fusion weight coefficients;

[0058] (6) Perform weighted superposition on the data fusion weight coefficients and the denoised and smoothed data, perform contrast enhancement calculation on the superposition result through data processing, and obtain the fusion image of the reflection data.

[0059] Specifically, extract feature points from the compensated multi-band reflection data. The feature point extraction adopts a method based on corner detection to identify significant feature points in the reflection data of each band. The corner detection finds the points with drastic gray level changes by calculating the gray level change gradient in the local area. These feature points include typical features such as the inflection points at the edges of bubbles and the mutation points of surface textures. During the spatial analysis process, screen and match the extracted feature points to determine the positions of the control point markers. The selection of control points needs to consider the uniformity of spatial distribution and the stability of features. By calculating the spatial relationship between the feature points, construct the topological structure of the marker points to provide a basis for subsequent coordinate transformation. The spatial distribution of the marker points is evaluated through neighborhood analysis and distance calculation to ensure the calculation accuracy of the transformation matrix.

[0060] The coordinate transformation matrix is calculated using the following formula:

[0061] ;

[0062] Where: is the transformation matrix element; , , is the scale factor; is the non-linear transformation function in the x direction; is the non-linear transformation function in the y direction; is the non-linear transformation function in the z direction.

[0063] The generation of registration reference data is based on the spatial correspondence of marker points. The least squares criterion is used to optimize the transformation parameters by iteratively calculating the sum of squared residuals. During each iteration, the mapping error under the current transformation parameters is calculated, and the parameters are adjusted until the error converges below a preset threshold. The spatial transformation parameters include basic transformation information such as translation, rotation, and scaling. During the coordinate mapping transformation, the spatial transformation parameters are applied to the original data points to calculate the new coordinate positions. The resampling calculation uses bicubic interpolation to ensure data continuity and smoothness. The initial registration data retains the detailed features of the original data while achieving precise spatial position correspondence. The wavelet decomposition operation adopts a multi-level decomposition strategy to analyze the data at different scales. After sorting the decomposition coefficients by energy magnitude, the coefficients with a relatively large energy proportion are selected for reconstruction, effectively removing noise interference. The denoised and smoothed data retains the main structural features, improving the stability of subsequent processing. The covariance operation analyzes the correlation between the components of each band, constructs a covariance matrix, and determines the main change directions through eigenvalue decomposition. The magnitudes of the eigenvalues reflect the significance of data changes in each direction, based on which the weight coefficients for data fusion are determined. The weighted superposition process multiplies the weight coefficients by the data of each band and sums them to obtain the fused image. The contrast enhancement is achieved through histogram equalization, improving the visual effect of the image.

[0064] For example, when detecting air bubbles in a precast concrete T-beam, first extract feature points from the compensated multi-band reflection data. Through the corner detection algorithm, a large number of edge feature points are identified in the visible light band data, and these feature points are mainly distributed in the areas where the air bubble edges and surface texture change significantly. The control point markers select points with uniform spatial distribution and stable features to construct a control point network covering the entire detection area. During the least squares optimization process, the initial error is large, and the error gradually decreases and stabilizes after multiple iterations. The spatial transformation parameters reflect the geometric correspondence between the data of different bands, including minor deformations and distortions. The coordinate mapping transformation and resampling processing ensure the spatial consistency of the data. After wavelet decomposition, the main coefficients are selected for reconstruction according to the energy distribution to obtain the denoised and smoothed data. The covariance analysis of the data of each band shows that the near-infrared and visible light bands have strong correlations, while the far-infrared band provides complementary information. Based on this correlation feature, reasonable fusion weights are determined, and the fused image clearly shows the positions and shapes of the air bubbles.

[0065] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0066] (1) Perform gradient calculation and analysis on the fused image of the reflection data, construct an edge feature point set according to the gradient intensity values through data operations, and perform connectivity analysis calculation on the edge feature point set to obtain the initial edge contour;

[0067] (2) Region segmentation calculation is performed on the initial edge contour according to the gray distribution. Boundary tracking analysis is carried out on the segmented regions through data processing, and the closure of the tracking results is judged to screen out the valid regions, obtaining the candidate pore regions.

[0068] (3) Calculate the perimeter and area of the contour features of the candidate pore regions. Through numerical analysis, compare the calculation results with the circularity index, and analyze the shape features of the comparison results to obtain the initial morphological data.

[0069] (4) Perform spatial correspondence analysis on the region of the initial morphological data and the fused image of the reflection data. Extract the gray distribution features and texture direction features of the corresponding regions through data extraction to obtain the texture feature data.

[0070] (5) Perform boundary refinement calculation on the candidate pore regions according to the texture feature data. Calculate and analyze the local contrast through numerical operations to determine the dynamic threshold range, obtaining the set of pore boundary points.

[0071] (6) Perform curve fitting operation on the set of pore boundary points. Extract geometric parameters and perform feature quantization calculation on the fitting curve through data processing to obtain the morphological parameter data of the pores.

[0072] Specifically, gradient calculation is performed on the fused image of the reflection data. The gradient calculation uses a multi-directional gradient operator to calculate the gray change rate of each pixel point in the horizontal and vertical directions. Through gradient intensity threshold screening, the pixel points with significant changes are marked as edge candidate points. These edge candidate points undergo connectivity analysis to determine the topological relationship between adjacent points, forming the initial edge contour. The region segmentation of the initial edge contour uses the watershed algorithm based on the gray distribution to divide the image into different regions. The watershed algorithm starts from the local minimum points and gradually grows the region boundaries until adjacent regions meet. During the boundary tracking process, perform boundary integrity check on each segmented region to evaluate the closure and continuity of the boundary. The closure judgment is based on the connection relationship and gap size of the boundary points, and the closed and complete regions are screened out as the candidate pore regions.

[0073] The contour feature analysis of the candidate pore regions includes the calculation of geometric parameters. The perimeter calculation is realized by accumulating the Euclidean distances between boundary points, and the area calculation uses Green's formula. The circularity index reflects the similarity between the region shape and the standard circle, and the calculation formula is:

[0074] ;

[0075] Where: is the region shape matrix; is the roundness weight factor; is the correction coefficient; is the boundary smoothness; is the shape feature vector; is the standard circle feature vector; is the normalization factor; R, S are the dimensions of the region matrix; T is the dimension of the feature vector.

[0076] Spatial correspondence analysis registers the initial morphological data with the fused image and extracts the gray-scale features of the corresponding regions. The gray-scale distribution features include statistics such as mean, variance, skewness, and kurtosis. The texture direction feature is obtained by calculating the directional gradient histogram of the local region, reflecting the directional characteristics of the surface structure.

[0077] Boundary refinement calculation is based on local contrast analysis and adopts an adaptive threshold method. By analyzing the gray-scale change trend near the pore boundary, a suitable threshold range is determined to achieve precise boundary positioning. The pore boundary point set records the coordinates of the refined boundary position.

[0078] Curve fitting uses the B-spline interpolation method to smoothly fit the boundary point set. Geometric parameter extraction includes feature quantities such as curvature and eccentricity, and standardized morphological parameter data is obtained through feature quantization.

[0079] For example: In the bubble detection on the surface of a certain precast concrete T-beam, obvious edge features are shown after the fused image undergoes gradient calculation. Connectivity analysis connects adjacent edge points into continuous contour lines. After region segmentation, multiple closed regions are found, and some of the regions are close to circular in shape. Through perimeter and area calculations and combined with circularity index analysis, the real bubble regions are screened out. The gray-scale distribution of these regions shows typical depression features, and the texture direction feature shows that the edges have a regular radial distribution. After boundary refinement, an accurate bubble contour is obtained, and the curve fitting result shows that the bubble shape is regular and the curvature changes smoothly. The final morphological parameter data accurately describes the geometric features of the bubble.

[0080] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0081] (1) Perform depth conversion calculation on the gray-scale distribution features in the morphological parameter data of the pores, establish a depth mapping relationship through data analysis, and perform boundary constraint operations on the mapping result to obtain the initial depth data;

[0082] (2) Perform coordinate transformation calculation on the initial depth data according to the spatial distribution law, perform repositioning analysis on the depth values through numerical operations, and perform error correction processing on the repositioning result to obtain the spatial coordinate data of the initial depth data;

[0083] (3) Perform joint operations on the spatial coordinate data and the surface reflection intensity. Through data processing, calculate and analyze the reflection intensity gradient to determine the depth correction coefficient, and perform weighted combination of the correction coefficient and the spatial coordinate data to obtain the initial reconstruction data;

[0084] (4) Compare and analyze the initial reconstruction data with the surface reflection intensity of the local area. Through numerical calculation, adjust the spatial distribution parameters of the neighborhood features, and perform consistency verification on the adjusted parameters to obtain the reconstruction correction data;

[0085] (5) Perform correlation calculations on the reconstruction correction data and the concrete density parameters. Through data analysis, perform compensation processing on the spatial coordinate data, and compare and calculate the compensation result with the theoretical distribution to obtain the correction parameter set;

[0086] (6) Perform spatial calibration calculations on the reconstruction correction data using the correction parameter set. Through data processing, perform multi-dimensional feature fusion localization on the pore positions, and perform statistical analysis on the localization results to obtain the spatial distribution data of the pores.

[0087] Specifically, perform depth conversion calculations on the morphological parameter data of the pores. The depth conversion calculations are based on the gray-scale distribution characteristics. By establishing the correspondence between the gray-scale values and the actual depths, quantitative description of the surface topography is realized. Among them, the formula for the depth conversion calculations is as follows:

[0088] ;

[0089] Among them: is the depth conversion value; is the gray-scale distribution matrix; is the depth mapping weight; is the depth conversion exponent; is the calibration curve parameter; is the reference depth value; is the depth offset constant; B, H are the dimensions of the gray-scale matrix; J is the number of calibration parameters, a and b are the index parameters.

[0090] The establishment of the depth mapping relationship adopts a piecewise linear mapping method, and different mapping coefficients are set for different depth ranges. The boundary constraint operation ensures the physical rationality of the depth value and prevents abnormal depth estimation results. The spatial coordinate transformation converts the initial depth data into a three-dimensional physical space. It should be noted that in the embodiments of this application, the depth value refers to the measured value of the distance that the air pores extend from the concrete surface into the interior, and it is a key parameter characterizing the position of the air pores in the three-dimensional space. Specifically, the depth value reflects the spatial position information of the air pores in the direction perpendicular to the concrete surface, and the unit is usually millimeters (mm). The coordinate transformation process takes into account the geometric characteristics of the T-beam surface and the scanning direction, and establishes the conversion relationship between the local coordinate system and the global coordinate system. The depth value repositioning analysis optimizes and adjusts the single-point depth value by considering the depth distribution of adjacent points. The error correction process uses a statistical filtering method to eliminate the random errors and systematic errors in the depth measurement.

[0091] The joint operation of the spatial coordinate data and the surface reflection intensity is a key link in the reconstruction process. The calculation of the concrete density correlation adopts the following formula:

[0092] ;

[0093] where: D is the density correlation coefficient, is the density distribution matrix; is the spatial weight function; is the depth compensation factor; is the standard density component; is the non-linear correction exponent; is the scale factor; is the compensation constant; U, V are the dimensions of the distribution matrix; W is the number of compensation factors.

[0094] The calculation of the surface reflection intensity gradient determines the depth correction coefficient by analyzing the change rate of the reflection intensity in the local area. The weighted combination process fuses the correction coefficient with the spatial coordinate data to generate the initial reconstruction data. The comparative analysis of the initial reconstruction data and the surface reflection intensity in the local area mainly focuses on the spatial distribution law of the reflection intensity. The adjustment of the spatial distribution parameters of the neighborhood features is based on the local consistency principle to ensure the continuity and smoothness of the reconstruction result. The correlation calculation between the reconstructed correction data and the concrete density parameters takes into account the influence of material properties on the formation of air bubbles. By establishing the correspondence between the density and the air bubble distribution, the spatial coordinate data is compensated. The comparison operation between the compensation result and the theoretical distribution helps to identify abnormal areas and ensure the accuracy of the reconstruction result.

[0095] The spatial calibration calculation of the corrected parameter set adopts a multi-level optimization strategy. The multi-dimensional feature fusion positioning comprehensively considers depth information, reflection intensity, and density characteristics, improving the accuracy of pore position determination. The statistical analysis focuses on evaluating the regularity and concentration degree of pore distribution.

[0096] For example, when detecting air bubbles in a precast concrete T-beam, first analyze the gray-scale distribution characteristics of the surface. Through depth conversion calculation, the mapping relationship between gray-scale values and actual depths is established. The spatial coordinate transformation converts the depth data into the actual physical space and performs error correction at the same time. The gradient analysis of the surface reflection intensity shows that the air bubble area has obvious intensity change characteristics. During the joint operation process, the depth correction coefficient is determined according to the change of the reflection intensity, and the initial reconstruction result is optimized. The correlation analysis of the concrete density parameters shows that there is an obvious correlation between the formation of air bubbles and the density of local areas. Through spatial parameter adjustment and consistency verification, the reliability of the reconstruction result is ensured. The final spatial distribution data clearly reflects the distribution characteristics of air bubbles on the surface of the T-beam, including key information such as depth, position, and density. This measurement and analysis process fully considers the material characteristics and structural features of the precast concrete T-beam, and through the fusion and analysis of multi-source data, realizes the precise positioning and characterization of air bubble defects. The processing results not only reflect the morphological characteristics of individual air bubbles but also reveal the distribution law of air bubble groups.

[0097] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0098] (1) Calculate the frequencies of the depth values and position coordinates in the spatial distribution data of the pores, perform aggregation analysis on the pore distribution density through interval division, and normalize the aggregation data to obtain initial statistical data;

[0099] (2) Numerically classify the initial statistical data according to spatial division, count the number of pores in each area through density distribution, and compare the statistical values with a preset threshold in an interval to obtain pore classification parameters;

[0100] (3) Hierarchically process the pore classification parameters, determine the distribution characteristics of the pore groups according to the spatial position relationship, perform regional connectivity analysis on the distribution characteristic data, and obtain the classification statistical results;

[0101] (4) Match the classification statistical results with the pore characteristic parameters, establish an association matrix of pore morphology and distribution through parameter mapping, and perform numerical calculation on the association matrix to obtain characteristic association data;

[0102] (5) Configure weights for the feature correlation data according to the spatial distribution law, construct an evaluation index system through multi-dimensional feature calculation, conduct hierarchical analysis on the evaluation indexes, and obtain an evaluation parameter set;

[0103] (6) Perform numerical operations on the evaluation parameter set, determine the evaluation value of the pore defect through the fusion of multiple indexes, compare and analyze the evaluation value with the standard range, and obtain the quantitative evaluation data of the pore defect.

[0104] Specifically, conduct in-depth analysis on the spatial distribution data of pores. The frequency calculation of the pore depth value and position coordinates adopts the statistical histogram method. Divide the entire detection area into regular spatial units, and count the number of pores appearing in each unit. The interval division process considers the typical size of the bubbles, divides the depth range into several equally spaced sections, and is convenient for observing the distribution of bubbles at different depths. The aggregation degree analysis evaluates the concentration degree of the bubble group by calculating the bubble density in the local area. The normalization process unifies the statistical data in different regions and at different depths into the same numerical range, which is convenient for comparative analysis. The spatial division of the initial statistical data adopts the adaptive grid method, and dynamically adjusts the grid size according to the density of the pore distribution. Use smaller grid units in the pore-dense area and larger grid units in the sparse area. This division method not only ensures the accuracy of the analysis but also improves the calculation efficiency. The weighted counting method is used for the quantity statistics of pores in each region, considering the influence of the pore size on the statistical results. The preset threshold is a reference value determined based on engineering experience and quality standards. The interval comparison process compares the statistical value with these thresholds to determine the severity of the pore distribution.

[0105] The hierarchical processing of stomatal classification parameters establishes a multi-level classification system. According to characteristics such as the depth, size, and density of stomata, the stomata are divided into different grades. The analysis of spatial position relationships focuses on examining the relative positions and clustering phenomena among stomata. The distribution characteristics of stomatal clusters include the size, shape, and directionality of the clusters. The regional connectivity analysis uses graph theory methods to connect adjacent stomata into a network and evaluate the overall distribution characteristics of stomatal clusters. The data matching process between the classification statistics results and the stomatal characteristic parameters establishes the correlation between morphological characteristics and spatial distribution. Parameter mapping uses multi-dimensional mapping methods to map morphological parameters and distribution parameters to the same feature space. The correlation matrix records the degree of correlation between different features, and the dependence relationship between features is determined through numerical calculations. Specifically, the stomatal characteristic parameters specifically include: Morphological characteristic parameters: including quantitative indicators such as the diameter, area, perimeter, circularity index, and aspect ratio of the stomata to describe the geometric shape of the stomata. Position characteristic parameters: including the three-dimensional coordinates of the stomata in the concrete T-beam. Spatial distribution characteristic parameters: including parameters such as the aggregation degree, spatial density, and adjacent stomatal spacing of the stomata to characterize the relative position relationship of multiple stomata. Physical characteristic parameters: including parameters such as the surface reflectivity, edge sharpness, and internal filling state of the stomata to reflect the physical properties of the stomata.

[0106] Data matching is achieved by establishing the correspondence between the classification statistics results and the stomatal characteristic parameters. It is expressed by the following formula:

[0107] ;

[0108] Where: is the element of the correlation matrix; is the classification statistics result matrix; is the feature mapping function; is the matching weight index; D and E are the dimensions of the statistical matrix.

[0109] The feature mapping function J(i,j,d,e) defines the correspondence between the statistical results and the characteristic parameters, and it can be expressed as:

[0110] ;

[0111] Where: is the stomatal characteristic parameter; is the classification statistics weight; A is the normalization coefficient; B is the feature dimension. The feature correlation data can be obtained through the following calculation:

[0112] ;

[0113] Where: is the feature correlation data; is the i-th component of the feature vector; is the eigenvalue; G is the number of eigenvalues.

[0114] The weight configuration of the feature correlation data takes into account the importance of different features for the evaluation of pore defects. The multi-dimensional feature calculation synthesizes morphological features, distribution features, and material property indicators to construct a complete evaluation index system. Specifically, the multi-dimensional feature calculation constructs an evaluation index system by comprehensively calculating the feature indicators of multiple dimensions such as the morphological features, distribution features, material properties, and environmental factors of the pores, forming a complete evaluation system. This process can be represented by the following mathematical model:

[0115] ;

[0116] where: is the evaluation index system; is the feature category weight; is the feature component; is the dimension coefficient; is the number of features in each dimension; u represents the feature category (1 - morphological feature, 2 - distribution feature, 3 - material property, 4 - environmental factor); n is the evaluation index index. First, the feature components within each feature category are weighted and summed, and then normalized; then, the calculation results of different feature categories are weighted and fused again to form the final evaluation index system.

[0117] Feature component The calculation of takes into account the difference between the relative value and the reference value of the feature, as well as the adaptive weight, and its calculation formula is:

[0118] ;

[0119] where: is the original eigenvalue; is the feature reference value; is the normalization factor; is the adaptive weight.

[0120] The analytic hierarchy process is adopted to determine the weight coefficients of various indicators. The evaluation parameter set includes quantitative evaluation indicators and corresponding weight values. The numerical operation process weights and integrates multiple indicators to obtain a comprehensive evaluation value. These evaluation values are compared with the pre-set standard range to determine the severity of pore defects. The quantitative evaluation data not only reflects the defect status of a single area but also the overall quality level. Specifically, the process of determining the evaluation value of pore defects by integrating multiple indicators is a process of integrating the indicators in each evaluation parameter set into a single evaluation value through weighted calculation. The specific implementation includes index normalization, index weight assignment, weighted integration calculation, and evaluation value calibration. Index normalization is to convert various indicators with different dimensions and value ranges (such as pore diameter, depth, density, aggregation degree, etc.) into a unified numerical interval [0,1] to ensure the comparability of each indicator. This is usually achieved by the maximum-minimum normalization method or the z-score normalization method.

[0121] Index weight assignment is to assign different weight coefficients according to the importance of each indicator to the performance of the concrete structure. For example, if the influence of pore diameter on structural strength may be greater than that of pore depth, the weight of the pore diameter indicator will increase accordingly. The weight assignment can be determined by methods such as expert scoring method, analytic hierarchy process, or historical data regression analysis.

[0122] Weighted integration calculation is to use the method of linear weighting or non-linear weighting to multiply all normalized indicators by their corresponding weights and then sum them to obtain a single evaluation value. If there are interactions between some indicators, cross terms will also be introduced for calculation to more comprehensively reflect the interaction relationship between indicators. Evaluation value calibration is to calibrate the initially obtained evaluation value according to historical data and engineering experience to ensure that the evaluation results conform to the actual situation. The calibration process often needs to be carried out by the comparison and verification method, comparing the algorithm results with the expert evaluation results and making necessary adjustments. The process of comparing and analyzing the evaluation value with the standard range to obtain quantitative evaluation data mainly includes several links: standard range definition, interval mapping transformation, defect influence factor calculation, defect distribution characteristic synthesis, quantitative evaluation data generation, and result reliability verification. Standard range definition is to determine the evaluation value range corresponding to different grades of defects based on national standards, industry specifications, and engineering experience. For example, the evaluation value can be divided into three grade intervals: 0.0 - 0.3 (minor defect), 0.3 - 0.6 (medium defect), 0.6 - 1.0 (severe defect) to form an evaluation standard. Interval mapping transformation is to map a single evaluation value to the corresponding defect grade interval to determine the severity level of pore defects. This process is usually achieved by the look-up table method or the function mapping method to ensure that the evaluation results have clear grade attributes.

[0123] The calculation of defect influence factors is to calculate the defect influence factors according to the defect level and the importance of the position of pores in the component, so as to quantify the influence degree of defects on the overall performance of the component. For pore defects located in the stress concentration area, their influence factors are often given higher weights. The defect distribution characteristics comprehensively consider the spatial distribution characteristics of pore defects, such as aggregated distribution, chain-like distribution or uniform distribution, etc., and adjust the defect influence factors. There are significant differences in the influence of different distribution characteristics on the structural performance, which need to be fully considered during the evaluation process. The generation of quantitative evaluation data is to comprehensively evaluate numerical values, defect levels, influence factors and distribution characteristics to generate multi-dimensional quantitative evaluation data, including defect severity index, structural influence coefficient, quality grade assessment and safety risk degree, etc. These data are expressed in specific numerical forms, which are convenient for subsequent analysis and decision-making. The verification of result reliability is to verify the accuracy and reliability of the evaluation data by comparing with actual engineering cases and test results, and perform parameter fine-tuning if necessary. The verification process helps to optimize the evaluation algorithm and improve the accuracy and practicality of the evaluation results.

[0124] For example: In the bubble detection of a precast concrete T-beam, the pore distribution data on the surface was first obtained. Frequency calculation showed that the pore depth was mainly concentrated in the range of 2-5 mm on the surface layer, and the position distribution showed the characteristics of local aggregation. Aggregation analysis found that in the stress concentration area of the beam, the pore density was significantly higher than that in other areas. After these data were normalized, the data in different areas became comparable. After spatial division, it was found that the distribution of pores showed obvious regional differences. The number of pores in some areas exceeded the preset threshold, and these areas were marked as key attention areas. Through hierarchical processing, the pores were divided into three grades: slight, medium and severe. Regional connectivity analysis showed that some pores formed a continuous chain-like distribution, and this distribution characteristic had a more significant impact on the structure. The correlation analysis between pore morphology and distribution showed that larger-sized pores tended to be concentrated in specific areas, while the distribution of smaller-sized pores was relatively uniform. These characteristics were quantified as numerical values in the correlation matrix. When configuring weights, considering that larger-sized pores had a greater impact on the structural performance, their weight coefficients were correspondingly increased. The final evaluation result clearly reflected the distribution characteristics and severity of pore defects, providing an important basis for quality control.

[0125] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0126] (1) Conduct numerical interval analysis on the quantitative evaluation data of pore defects, divide the pore defects in each interval into grades through distribution characteristics, establish early warning index data for the division results, and obtain initial early warning parameters;

[0127] (2)Hierarchically classify the initial warning parameters according to their spatial distribution, calculate the hazard level coefficient through the defect degree, perform identification mapping on defects at each level, and obtain the warning judgment result;

[0128] (3)Perform regional annotation on the warning judgment result, spatially locate the defect information through position coordinates, and convert the positioning data into visual markers to obtain the annotation data set;

[0129] (4)Graphically process the annotation data set, construct a visual interface through multi-dimensional feature expression, and optimize the layout of interface elements to obtain the visual result;

[0130] (5)Match the visual result with the quality parameters of precast components, establish a quality assessment correlation matrix through parameter mapping, and perform comprehensive calculation on the correlation data to obtain the quality comparison data;

[0131] (6)Analyze and organize the quality comparison data according to the evaluation criteria, generate inspection and evaluation content through the fusion of multiple indicators, and format the evaluation content to obtain the inspection report data.

[0132] Specifically, perform numerical interval analysis on the quantitative evaluation data of pore defects. The numerical interval analysis uses a hierarchical classification method. According to the characteristics such as the size, depth, and distribution density of pores, different evaluation intervals are set. The distribution characteristic analysis considers the aggregation degree and continuity of pores in space, and determines the characteristic parameters of pore distribution through statistical methods. The grade division adopts a multi-level classification system, and the pore defects are divided into different hazard levels. The warning index data includes thresholds in multiple dimensions such as pore size, depth, and density, and these thresholds are determined based on engineering standards and practical experience. The spatial distribution analysis of the initial warning parameters focuses on the position characteristics of the defects. The hierarchical classification process adopts a top-down classification strategy. First, determine the large-scale hazard area, and then gradually refine to the specific position. The calculation of the defect degree comprehensively considers the geometric characteristics and material properties of the pores, and obtains the hazard level coefficient through weighted calculation. The identification mapping represents defects at different levels with different identifiers to form a clear warning judgment result.

[0133] The regional annotation of the early warning judgment results adopts a multi-level annotation strategy. The spatial positioning of the position coordinates uses a combination of absolute coordinates and relative coordinates to ensure the accuracy of the annotation. The visual marking of the defect information uses different graphic symbols and color coding to intuitively express the type and severity of the defect. The annotation data set contains multi-dimensional data such as position information, defect characteristics, and early warning levels. The graphical processing of the annotation data set uses a variety of visualization techniques. The multi-dimensional feature expression uses visual elements such as color, shape, and size to display different data dimensions. The construction of the visualization interface takes into account the hierarchical structure and correlation of the data, and adopts a combination of hierarchical layout and grid layout. The layout optimization of the interface elements ensures the clear display and intuitive understanding of the information. The data matching process between the quality parameters of the precast components and the visualization results establishes a mapping relationship between the defect characteristics and the quality indicators. The quality assessment correlation matrix records the mutual influence between different parameters, and determines the weights and correlations of each indicator through matrix operations. The comprehensive calculation process weights and fuses multiple quality indicators to form an overall quality assessment result.

[0134] The analysis and collation of the quality comparison data are based on the preset evaluation criteria, and the test results are compared with the quality requirements. The fusion of multiple indicators adopts the analytic hierarchy process to reasonably allocate the weights of each indicator. The inspection and evaluation content includes multiple parts such as numerical analysis results, graphical display, and quality suggestions. The formatting process ensures the standardization and readability of the report.

[0135] For example: In the bubble detection of a certain precast concrete T-beam, according to the quantitative evaluation data, numerical interval analysis is carried out, and it is found that the pore distribution has obvious regularity. Larger-sized pores (diameter greater than 5mm) are mainly distributed in the web area of the beam, while smaller-sized pores are more scattered. Through the analysis of the distribution characteristics, a two-dimensional classification system based on position and size is established. The spatial distribution analysis shows that the pores form a continuous chain-like distribution in some areas, and this situation is marked as a high hazard level. The calculation of the hazard level coefficient takes into account the size, depth, and continuity of the pores, and the weighted average method is used to obtain a comprehensive score. Different levels of defects are mapped using different colors and shapes of identifiers, with red indicating serious defects, yellow indicating medium defects, and green indicating minor defects.

[0136] The visualization process adopts a three-dimensional display method, representing the depth distribution of pores by different shades of color, and representing the bubble sizes by circular marks of different sizes, forming an intuitive spatial distribution map. The interface layout adopts a partition display method, with the overall distribution map shown on the left and the detailed parameters and statistical data shown on the right. During the quality assessment process, the detected data is compared with quality parameters such as the strength requirements and appearance requirements of the precast components. Through correlation analysis, it is found that there is an obvious correlation between the pore density and the concrete strength, and this finding is used to optimize the production process. The final inspection report includes quantitative analysis results, distribution maps, quality assessment, and improvement suggestions, providing comprehensive technical support for quality control.

[0137] The above describes the method for detecting bubbles in precast concrete T-beams in the embodiments of the present application. Next, the system for detecting bubbles in precast concrete T-beams in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the system for detecting bubbles in precast concrete T-beams in the embodiments of the present application includes:

[0138] An acquisition module 201, configured to collect according to the multi-band reflection data on the surface of the precast concrete T-beam, perform position mapping calibration on the collected reflection data according to the calibration plate parameters, perform light intensity compensation processing and signal-to-noise ratio optimization on the calibrated reflection data of each band, and screen the optimal band combination through cross-validation to obtain the compensated multi-band reflection data;

[0139] A registration module 202, configured to perform spatial registration according to the compensated multi-band reflection data, perform noise reduction and smoothing processing on the registered data, perform principal component calculation and data integration on the smoothed multi-band data to obtain a fused image of the reflection data;

[0140] A segmentation module 203, configured to perform edge detection and image segmentation according to the fused image of the reflection data, perform numerical conversion on the segmented image regions according to the contour features, perform joint analysis on the converted numerical values and the texture features, and use an adaptive threshold to refine the extraction of the pore boundaries to obtain the morphological parameter data of the pores;

[0141] A calculation module 204, configured to perform depth calculation according to the morphological parameter data of the pores, perform spatial coordinate conversion on the calculated depth data, perform double-constraint reconstruction on the converted spatial coordinate data and the surface reflection intensity, and correct it in combination with the concrete density parameter to obtain the spatial distribution data of the pores;

[0142] A statistics module 205, configured to perform numerical statistics according to the spatial distribution data of the pores, classify the statistical results according to a preset threshold, perform correlation operations on the classification results and the pore characteristic parameters to obtain the quantitative evaluation data of the pore defects;

[0143] A determination module 206, configured to determine an early warning level according to the quantitative evaluation data of the pore defects, automatically label and visually process the determination result, compare and analyze the processed result with the quality parameters of the precast component, and obtain the inspection report data.

[0144] Through the collaborative cooperation of the above-mentioned various components, by using multi-band reflection data for acquisition and position mapping calibration of the calibration plate parameters, accurate spatial positioning and light intensity compensation in the bubble detection process are achieved, improving the accuracy of data acquisition; through spatial registration and noise reduction and smoothing processing of the compensated multi-band reflection data, combined with principal component calculation and data integration technology, the quality and reliability of the fused image of the reflection data are effectively improved; edge detection and image segmentation methods are adopted, combined with the numerical conversion of contour features and the joint analysis of texture features, and the pore boundaries are refined by an adaptive threshold to accurately obtain the pore morphology parameters; through depth calculation and spatial coordinate conversion of the pore morphology parameter data, combined with the double-constraint reconstruction of the surface reflection intensity and the correction of the concrete density parameters, the accuracy of the pore spatial distribution data is improved; numerical statistics and classification and division technologies are used, combined with the correlation operation of the pore characteristic parameters, to achieve the quantitative evaluation of the pore defects; based on the quantitative evaluation data of the pore defects, the early warning level is determined, through automatic labeling and visual processing, combined with the comparative analysis of the quality parameters of the precast component, and complete inspection report data are formed, providing a scientific basis for quality control. The entire detection process realizes the full-process automation and standardization from data acquisition to result output, significantly improving the efficiency and accuracy of the bubble detection of precast concrete T-beams, and providing strong technical support for quality control in engineering practice. Through the fusion analysis of multi-source data and the multi-level evaluation system, the limitations of traditional single detection methods are overcome, the accurate identification and quantitative evaluation of bubble defects are realized, and the reliability and practical value of the detection results are improved.

[0145] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A method for detecting bubbles in precast concrete T-beams, characterized in that: The precast concrete T-beam bubble detection method comprises: According to the multi-band reflection data of the surface of the precast concrete T-beam, the collected reflection data is calibrated by position mapping according to the calibration plate parameters, the calibrated reflection data of each band is subjected to light intensity compensation processing and signal-to-noise ratio optimization, and the optimal band combination is selected by cross-validation to obtain the compensated multi-band reflection data; Perform spatial registration based on the compensated multi-band reflection data, perform noise reduction and smoothing on the registered data, perform principal component calculation and data integration on the smoothed multi-band data, and obtain a fused image of the reflection data; Edge detection and image segmentation are performed based on the fused image of the reflection data. The segmented image area is converted into numerical values ​​according to the contour features. The converted numerical values ​​are jointly analyzed with the texture features. The pore boundaries are refined and extracted using an adaptive threshold to obtain the morphological parameter data of the pores. The depth is calculated based on the morphological parameter data of the pores, the calculated depth data is transformed into spatial coordinates, the transformed spatial coordinate data and the surface reflection intensity are reconstructed with dual constraints, and the data are corrected in combination with the concrete density parameters to obtain the spatial distribution data of the pores; Perform numerical statistics based on the spatial distribution data of pores, classify and divide the statistical results according to the preset threshold, and perform correlation operation on the division results and the pore characteristic parameters to obtain quantitative evaluation data of pore defects; The warning level is determined based on the quantitative evaluation data of the pore defects, the determination results are automatically labeled and visualized, and the processing results are compared and analyzed with the quality parameters of the prefabricated components to obtain the test report data.

2. The method for detecting bubbles in precast concrete T-beams according to claim 1, characterized in that: The method collects multi-band reflection data from the surface of the precast concrete T-beam, performs position mapping calibration on the collected reflection data according to the calibration plate parameters, performs light intensity compensation processing and signal-to-noise ratio optimization on the calibrated reflection data of each band, selects the optimal band combination by cross-validation, and obtains compensated multi-band reflection data, including: Collecting multi-band reflection data on the surface of the precast concrete T-beam, and continuously scanning the surface of the precast concrete T-beam with a multi-band sensor to obtain original multi-band reflection data; The original multi-band reflection data is spatially corresponded to the position reference point of the calibration plate parameter, and a position mapping relationship is established through spatial mapping calculation to obtain initial position mapping data; Performing light intensity deviation calculation on each band reflection data in the initial position mapping data, comparing the calculated deviation value with the real-time ambient light intensity to obtain a light intensity correction factor; Numerically correcting the light intensity correction factor and the initial position mapping data, and calculating the signal-to-noise ratio of the corrected multi-band reflection data through data calculation to obtain a band quality assessment value; The band quality assessment values ​​are grouped into bands according to the numerical intervals, and a cross-validation operation is performed on each group of multi-band reflection data through data analysis to obtain a band combination score; The band combination scores are arranged in descending order, the band combination with the highest score is selected by numerical comparison, and the multi-band reflection data of the band combination is normalized to obtain the compensated multi-band reflection data.

3. The method for detecting bubbles in precast concrete T-beams according to claim 1, characterized in that: The method of performing spatial registration according to the compensated multi-band reflection data, performing noise reduction and smoothing processing on the registered data, and performing principal component calculation and data integration on the smoothed multi-band data to obtain a fused image of the reflection data includes: Extracting feature points from the compensated multi-band reflection data, marking control points of the multi-band reflection data of each band through spatial analysis, calculating a coordinate transformation matrix according to the spatial distribution of the marking points, and obtaining registration reference data; The error calculation of the marked points in the registration reference data is performed according to the least squares criterion, and the calculation results are iteratively optimized to obtain the spatial transformation parameters; Using the spatial transformation parameters to perform coordinate mapping conversion on the compensated multi-band reflection data, and obtaining initial registration data through resampling calculation; Performing wavelet decomposition operation on the initial registration data, sorting the decomposition coefficients according to energy size through data analysis, selecting main coefficients for reconstruction calculation, and obtaining denoised and smoothed data of the registered data; Performing covariance calculation on each band component in the noise reduction and smoothing data, determining the principal component weight based on eigenvalue decomposition through feature analysis, and obtaining the data fusion weight coefficient; The data fusion weight coefficient and the noise reduction smoothing data are weightedly superimposed, and contrast enhancement calculation is performed on the superposition result through data processing to obtain a fused image of the reflection data.

4. The method for detecting bubbles in precast concrete T-beams according to claim 1, characterized in that: The method performs edge detection and image segmentation based on the fused image of the reflection data, performs numerical conversion on the segmented image area according to the contour features, jointly analyzes the converted numerical values ​​with the texture features, and uses an adaptive threshold to extract the pore boundaries in a refined manner to obtain the morphological parameter data of the pores, including: Performing gradient calculation analysis on the fused image of the reflection data, constructing an edge feature point set according to the gradient intensity value through data calculation, and performing connectivity analysis calculation on the edge feature point set to obtain an initial edge contour; The initial edge contour is segmented and calculated according to the grayscale distribution, the segmented area is tracked and analyzed by data processing, and the tracking result is judged for closure to select the effective area to obtain the candidate pore area; Calculating the perimeter and area of ​​the contour features of the candidate pore area, comparing the calculated results with the circularity index through numerical analysis, and analyzing the shape features of the compared results to obtain initial morphological data; Performing spatial correspondence analysis on the region of the initial morphological data and the fused image of the reflection data, and extracting grayscale distribution features and texture direction features of the corresponding region through data to obtain texture feature data; Perform boundary refinement calculation on the candidate pore area according to the texture feature data, calculate and analyze the local contrast through numerical calculation to determine the dynamic threshold range, and obtain a pore boundary point set; The pore boundary point set is subjected to curve fitting operation, and geometric parameter extraction and feature quantification calculation are performed on the fitting curve through data processing to obtain the morphological parameter data of the pores.

5. The method for detecting bubbles in precast concrete T beams according to claim 1, characterized in that: The depth calculation is performed according to the morphological parameter data of the pores, the calculated depth data is converted into spatial coordinates, the converted spatial coordinate data and the surface reflection intensity are double-constrained reconstructed, and the spatial distribution data of the pores is obtained by combining the concrete density parameter for correction, including: Performing depth conversion calculation on the grayscale distribution features in the morphological parameter data of the pores, establishing a depth mapping relationship through data analysis, performing boundary constraint calculation on the mapping results, and obtaining initial depth data; The initial depth data is subjected to coordinate transformation calculation according to the spatial distribution law, the depth value is relocated and analyzed through numerical calculation, and the relocation result is subjected to error correction processing to obtain the spatial coordinate data of the initial depth data; Performing a joint operation on the spatial coordinate data and the surface reflection intensity, calculating and analyzing the reflection intensity gradient through data processing to determine the depth correction coefficient, and performing a weighted combination of the correction coefficient and the spatial coordinate data to obtain initial reconstructed data; Comparing and analyzing the initial reconstructed data with the surface reflection intensity of the local area, adjusting the spatial distribution parameters of the neighborhood features through numerical calculation, and verifying the consistency of the adjusted parameters to obtain reconstructed corrected data; The reconstructed correction data and the concrete density parameter are correlated and calculated, the spatial coordinate data is compensated through data analysis, and the compensation result is compared with the theoretical distribution to obtain a correction parameter set; The correction parameter set is used to perform spatial calibration calculation on the reconstructed correction data, the pore positions are located by multi-dimensional feature fusion through data processing, and the positioning results are statistically analyzed to obtain the spatial distribution data of the pores.

6. The method for detecting bubbles in precast concrete T-beams according to claim 1, characterized in that: The method of performing numerical statistics based on the spatial distribution data of the pores, classifying and dividing the statistical results according to a preset threshold, and correlating the division results with the pore characteristic parameters to obtain quantitative evaluation data of the pore defects includes: Performing frequency calculation on the depth values ​​and position coordinates in the spatial distribution data of the pores, performing aggregation analysis on the pore distribution density by interval division, and normalizing the aggregation data to obtain initial statistical data; The initial statistical data is numerically classified according to spatial division, the number of stomata in each region is counted by density distribution, and the statistical value is compared with a preset threshold value to obtain a stomata classification parameter; The stomatal classification parameters are hierarchically processed, the distribution characteristics of the stomatal groups are determined according to the spatial position relationship, and the regional connectivity analysis is performed on the distribution characteristic data to obtain the classification statistical results; Performing data matching on the classification statistical results and the stomatal characteristic parameters, establishing a correlation matrix of stomatal morphology and distribution through parameter mapping, performing numerical calculation on the correlation matrix, and obtaining characteristic correlation data; The feature-related data are weighted according to the spatial distribution law, an evaluation index system is constructed through multi-dimensional feature calculation, and a hierarchical analysis is performed on the evaluation index to obtain an evaluation parameter set; Numerical calculation is performed on the evaluation parameter set, and the evaluation value of the pore defect is determined by fusing multiple indicators. The evaluation value is compared and analyzed with the standard range to obtain quantitative evaluation data of the pore defect.

7. The method for detecting bubbles in precast concrete T-beams according to claim 1, characterized in that: The method of determining the warning level based on the quantitative evaluation data of the pore defects, automatically marking and visualizing the determination results, and comparing and analyzing the processing results with the quality parameters of the prefabricated components to obtain the test report data includes: Performing numerical interval analysis on the quantitative evaluation data of the pore defects, classifying the pore defects in each interval according to the distribution characteristics, establishing early warning index data for the classification results, and obtaining initial early warning parameters; The initial warning parameters are hierarchically classified according to spatial distribution, the hazard level coefficient is determined by defect degree calculation, and each level of defects is identified and mapped to obtain a warning judgment result; Performing regional annotation on the warning determination result, spatially locating the defect information through position coordinates, converting the positioning data into visual marks, and obtaining annotated data sets; Graphically processing the annotated data set, constructing a visualization interface through multi-dimensional feature expression, optimizing the layout of interface elements, and obtaining a visualization result; Performing data matching between the visualization result and the quality parameters of the prefabricated component, establishing a quality assessment association matrix through parameter mapping, and performing comprehensive calculation on the associated data to obtain quality comparison data; The quality comparison data is analyzed and sorted according to the evaluation criteria, and the detection evaluation content is generated by integrating multiple indicators. The evaluation content is formatted to obtain the detection report data.

8. A precast concrete T-beam bubble detection system, used to implement the precast concrete T-beam bubble detection method according to any one of claims 1 to 7, characterized in that: The precast concrete T-beam bubble detection system comprises: The acquisition module is used to collect multi-band reflection data from the surface of the precast concrete T-beam, perform position mapping calibration on the collected reflection data according to the calibration plate parameters, perform light intensity compensation processing and signal-to-noise ratio optimization on the calibrated reflection data of each band, select the optimal band combination by cross-validation, and obtain compensated multi-band reflection data; A registration module is used to perform spatial registration based on the compensated multi-band reflection data, perform noise reduction and smoothing on the registered data, perform principal component calculation and data integration on the smoothed multi-band data, and obtain a fused image of the reflection data; The segmentation module is used to perform edge detection and image segmentation based on the fused image of the reflection data, perform numerical conversion on the segmented image area according to the contour features, jointly analyze the converted numerical values ​​with the texture features, and use the adaptive threshold to extract the pore boundaries in a refined manner to obtain the morphological parameter data of the pores; A calculation module is used to perform depth calculation based on the morphological parameter data of the pores, perform spatial coordinate conversion on the calculated depth data, perform dual constraint reconstruction on the converted spatial coordinate data and the surface reflection intensity, and perform correction in combination with the concrete density parameter to obtain the spatial distribution data of the pores; A statistical module is used to perform numerical statistics based on the spatial distribution data of pores, classify and divide the statistical results according to a preset threshold, and associate the division results with the pore characteristic parameters to obtain quantitative evaluation data of pore defects; The determination module is used to determine the warning level based on the quantitative evaluation data of the pore defects, automatically mark and visualize the determination results, compare and analyze the processing results with the quality parameters of the prefabricated components, and obtain the test report data.

Images