A fully automated method and system for measuring and testing the strength of concrete core samples

Abstract

This invention discloses a fully automated method and system for measuring and testing the strength of concrete core samples, comprising: acquiring end-face and cylindrical surface images of the concrete core sample and extracting the boundary contours of the images; identifying end-face deformation anomalies and cylindrical surface deformation anomalies based on the boundary point set of the core sample images; constructing a core sample end-face defect identification model to identify the type and severity of core sample end-face defects; extracting core sample texture features, color features, and geometric features to construct a joint feature matrix; constructing a core sample cylindrical surface defect identification model to identify the type of core sample cylindrical surface defects; allocating core sample disposal, repair, and testing processes based on the core sample end-face and cylindrical surface deformation and defect identification information; and using an automatic centering and servo loading device to determine the compressive strength of the concrete core sample. This invention achieves an integrated, fully automated, closed-loop operation of dimensional inspection and strength testing, meeting the efficiency, consistency, and traceability requirements for large-scale engineering core sample testing.

Description

Technical Field

[0001] This invention relates to the field of civil engineering and building materials testing technology, and in particular to a fully automated method and system for measuring and testing the strength of concrete core samples. Background Technology

[0002] Concrete core sample testing is one of the most critical steps in engineering quality assessment. Its results are widely used in assessing the load-bearing capacity of concrete structures, accepting construction quality, and evaluating material properties. The core components of core sample testing include dimensional measurement, end-face quality judgment, cylindrical surface deformation inspection, and finally, compressive strength testing. However, in the current testing system, problems such as decentralization, manual processes, and a lack of unified data transfer mechanisms are prevalent, leading to low testing efficiency, poor result consistency, and difficulty in meeting the high efficiency, high precision, and high reliability requirements of engineering sites. Firstly, many core samples exhibit geometric anomalies during drilling and cutting, such as cylindrical surface eccentricity, taper, curvature, or ellipticity changes. Traditional visual inspection methods cannot quantitatively assess these deformations, resulting in insufficient objectivity. Secondly, in terms of end-face defect identification, most current testing relies on manual visual observation or simple image review. Since end-face images are easily affected by factors such as lighting, dust cover, and changes in surface humidity, defect judgment is highly subjective. Especially when distinguishing between inherent defects such as porosity and aggregate exposure, and man-made defects such as drilling deviation and improper cutting angles, existing methods lack effective evidence, making it difficult to trace the source of defects and affecting engineering quality analysis. Secondly, current engineering projects commonly separate dimensional inspection from compressive strength testing. Core samples typically require manual handling between dimensional inspection equipment, trimming equipment, and compressive strength testing machines. The lack of a unified data interface between these devices forces inspectors to manually record information such as diameter, height, and end-face conditions before inputting it into the compressive strength testing system, resulting in low data integration efficiency and a high risk of errors. When core sample batches are large, the burden of manual handling and data processing further increases, not only extending the testing cycle but also hindering continuous, large-scale engineering testing processes. Finally, when core samples require trimming, determining the trimming amount relies on manual experience, making it difficult to guarantee the stability of end-face flatness and perpendicularity. Post-trimming re-inspections also often lack objective quantitative standards, leading to uncertainty about whether the core sample meets the compressive strength test requirements. These problems collectively increase the dispersion of compressive strength results and reduce the reliability of the testing system. Therefore, current testing processes typically separate dimensional measurement, defect identification, repair and re-inspection, and compression testing into different equipment and manual processes, resulting in difficulties in process coordination, poor data correlation, and limited efficiency. There is a lack of a fully automated and integrated testing solution that can realize the entire process from sample input to test report output. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides a fully automated method and system for measuring and testing the strength of concrete core samples.

[0004] The first aspect of this invention provides a fully automated method for measuring and testing the strength of concrete core samples, mainly comprising:

[0005] By using a camera installed at the inspection station, multi-view end face images and cylindrical surface images of the concrete core sample are acquired. The boundary contours of the end face images and cylindrical surface images are extracted, and the diameter and height of the core sample are determined.

[0006] Based on the set of boundary points of the end face image and the cylindrical surface image, identify the end face deformation anomaly and the cylindrical surface deformation anomaly of the core sample;

[0007] Based on historical images of concrete core sample end faces, a core sample end face defect recognition model is constructed to identify the types and severity of defects acquired in real time.

[0008] Based on multi-view cylindrical surface images and time-coded registered 3D point cloud data, texture features, color features, and geometric features of concrete core samples are extracted, and a joint feature matrix is ​​constructed.

[0009] Based on the joint feature matrix of the cylindrical surface and the drilling and cutting process parameter records, combined with the defect type annotation information of the core sample cylindrical surface, a defect identification model for the core sample cylindrical surface is constructed to identify the defect type of the concrete core sample cylindrical surface.

[0010] Based on the diameter, end face deformation, and cylindrical surface deformation of the concrete core sample, combined with the severity of end face defects and information on human-caused defects, the process of discarding, repairing, and sending the core sample for testing is allocated.

[0011] Based on the re-inspection results of the end face and cylindrical surface of the repaired concrete core sample, combined with the loading information of the automatic centering and servo loading device, the compressive strength of the concrete core sample is determined.

[0012] Furthermore, the step of acquiring multi-view end face images and cylindrical surface images of the concrete core sample using a camera installed at the inspection station, extracting the boundary contours of the end face images and cylindrical surface images, and determining the diameter and height of the core sample includes:

[0013] A positioning and clamping device is used to limit and stabilize the concrete core samples, which are then sequentially transferred to an integrated dimensional and quality inspection station. Multi-view end-face and cylindrical surface images of the concrete core samples are acquired using a camera installed at the inspection station. The acquired images undergo preprocessing and are stored in the core sample inspection database. Image preprocessing includes histogram equalization, edge-preserving filtering, and image noise suppression. The boundary contours of the end-face and cylindrical surface images are extracted using the Canny edge detection algorithm to determine the set of boundary points. By calculating the camera extrinsic matrix and the three-dimensional point cloud data of the image pixels in space using calibration parameters, the pixel values ​​of the end-face and cylindrical surface images are converted into actual dimensions, yielding the diameter and height of each concrete core sample.

[0014] Further, the step of identifying end-face deformation anomalies and cylindrical surface deformation anomalies based on the boundary point set of the end-face image and the cylindrical surface image includes:

[0015] Based on the boundary point set of the end face image, the Hough transform method is used to fit the ideal end face contour, locate the center coordinates of the ideal end face contour, and compare them with the center coordinates of the core sample to calculate the roundness deviation of the end face. If the roundness deviation of the end face exceeds the preset first deviation threshold, the core sample end face is determined to be abnormally deformed. The perpendicularity deviation between the end face and the axis of the core sample is calculated by using the angle between the end face normal and the axis to determine whether there is a tilt phenomenon of the end face. Based on the boundary point set of the multi-view cylindrical surface image of the concrete core sample, the camera extrinsic parameter matrix and the three-dimensional point cloud data of the image pixels in space are solved by calibration parameters to construct the cylindrical surface axis point set. The global axis direction vector is obtained by least squares fitting, and the point cloud is processed by axial layering. The offset value of each layer from the axis is calculated to obtain the taper and straightness of the entire core sample. If the taper or straightness of the core sample exceeds the preset standard value, the cylindrical surface of the core sample is determined to be abnormally deformed.

[0016] Furthermore, based on historical images of the concrete core sample end face, a core sample end face defect recognition model is constructed to identify the real-time acquired concrete core sample end face defect types and severity, including:

[0017] Historical images of concrete core sample end faces are obtained through a core sample inspection database. The types and severity of defects on the core sample end faces are labeled. A convolutional neural network is used to train the model and construct a core sample end face defect recognition model. The original defect types of the end faces include honeycomb, pitting, porosity, and uneven aggregate distribution. The defect severity includes severe, moderate, and ordinary. Based on real-time acquired images of concrete core sample end faces, the core sample end face defect recognition model is used to identify the type and severity of defects on the core sample end faces.

[0018] Furthermore, the step of extracting texture, color, and geometric features of concrete core samples from multi-view cylindrical surface images and time-coded registered 3D point cloud data, and constructing a joint feature matrix, includes:

[0019] By using timestamps and camera angle encoding, the 3D point cloud data of the cylindrical surface image of the concrete core sample is registered accordingly, aligning the same circumferential position and the same height section to obtain a cylindrical surface registration multimodal dataset. Based on the cylindrical surface registration multimodal dataset, image denoising and white balance correction are used to preprocess the original image sequence to obtain the corrected cylindrical surface image. Sub-pixel edge detection is used to extract the cylindrical surface texture boundary and color gradient field, and the gray-level co-occurrence matrix, local binary mode, and orientation gradient histogram are calculated to obtain the cylindrical surface texture feature vector. The mean, variance, and spatial continuity index of the chromaticity of each height section of the core sample are statistically analyzed using the RGB channels to obtain... The color feature vector of the cylindrical surface is obtained; the cylindrical surface shape mesh is generated by removing outliers from the point cloud and reconstructing the surface, and the cylindricity error, taper error, axial curvature, ellipticity, local curvature amplitude and height distribution uniformity are calculated to obtain the geometric feature vector of the cylindrical surface; based on the texture feature vector and color feature vector of the concrete core sample cylindrical surface, the theoretical predicted value of the geometric features is predicted, the coupling perturbation coefficient is calculated, conflict samples are identified and the texture feature, color feature and geometric feature data are corrected; the texture feature vector, color feature vector and geometric feature vector of the cylindrical surface are normalized and combined into a joint feature matrix of the cylindrical surface, and stored in the core sample detection database.

[0020] It also includes predicting theoretical values ​​of geometric features based on the texture and color feature vectors of the cylindrical surface of the concrete core sample, calculating the coupling perturbation coefficient, identifying conflicting samples, and correcting the texture, color, and geometric feature data. Specifically, this includes:

[0021] By using a core sample inspection database, texture and color feature vectors of historical concrete core sample cylindrical surfaces are obtained, and geometric feature vectors are labeled. A recurrent neural network is used for model training to construct a geometric feature prediction model, predicting theoretical values ​​of geometric features. Based on the theoretical predicted values ​​of geometric feature vectors of the concrete core sample cylindrical surfaces and the actual geometric feature vectors obtained from current inspection, the coupling perturbation coefficient formula is used. Calculate the coupling perturbation coefficient ,in, This represents the actual value of the i-th geometric feature in the current sample. Let i be the predicted value of the i-th geometric feature of the current sample. The standard geometric reference value for the i-th geometric feature is obtained from the feature mean of a large number of qualified core samples in history, where n is the number of geometric feature dimensions. If the calculated coupling perturbation coefficient is greater than the preset coefficient threshold, the concrete core sample is marked as a conflict sample. Based on the marked conflict samples, high-pass filtering in the image frequency domain is used to enhance boundary details, affine registration is performed to correct texture features, white balance and color difference correction are performed on the RGB channels to optimize color features, outlier removal and B-spline interpolation are performed on the point cloud to correct local geometry, and perturbation coefficients are calculated based on the corrected texture features, color features, and geometric features to determine the effectiveness of feature correction. Based on the corrected texture features, color features, and geometric features, each feature vector is re-normalized and merged into a corrected joint feature matrix.

[0022] Furthermore, based on the joint feature matrix of the cylindrical surface and the drilling and cutting process parameter records, combined with the core sample cylindrical surface defect type annotation information, a core sample cylindrical surface defect identification model is constructed to identify the defect types of the concrete core sample cylindrical surface, including:

[0023] By using a core sample inspection database, the joint feature matrix of the cylindrical surface of historical concrete core samples and the records of drilling and cutting process parameters are obtained. The types of defects on the cylindrical surface of the core samples are labeled. A decision tree algorithm is used to train the model and construct a core sample cylindrical surface defect recognition model. The drilling and cutting process parameter records include drilling process parameters and cutting and grinding parameters. The drilling process parameters include drill bit diameter, rotation speed, feed rate, cooling method, equipment vibration spectrum, and drilling inclination angle. The cutting and grinding parameters include cutting blade wear, cutting speed, end face grinding pressure and time. The cylindrical surface defect types include man-made defects and native defects. Based on the joint feature matrix of the core sample cylindrical surface acquired in real time, the core sample cylindrical surface defect type recognition model is used to identify the defect types of the concrete core sample cylindrical surface.

[0024] Furthermore, the process of allocating core sample disposal, repair, and testing based on the concrete core sample diameter, end face deformation, cylindrical surface deformation, and the severity of end face defects and information on human-caused defects includes:

[0025] Using a positioning and clamping device, unqualified concrete core samples with diameters that do not meet the preset diameter requirements, abnormal end-face deformation or abnormal cylindrical surface deformation, severe end-face defects, or human-caused defects are conveyed to the waste core sample conveyor belt for waste disposal. Using the positioning and clamping device, concrete core samples that do not meet the preset height requirements, exhibit tilting, or have end-face defects that are not severe are conveyed to the core sample trimming station, while the remaining qualified core samples are conveyed to the compressive strength testing unit. Based on the measured concrete core sample height and the perpendicularity deviation between the end face and the axis, the required grinding amount is calculated, and the end face of the core sample is trimmed. Automatic spraying or heat pressing methods are used to trim the end-face defects.

[0026] Furthermore, the determination of the compressive strength of the concrete core sample based on the re-inspection results of the end face and cylindrical surface of the trimmed concrete core sample, combined with the loading information of the automatic centering and servo loading device, includes:

[0027] Using a positioning and clamping device, the modified concrete core sample is transferred to an integrated dimensional and quality inspection station for re-inspection of the end face and cylindrical surface. If the inspection is qualified, it is transferred to the compressive strength testing unit. Using an automatic centering and servo loading device, the loading axis is aligned with the core sample axis to conduct a loading test, and load and displacement data are collected in real time. By calculating the core sample failure load and end face area, the compressive strength data of the concrete core sample is obtained, and a concrete core sample test report and core sample strength index are generated.

[0028] A second aspect of the present invention provides a fully automated concrete core sample measurement and strength testing system, mainly comprising:

[0029] The core sample image acquisition module is used to acquire multi-view end face images and cylindrical surface images of concrete core samples through a camera installed at the inspection station, extract the boundary contours of the end face images and cylindrical surface images, and determine the diameter and height of the core sample.

[0030] The core sample deformation anomaly identification module is used to identify core sample end face deformation anomalies and cylindrical surface deformation anomalies based on the boundary point set of the end face image and the cylindrical surface image;

[0031] The end-face defect recognition module is used to construct a core sample end-face defect recognition model based on historical images of the concrete core sample end face, and to identify the type and severity of defects in the concrete core sample end face acquired in real time.

[0032] The cylindrical surface feature analysis module is used to extract the texture features, color features, and geometric features of concrete core samples based on multi-view cylindrical surface images and time-coded registered 3D point cloud data, and to construct a joint feature matrix.

[0033] The cylindrical surface defect recognition module is used to construct a core sample cylindrical surface defect recognition model based on the joint feature matrix of the cylindrical surface and the drilling and cutting process parameter records, combined with the core sample cylindrical surface defect type annotation information, to identify the defect type of the concrete core sample cylindrical surface;

[0034] The core sample screening and process allocation module is used to allocate the core sample rejection, repair and testing process based on the concrete core sample diameter, end face deformation, cylindrical surface deformation, combined with the severity of end face defects and human defect judgment information.

[0035] The core sample compressive strength testing module is used to determine the compressive strength of concrete core samples based on the re-inspection results of the end face and cylindrical surface of the repaired concrete core sample, combined with the loading information of the automatic centering and servo loading device.

[0036] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0037] This invention provides a fully automated method and system for measuring and testing the strength of concrete core samples. By acquiring multi-view end-face and cylindrical surface images and extracting boundary contours, this invention can stably obtain the diameter, height, and deformation information of core samples under different lighting and surface quality conditions, reducing errors caused by manual measurement. The defect identification mechanism constructed using historical end-face images can accurately distinguish the type and severity of end-face defects, improving the objectivity of end-face quality assessment. Based on the extraction of texture, color, and geometric features from 3D point cloud data with multi-view images and time-coded registration, and combined with geometric theory predictions for coupled perturbation analysis, this invention can automatically correct conflicting features between images and point clouds, thereby improving the stability and consistency of the joint features of the cylindrical surface. The defect type identification mechanism constructed using a joint feature matrix combined with drilling and cutting process parameters improves the traceability of defect attribution, thereby distinguishing between human-caused defects and native defects, enhancing the continuity of the testing process, the accuracy of results, and the ability to manage data in a correlated manner. This invention reduces process fragmentation and recording errors caused by multi-device transfer and manual data docking, strengthens the correlation and traceability of multimodal features and process information, and improves the consistency of strength test conditions through re-inspection closed loop and centering loading constraints, thereby improving overall testing efficiency and result consistency, and meeting the efficiency, consistency and traceability requirements of large-scale engineering core sample testing. Attached Figure Description

[0038] Figure 1 This is a flowchart of a fully automated method for measuring and testing the strength of concrete core samples according to the present invention;

[0039] Figure 2 This is a schematic diagram of a fully automated method for measuring and testing the strength of concrete core samples according to the present invention;

[0040] Figure 3This is a schematic diagram of a fully automatic measurement and strength testing system for concrete core samples according to the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] like Figures 1-2 This embodiment of a fully automated method for measuring and testing the strength of concrete core samples may specifically include:

[0043] Step S101: Using a camera installed at the inspection station, acquire multi-view end face images and cylindrical surface images of the concrete core sample, extract the boundary contours of the end face images and cylindrical surface images, and determine the diameter and height of the core sample.

[0044] A positioning and clamping device is used to limit and stabilize the concrete core samples, which are then sequentially transferred to an integrated dimensional and quality inspection station. Multi-view images of the end face and cylindrical surface of the concrete core samples are acquired using a camera installed at the inspection station. These images undergo preprocessing and are stored in the core sample inspection database. Image preprocessing includes histogram equalization, edge-preserving filtering, and image noise suppression. The boundary contours of the end face and cylindrical surface images are extracted using the Canny edge detection algorithm to determine the set of boundary points. By calibrating parameters, the camera extrinsic matrix and the 3D point cloud data of the image pixels in space are calculated. The pixel values ​​of the end face and cylindrical surface images are then converted into actual dimensions to obtain the diameter and height of each concrete core sample.

[0045] For example, when inspecting a concrete core sample, the operator places it in a positioning clamping device for limiting and fixing, keeping it stable during the inspection process. It is then moved by a conveying mechanism to an integrated dimensional and quality inspection station. The core sample is approximately 150.2 mm long and 74.5 mm in diameter, and is cylindrical in shape. The inspection station is equipped with two high-resolution industrial cameras arranged in a circular pattern, acquiring end-face and cylindrical surface images of the core sample at 30° and 60° angles, respectively. After image acquisition, the images are preprocessed and stored in the core sample inspection database. Image preprocessing includes performing histogram equalization to enhance image contrast, using an edge-preserving filtering algorithm to suppress texture noise while preserving boundary details, and finally applying medium-level filtering to remove isolated noise points. The Canny edge detection algorithm is used to extract boundary lines from the image, obtaining the closed contour curve of the core sample end face image. Vertical edge information is extracted from the cylindrical surface image to obtain a set of boundary points. This set of boundary points refers to the set of pixels representing the cylindrical edge contour in the image. For example, the detected radius of the end face contour is 124.7 pixels in pixel coordinates, while the height span of the cylindrical surface is 418.5 pixels. Based on the intrinsic and extrinsic parameter matrices obtained from previous camera calibration, the spatial geometric relationship between the camera and the sample is calculated, and the pixel coordinates are converted into actual physical coordinates in three-dimensional space. If the camera calibration parameters are: pixel size 0.05 mm / pixel, distance from the image sensor to the object 250 mm, and focal length 8 mm. After spatial calibration conversion, the end face radius of the aforementioned 124.7 pixels was converted to 62.35 mm, and the calculated diameter of the core sample was 124.7 × 0.05 × 2 = 12.47 mm. However, the actual conversion factor of the core sample was different, so the final conversion was a diameter of 74.5 mm and a height of 150.2 mm. The dimensions of the concrete core sample were then stored in the core sample detection database.

[0046] Step S102: Identify the core sample end face deformation anomaly and cylindrical surface deformation anomaly based on the boundary point set of the end face image and the cylindrical surface image.

[0047] Based on the boundary point set of the end face image, the Hough transform method is used to fit the ideal end face contour, locate the center coordinates of the ideal end face contour, and compare them with the center coordinates of the core sample to calculate the roundness deviation of the end face. If the roundness deviation of the end face exceeds the preset first deviation threshold, the core sample end face is judged to be abnormally deformed. The perpendicularity deviation between the end face and the axis of the core sample is calculated by using the angle between the end face normal and the axis to determine whether there is a tilt phenomenon of the end face. Based on the boundary point set of the multi-view cylindrical surface image of the concrete core sample, the camera extrinsic parameter matrix and the three-dimensional point cloud data of the image pixels in space are solved by calibration parameters to construct the cylindrical surface axis point set. The global axis direction vector is obtained by least squares fitting, and the point cloud is processed by axial layering. The offset value of each layer from the axis is calculated to obtain the taper and straightness of the entire core sample. If the taper or straightness of the core sample exceeds the preset standard value, the cylindrical surface of the core sample is judged to be abnormally deformed.

[0048] For example, an end-face image of the core sample is acquired using a camera, and an ideal circular end-face contour is fitted using the Hough transform method. The coordinates of the ideal end-face center obtained by this method are (125.6 mm, 128.3 mm), while the coordinates of the end-face center calculated from the actual boundary point set extracted from the image are (125.8 mm, 128.5 mm). By calculating the deviation between these two center coordinates, the end-face roundness deviation is found to be 0.2 mm. If the preset roundness deviation threshold is 0.1 mm, this deviation obviously exceeds the threshold, therefore the end-face of the core sample is determined to be abnormally deformed. Image analysis calculates the angle between the normal vector of the end face and the core sample axis to be 2.3°, while the preset maximum perpendicularity deviation is 2°. Since the calculated angle exceeds the preset perpendicularity deviation value, the end-face of the core sample is determined to be tilted. By acquiring multi-view images of the cylindrical surface, and combining camera calibration parameters and extrinsic matrix, the 3D point cloud data in the images is calculated, and the axis point set on the cylindrical surface is extracted. If the global axis direction vector obtained by least squares fitting is (0.998, 0.056, 0.000), and the point cloud data is processed by axial layering, the axis offset values ​​of each slice of the cylindrical surface are obtained. When calculating the 10th layer of the core sample, the offset value is 0.6 mm, while the offset value of the 15th layer is 0.3 mm. By calculating the average offset value of each layer, the taper of the core sample is found to be 0.45 mm, and the straightness is 0.35 mm. If the preset standard value for taper is 0.3 mm and the standard value for straightness is 0.25 mm, then the taper and straightness of the core sample both exceed the standard values, therefore the cylindrical surface of the core sample is judged to be abnormally deformed.

[0049] Step S103: Based on historical images of the concrete core sample end face, construct a core sample end face defect recognition model to identify the type and severity of defects in the concrete core sample end face acquired in real time.

[0050] Historical images of concrete core sample end faces are obtained from a core sample inspection database. The types and severity of defects on the core sample end faces are labeled. A convolutional neural network is used to train a model to construct a core sample end face defect recognition model. The original defect types on the end faces include honeycombing, pitting, porosity, and uneven aggregate distribution. Defect severity is categorized as severe, moderate, and ordinary. Based on real-time acquired images of concrete core sample end faces, the core sample end face defect recognition model is used to identify the type and severity of defects.

[0051] For example, historical images of concrete core sample end faces are obtained through a core sample inspection database. If the historical images show multiple defects on the core sample end face, including honeycomb, pitting, porosity, and uneven aggregate distribution, and these defects are manually labeled as moderate or moderate in severity, a convolutional neural network is used to train a model to construct a core sample end face defect recognition model. During training, the model learns the characteristics of different defect types. For example, honeycomb typically manifests as irregular surface voids, while pitting manifests as relatively uniform small holes or inconsistent textures, porosity is a large void area, and uneven aggregate distribution manifests as abnormal distribution of coarse aggregate. During real-time detection, based on the acquired current core sample end face image, the trained core sample end face defect recognition model is used to identify the type and severity of the core sample end face defects. Two defects are identified in the image: the first is a porosity with an area of ​​5 square millimeters, which the model assesses as moderate in severity; the second is a honeycomb with an area of ​​2 square millimeters, which the model assesses as moderate in severity.

[0052] Step S104: Based on the multi-view cylindrical surface image and the time-coded registered 3D point cloud data, extract the texture features, color features and geometric features of the concrete core sample, and construct a joint feature matrix.

[0053] By using timestamps and camera angle encoding, the 3D point cloud data of the cylindrical surface image of the concrete core sample is registered accordingly, aligning the same circumferential position and the same height section to obtain a cylindrical surface registration multimodal dataset. Based on the cylindrical surface registration multimodal dataset, image denoising and white balance correction are used to preprocess the original image sequence to obtain the corrected cylindrical surface image. Sub-pixel edge detection is used to extract the cylindrical surface texture boundary and color gradient field, and the gray-level co-occurrence matrix, local binary mode, and directional gradient histogram are calculated to obtain the cylindrical surface texture feature vector. The chromaticity mean, variance, and spatial continuity index of each height section of the core sample are statistically analyzed using the RGB channels to obtain the cylindrical surface color feature vector. The cylindrical surface shape mesh is generated by point cloud outlier removal and surface reconstruction, and the cylindricity error, taper error, axial curvature, ellipticity, local curvature amplitude, and height distribution uniformity are calculated to obtain the cylindrical surface geometric feature vector. Based on the texture and color feature vectors of the cylindrical surface of the concrete core sample, the theoretical predicted values ​​of the geometric features are predicted, the coupling perturbation coefficient is calculated, conflicting samples are identified, and the texture, color, and geometric feature data are corrected. The texture, color, and geometric feature vectors of the cylindrical surface are normalized, combined into a joint feature matrix of the cylindrical surface, and stored in the core sample detection database.

[0054] For example, the core sample is fixed by a positioning clamping device, and two cameras installed at the inspection station capture images of the cylindrical surface from different angles. The core sample has a height of 150 mm and a diameter of 75 mm. Based on the camera timestamps and angle codes, multimodal data registration aligns the images and point cloud data from different perspectives. The system ensures that the images and point cloud data are aligned on the same height section, resulting in clear cylindrical surface images and 3D point cloud data. The acquired images are preprocessed, including white balance correction and noise reduction, to enhance image details and remove illumination variations and noise interference, yielding a clear cylindrical surface image. A subpixel edge detection algorithm was used to extract the boundary of the cylindrical surface texture, and the gray-level co-occurrence matrix, local binary mode, and histogram of directional gradients of the image were calculated. The texture feature vector of the core sample was obtained as [0.38, 0.45, 0.32], representing the uniformity, contrast, and texture orientation of gray levels in the image. The gray-level co-occurrence matrix features include contrast, energy, and entropy, totaling 3 dimensions. The local binary mode features are the mean histogram values, totaling 1 dimension. The histogram of directional gradients features include the principal direction and the directional variance, totaling 2 dimensions. The chromaticity mean, variance, and spatial continuity of each height section were statistically analyzed using the RGB channels, and the color feature vector was extracted as [0.22, 0.18, 0.25], representing the uniformity, variance, and color gradient information of the color change on the core sample surface. Based on point cloud data, outlier removal and surface reconstruction were performed to obtain the cylindrical shape mesh of the core sample. Through geometric analysis, the cylindricity error of the core sample was calculated to be 0.03 mm, the taper error to be 0.02 mm, the axial curvature to be 0.15 mm, the ellipticity to be 0.98, and the local curvature amplitude to be 0.01 mm, indicating a slight deviation in the geometric shape of the core sample. However, certain geometric defects, such as local curvature fluctuations, local protrusions, or depressions, can alter the surface reflection state and color gradient distribution, interfering with the stability of texture details and chromaticity fields. This causes the model to misidentify normal texture and color changes caused by lighting, construction marks, etc., as defects on the cylindrical surface. Conversely, certain texture and color anomalies, such as large-area honeycombing, severe pitting, color difference bands caused by pollution, or local discoloration areas, can affect the stability of image boundary extraction and point cloud contour reconstruction. This causes geometric features such as cylindricity error, axial curvature, ellipticity, and local curvature amplitude to deviate from the true geometric state, thus introducing non-realistic geometric deviation samples into the joint feature space. Therefore, it is necessary to combine texture and color features to predict the theoretical values ​​of the geometric features of the core sample and calculate the coupling perturbation coefficient. If the calculated perturbation coefficient is 0.07, which is lower than the preset threshold of 0.2, it indicates that the consistency between texture features and geometric features is good and does not meet the standard of conflicting samples, so no further correction is needed.The texture features, color features, and geometric feature vectors were normalized to generate a unified joint feature matrix [0.38, 0.45, 0.32, 0.22, 0.18, 0.25, 0.03, 0.02, 0.15], and this feature matrix was stored in the core sample detection database.

[0055] Based on the texture and color feature vectors of the cylindrical surface of the concrete core sample, the theoretical predicted values ​​of the geometric features are predicted, the coupling perturbation coefficient is calculated, conflict samples are identified, and the texture, color, and geometric feature data are corrected.

[0056] By using a core sample inspection database, texture and color feature vectors of historical concrete core sample cylindrical surfaces are obtained, and geometric feature vectors are labeled. A recurrent neural network is used for model training to construct a geometric feature prediction model, predicting theoretical values ​​for geometric features. Based on the theoretical predicted values ​​of the geometric feature vectors of the concrete core sample cylindrical surfaces and the actual geometric feature vectors obtained from current inspection, a coupling perturbation coefficient formula is used. Calculate the coupling perturbation coefficient ,in, This represents the actual value of the i-th geometric feature in the current sample. Let i be the predicted value of the i-th geometric feature of the current sample. Let be the standard geometric reference value for the i-th geometric feature, obtained from the feature mean of a large number of qualified core samples in history, where n is the number of geometric feature dimensions. If the calculated coupling perturbation coefficient is greater than a preset threshold, the concrete core sample is marked as a conflict sample. Based on the marked conflict samples, high-pass filtering in the image frequency domain is used to enhance boundary details, affine registration is performed to correct texture features, white balance and color difference correction is performed on the RGB channels to optimize color features, outlier removal and B-spline interpolation are performed on the point cloud to correct local geometry, and perturbation coefficients are calculated based on the corrected texture, color, and geometric feature data to determine the effectiveness of feature correction. Based on the corrected texture, color, and geometric feature data, each feature vector is re-normalized and merged into a corrected joint feature matrix.

[0057] For example, historical data, including labeled texture and color feature vectors, and corresponding geometric feature vectors, is retrieved from the core sample inspection database. A recurrent neural network is used to train the model, constructing a geometric feature prediction model to predict the theoretical values ​​of the geometric features. In real-time inspection, after an image of a core sample is acquired and multimodal registration is completed, its cylindrical surface texture feature vector T=[0.35, 0.21, 0.65, 0.48, 0.72, 0.19] and color feature vector C=[0.42, 0.18, 0.55] are extracted. The geometric feature prediction model is then used to predict the geometric feature values. Given [0.018, 0.031, 0.005, 0.96, 0.004, 0.89], the actual geometric feature value G of the core sample is calculated as [0.024, 0.036, 0.007, 0.94, 0.006, 0.80] using point cloud data. The mean geometric feature value [0.020, 0.030, 0.004, 1.00, 0.005, 0.85] of qualified samples is retrieved from the core sample detection database as a reference value, and the coupling perturbation coefficient formula is used. The coupling perturbation coefficient was calculated. It is 0.249, of which, This represents the actual value of the i-th geometric feature in the current sample. Let i be the predicted value of the i-th geometric feature of the current sample. The standard geometric reference value for the i-th geometric feature is obtained from the feature mean of a large number of qualified core samples from history, where n is the number of geometric feature dimensions. If the preset coefficient threshold is 0.2, and 0.249 > 0.20, then the sample is marked as a conflict sample. Frequency domain high-pass filtering and affine registration are performed on the image to correct edge blurring and alignment deviations. White balance adjustment and color difference correction are performed on the RGB channels. Outlier removal and B-spline surface fitting were performed on the point cloud. The corrected features were extracted again, and the corrected actual geometric features were obtained as [0.021, 0.032, 0.005, 0.98, 0.005, 0.84]. The corrected predicted geometric features were [0.020, 0.030, 0.004, 0.99, 0.005, 0.85]. The corrected perturbation coefficient was calculated to be 0.065 < 0.20, so the correction was deemed effective. Each normalized feature vector was then merged into a corrected joint feature matrix and stored in the core sample detection database.

[0058] Step S105: Based on the joint feature matrix of the cylindrical surface and the drilling and cutting process parameter records, combined with the defect type annotation information of the core sample cylindrical surface, construct a core sample cylindrical surface defect identification model to identify the defect type of the concrete core sample cylindrical surface.

[0059] By utilizing a core sample inspection database, the joint feature matrix of the cylindrical surface of historical concrete core samples and records of drilling and cutting process parameters were obtained. The types of defects on the cylindrical surface of the core samples were labeled. A decision tree algorithm was used to train the model, constructing a core sample cylindrical surface defect identification model. The drilling and cutting process parameter records included drilling process parameters and cutting and grinding parameters. Drilling process parameters included drill bit diameter, rotation speed, feed rate, cooling method, equipment vibration spectrum, and drilling inclination angle. Cutting and grinding parameters included cutting blade wear, cutting speed, end face grinding pressure, and time. Cylindrical surface defect types included man-made defects and native defects. Based on the real-time acquired joint feature matrix of the core sample cylindrical surface, the core sample cylindrical surface defect type identification model was used to identify the defect types of the concrete core sample cylindrical surface.

[0060] For example, a batch of historical core samples' joint feature matrix and drilling and cutting process parameter data were retrieved from the core sample inspection database. In the historical samples, the cylindrical surface defect type has been marked by experts as either a native defect or a man-made defect. One set of historical samples has texture features of [0.33, 0.26, 0.58, 0.47, 0.61, 0.20], color features of [0.39, 0.12, 0.55], and geometric features of [0.023, 0.028, 0]. [005, 0.95, 0.008, 0.80], the drilling and cutting process parameters of this core sample include a drill bit diameter of 76 mm, a rotation speed of 800 rpm, a feed rate of 17 mm / min, water cooling, a root mean square value of the equipment vibration spectrum of 0.26 g, and a drilling inclination angle of 1.5 degrees. The cutting and grinding process recorded a blade wear of 0.42, a cutting speed of 1.9 m / min, an end face grinding pressure of 30 N, and a grinding time of 50 seconds. Based on the joint feature matrix of historical core samples and the drilling and cutting process parameter data, a decision tree algorithm was used to train the model and construct a core sample cylindrical surface defect recognition model. The joint feature matrix of the newly acquired core sample and the process parameters were input into the core sample cylindrical surface defect type identification model to determine whether defects exist on the cylindrical surface of the core sample and the nature of the defects. During the analysis, the model detected that the local curvature amplitude of the core sample was too high and the axial curvature was close to the boundary threshold. Combined with the high vibration spectrum of 0.26g and the large drilling inclination angle of 1.5°, it was inferred that the deformation was most likely caused by mechanical disturbance during the drilling equipment operation. In addition, the large HOG direction variance of 0.17 also suggested that the texture direction was discontinuous, which is common in structural anomalies caused by artificial cutting. After comprehensive judgment, the model determined that the core sample had artificial defects with a confidence level of 92.6%, and wrote the judgment result along with the complete features into the core sample detection database.

[0061] Step S106: Based on the diameter, end face deformation, and cylindrical surface deformation of the concrete core sample, combined with the severity of end face defects and information on human-caused defects, allocate the core sample disposal, repair, and testing process.

[0062] Using a positioning and clamping device, unqualified concrete core samples with diameters that do not meet the preset diameter requirements, abnormal end-face deformation or cylindrical surface deformation, severe end-face defects, or human-caused defects are conveyed to the waste core sample conveyor belt for waste disposal. Using the same positioning and clamping device, concrete core samples that do not meet the preset height requirements, exhibit tilting, or have end-face defects of a lesser severity are conveyed to the core sample trimming station. The remaining qualified core samples are then conveyed to the compressive strength testing unit. Based on the measured concrete core sample height and the perpendicularity deviation between the end face and the axis, the required grinding amount is calculated, and the core sample end face is trimmed. Automatic spraying or heat-pressing methods are used to trim the end-face defects.

[0063] For example, the core sample numbered MX-307 is automatically fixed in the positioning and clamping device, and the dimensional detection, deformation recognition, image recognition, and defect classification are completed sequentially. The detection results show that the diameter of the core sample is 72.1 mm, which is lower than the preset minimum diameter requirement of 74 mm. The end face image analysis shows that its roundness deviation is 1.95 mm, which exceeds the set threshold of 1.5 mm, and is judged to be an abnormal end face deformation. In addition, the core sample end face defect recognition model determines that the core sample end face has a composite defect of honeycomb and pores, with a severity level of severe. The core sample cylindrical surface defect type recognition model identifies that the core sample has cutting tilt and edge chipping caused by equipment vibration, and judges it to be a human-caused defect. Since the core sample meets all four rejection criteria of not meeting the diameter requirement, abnormal end face deformation, severe defect level, and human-caused defect, it is marked as a non-conforming core sample and transferred to the waste core sample conveyor belt through the positioning and clamping device for waste disposal. In the same batch, core sample MX-312, while meeting the specifications for diameter and cylindricity, had a measured total height of 142.6 mm, outside the required height range of 150 ± 2 mm. Its end face angle with the axis was calculated to be 2.8 degrees, exceeding the set upper limit of 2 degrees for perpendicularity deviation. Furthermore, the end face image showed slight surface roughness and localized color difference. The model determined the defect severity to be moderate, not constituting a rejection standard, but affecting the accuracy of subsequent compressive strength tests. Therefore, this core sample was deemed to require trimming and was sent to the core sample trimming station via a clamping device. At the trimming station, based on the end face tilt angle of 2.8 degrees and the current height gap of 7.4 mm, the required trimming amount was calculated to be 8.2 mm relative to the 150 mm standard. Considering material springback and end face unevenness during grinding, the actual grinding amount was set to 8.5 mm. Subsequently, an automated grinding device was activated to perform planar grinding on the end face of the core sample. Simultaneously, RGB analysis detected color difference bands on the end face, which were automatically covered and repaired using a spraying device, completing the color difference correction. After repair, the core sample was remeasured and its deformation correction was confirmed to be acceptable. It was then transported to the compressive strength testing unit by a conveyor. Other qualified core samples in this batch, such as MX-310, met all standards for size, geometry, end face quality, and defect identification. They required no repair and were directly transferred to the compressive strength testing process for the next stage of performance evaluation after joint feature archiving was completed.

[0064] Step S107: Based on the re-inspection results of the end face and cylindrical surface of the repaired concrete core sample, and combined with the loading information of the automatic centering and servo loading device, the compressive strength of the concrete core sample is determined.

[0065] A positioning and clamping device is used to transfer the shaped concrete core sample to an integrated dimensional and quality inspection station for re-inspection of the end face and cylindrical surface. If the inspection is qualified, it is transferred to the compressive strength testing unit. Using an automatic centering and servo loading device, the loading axis is aligned with the core sample axis for loading tests, and load and displacement data are collected in real time. By calculating the core sample failure load and end face area, the compressive strength data of the concrete core sample is obtained, and a concrete core sample test report and core sample strength index are generated.

[0066] For example, after the end face of concrete core sample CX-118 was trimmed, it was transferred to an integrated dimensional and quality inspection station via a positioning and clamping device for re-inspection of its end face flatness, cylindrical deformation, and dimensional parameters. The re-inspection results showed that the trimmed core sample had a height of 149.8 mm, meeting the set height range requirement of 150 ± 2 mm; the perpendicularity deviation between the end face and the axis was 1.3 degrees, below the allowable upper limit of 2 degrees; the cylindricity error was 0.018 mm; the taper error was 0.027 mm; and the axial curvature was 0.004 mm, all within the acceptable thresholds. Therefore, the core sample was determined to have met the quality standards after trimming and was marked as suitable for strength testing. Subsequently, the core sample was automatically transferred to the compressive strength testing unit. In this unit, an automatic alignment device aligned the loading axis with the core sample axis to ensure consistent force distribution. A servo loading system then applied pressure at a constant speed until the core sample failed. Throughout the loading process, load and displacement data were recorded in real time. Ultimately, the maximum load the core sample withstood upon failure was 141.6 kN. Measurements showed the core sample's end face diameter was 75.0 mm, resulting in an end face area of ​​4417.9 square millimeters. Based on the compressive strength calculation formula, the core sample's compressive strength was determined to be 32.05 MPa. This result was automatically entered into the core sample testing report, which also simultaneously generated a load-displacement curve, image recordings, a 3D point cloud model, and various structural indicators for subsequent quality traceability and concrete performance evaluation.

[0067] like Figure 3 This embodiment discloses a fully automated concrete core sample measurement and strength testing system, which may specifically include:

[0068] The core sample image acquisition module is used to acquire multi-view end face images and cylindrical surface images of concrete core samples through a camera installed at the inspection station, extract the boundary contours of the end face images and cylindrical surface images, and determine the diameter and height of the core sample.

[0069] The core sample deformation anomaly identification module is used to identify core sample end face deformation anomalies and cylindrical surface deformation anomalies based on the boundary point set of the end face image and the cylindrical surface image.

[0070] The end-face defect recognition module is used to construct a core sample end-face defect recognition model based on historical images of the concrete core sample end face, and to identify the type and severity of defects in the concrete core sample end face acquired in real time.

[0071] The cylindrical surface feature analysis module is used to extract the texture, color, and geometric features of concrete core samples based on multi-view cylindrical surface images and time-coded registered 3D point cloud data, and to construct a joint feature matrix.

[0072] The cylindrical surface defect recognition module is used to construct a core sample cylindrical surface defect recognition model based on the joint feature matrix of the cylindrical surface and the drilling and cutting process parameter records, combined with the core sample cylindrical surface defect type annotation information, to identify the defect type of the concrete core sample cylindrical surface.

[0073] The core sample screening and process allocation module is used to allocate the core sample rejection, repair and testing processes based on the concrete core sample diameter, end face deformation, cylindrical surface deformation, combined with the severity of end face defects and human defect judgment information.

[0074] The core sample compressive strength testing module is used to determine the compressive strength of concrete core samples based on the re-inspection results of the end face and cylindrical surface of the repaired concrete core sample, combined with the loading information of the automatic centering and servo loading device.

[0075] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the concept of this application. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A fully automated method for measuring and testing the strength of concrete core samples, characterized in that, The method includes: By using a camera installed at the inspection station, multi-view end face images and cylindrical surface images of the concrete core sample are acquired. The boundary contours of the end face images and cylindrical surface images are extracted, and the diameter and height of the core sample are determined. Based on the set of boundary points of the end face image and the cylindrical surface image, identify the end face deformation anomaly and the cylindrical surface deformation anomaly of the core sample; Based on historical images of concrete core sample end faces, a core sample end face defect recognition model is constructed to identify the types and severity of defects acquired in real time. Based on multi-view cylindrical surface images and time-coded registered 3D point cloud data, texture features, color features, and geometric features of concrete core samples are extracted, and a joint feature matrix is ​​constructed. Based on the joint feature matrix of the cylindrical surface and the drilling and cutting process parameter records, combined with the defect type annotation information of the core sample cylindrical surface, a defect identification model for the core sample cylindrical surface is constructed to identify the defect type of the concrete core sample cylindrical surface. Based on the diameter, end face deformation, and cylindrical surface deformation of the concrete core sample, combined with the severity of end face defects and information on human-caused defects, the process of discarding, repairing, and sending the core sample for testing is allocated. Based on the re-inspection results of the end face and cylindrical surface of the repaired concrete core sample, combined with the loading information of the automatic centering and servo loading device, the compressive strength of the concrete core sample is determined. The step of extracting texture, color, and geometric features of concrete core samples from multi-view cylindrical surface images and time-coded registered 3D point cloud data, and constructing a joint feature matrix, includes: By using timestamps and camera angle encoding, the 3D point cloud data of the cylindrical surface image of the concrete core sample is registered accordingly, aligning the same circumferential position and the same height section to obtain a cylindrical surface registration multimodal dataset. Based on the cylindrical surface registration multimodal dataset, image denoising and white balance correction are used to preprocess the original image sequence to obtain the corrected cylindrical surface image. Sub-pixel edge detection is used to extract the cylindrical surface texture boundary and color gradient field, and the gray-level co-occurrence matrix, local binary mode, and orientation gradient histogram are calculated to obtain the cylindrical surface texture feature vector. The mean, variance, and spatial continuity index of the chromaticity of each height section of the core sample are statistically analyzed using the RGB channels to obtain... The process begins by obtaining the color feature vector of the cylindrical surface. Outlier removal from the point cloud and surface reconstruction generate the cylindrical surface mesh. Cylindricity error, taper error, axial curvature, ellipticity, local curvature amplitude, and height distribution uniformity are calculated to obtain the cylindrical surface geometric feature vector. Based on the texture and color feature vectors of the concrete core sample's cylindrical surface, theoretical predictions of geometric features are made. Coupling perturbation coefficients are calculated, conflicting samples are identified, and texture, color, and geometric feature data are corrected. Finally, the texture, color, and geometric feature vectors of the cylindrical surface are normalized, combined into a joint feature matrix, and stored in the core sample detection database. The process of predicting theoretical values ​​of geometric features based on the texture and color feature vectors of the cylindrical surface of the concrete core sample, calculating the coupling perturbation coefficient, identifying conflicting samples, and correcting the texture, color, and geometric feature data includes: By using a core sample inspection database, texture and color feature vectors of historical concrete core sample cylindrical surfaces are obtained, and geometric feature vectors are labeled. A recurrent neural network is used for model training to construct a geometric feature prediction model, predicting theoretical values ​​of geometric features. Based on the theoretical predicted values ​​of geometric feature vectors of the concrete core sample cylindrical surfaces and the actual geometric feature vectors obtained from current inspection, the coupling perturbation coefficient formula is used. Calculate the coupling perturbation coefficient ,in, This represents the actual value of the i-th geometric feature in the current sample. Let i be the predicted value of the i-th geometric feature of the current sample. The standard geometric reference value for the i-th geometric feature is obtained from the feature mean of a large number of qualified core samples in history, where n is the number of geometric feature dimensions. If the calculated coupling perturbation coefficient is greater than the preset coefficient threshold, the concrete core sample is marked as a conflict sample. Based on the marked conflict samples, high-pass filtering in the image frequency domain is used to enhance boundary details, affine registration is performed to correct texture features, white balance and color difference correction are performed on the RGB channels to optimize color features, outlier removal and B-spline interpolation are performed on the point cloud to correct local geometry, and perturbation coefficients are calculated based on the corrected texture features, color features, and geometric features to determine the effectiveness of feature correction. Based on the corrected texture features, color features, and geometric features, each feature vector is re-normalized and merged into a corrected joint feature matrix.

2. The method according to claim 1, characterized in that, The process involves acquiring multi-view end-face and cylindrical images of the concrete core sample using a camera installed at the inspection station, extracting the boundary contours of the end-face and cylindrical images, and determining the diameter and height of the core sample, including: A positioning and clamping device is used to limit and stabilize the concrete core samples, which are then sequentially transferred to an integrated dimensional and quality inspection station. Multi-view end-face and cylindrical surface images of the concrete core samples are acquired using a camera installed at the inspection station. The acquired images undergo preprocessing and are stored in the core sample inspection database. Image preprocessing includes histogram equalization, edge-preserving filtering, and image noise suppression. The boundary contours of the end-face and cylindrical surface images are extracted using the Canny edge detection algorithm to determine the set of boundary points. By calculating the camera extrinsic matrix and the three-dimensional point cloud data of the image pixels in space using calibration parameters, the pixel values ​​of the end-face and cylindrical surface images are converted into actual dimensions, yielding the diameter and height of each concrete core sample.

3. The method according to claim 1, characterized in that, The step of identifying end-face deformation anomalies and cylindrical surface deformation anomalies based on the boundary point set of the end-face image and the cylindrical surface image includes: Based on the boundary point set of the end face image, the Hough transform method is used to fit the ideal end face contour, locate the center coordinates of the ideal end face contour, and compare them with the center coordinates of the core sample to calculate the roundness deviation of the end face. If the roundness deviation of the end face exceeds the preset first deviation threshold, the core sample end face is determined to be abnormally deformed. The perpendicularity deviation between the end face and the axis of the core sample is calculated by using the angle between the end face normal and the axis to determine whether there is a tilt phenomenon of the end face. Based on the boundary point set of the multi-view cylindrical surface image of the concrete core sample, the camera extrinsic parameter matrix and the three-dimensional point cloud data of the image pixels in space are solved by calibration parameters to construct the cylindrical surface axis point set. The global axis direction vector is obtained by least squares fitting, and the point cloud is processed by axial layering. The offset value of each layer from the axis is calculated to obtain the taper and straightness of the entire core sample. If the taper or straightness of the core sample exceeds the preset standard value, the cylindrical surface of the core sample is determined to be abnormally deformed.

4. The method according to claim 1, characterized in that, The process involves constructing a core sample end-face defect identification model based on historical images of the concrete core sample end face, identifying the types and severity of defects acquired in real time, including: Historical images of concrete core sample end faces are obtained through a core sample inspection database. The types and severity of defects on the core sample end faces are labeled. A convolutional neural network is used to train the model and construct a core sample end face defect recognition model. The original defect types of the end faces include honeycomb, pitting, porosity, and uneven aggregate distribution. The defect severity includes severe, moderate, and ordinary. Based on real-time acquired images of concrete core sample end faces, the core sample end face defect recognition model is used to identify the type and severity of defects on the core sample end faces.

5. The method according to claim 1, characterized in that, Based on the joint feature matrix of the cylindrical surface and the drilling and cutting process parameter records, combined with the defect type annotation information of the core sample cylindrical surface, a defect identification model for the core sample cylindrical surface is constructed to identify the defect types of the concrete core sample cylindrical surface, including: By using a core sample inspection database, the joint feature matrix of the cylindrical surface of historical concrete core samples and the records of drilling and cutting process parameters are obtained. The types of defects on the cylindrical surface of the core samples are labeled. A decision tree algorithm is used to train the model and construct a core sample cylindrical surface defect recognition model. The drilling and cutting process parameter records include drilling process parameters and cutting and grinding parameters. The drilling process parameters include drill bit diameter, rotation speed, feed rate, cooling method, equipment vibration spectrum, and drilling inclination angle. The cutting and grinding parameters include cutting blade wear, cutting speed, end face grinding pressure and time. The cylindrical surface defect types include man-made defects and native defects. Based on the joint feature matrix of the core sample cylindrical surface acquired in real time, the core sample cylindrical surface defect type recognition model is used to identify the defect types of the concrete core sample cylindrical surface.

6. The method according to claim 1, characterized in that, The process of disposing of, repairing, and sending for testing of concrete core samples is allocated based on the core sample diameter, end face deformation, cylindrical surface deformation, and the severity of end face defects and information on human-caused defects. This includes: Using a positioning and clamping device, unqualified concrete core samples with diameters that do not meet the preset diameter requirements, abnormal end-face deformation or abnormal cylindrical surface deformation, severe end-face defects, or human-caused defects are conveyed to the waste core sample conveyor belt for waste disposal. Using the positioning and clamping device, concrete core samples that do not meet the preset height requirements, exhibit tilting, or have end-face defects that are not severe are conveyed to the core sample trimming station, while the remaining qualified core samples are conveyed to the compressive strength testing unit. Based on the measured concrete core sample height and the perpendicularity deviation between the end face and the axis, the required grinding amount is calculated, and the end face of the core sample is trimmed. Automatic spraying or heat pressing methods are used to trim the end-face defects.

7. The method according to claim 1, characterized in that, The determination of the compressive strength of the concrete core sample is based on the re-inspection results of the end face and cylindrical surface of the repaired concrete core sample, combined with the loading information of the automatic centering and servo loading device, including: Using a positioning and clamping device, the modified concrete core sample is transferred to an integrated dimensional and quality inspection station for re-inspection of the end face and cylindrical surface. If the inspection is qualified, it is transferred to the compressive strength testing unit. Using an automatic centering and servo loading device, the loading axis is aligned with the core sample axis to conduct a loading test, and load and displacement data are collected in real time. By calculating the core sample failure load and end face area, the compressive strength data of the concrete core sample is obtained, and a concrete core sample test report and core sample strength index are generated.

8. A fully automated concrete core sample measurement and strength testing system, which is based on the fully automated concrete core sample measurement and strength testing method as described in any one of claims 1-7, characterized in that, The system includes the following modules: The core sample image acquisition module is used to acquire multi-view end face images and cylindrical surface images of concrete core samples through a camera installed at the inspection station, extract the boundary contours of the end face images and cylindrical surface images, and determine the diameter and height of the core sample. The core sample deformation anomaly identification module is used to identify core sample end face deformation anomalies and cylindrical surface deformation anomalies based on the boundary point set of the end face image and the cylindrical surface image; The end-face defect recognition module is used to construct a core sample end-face defect recognition model based on historical images of the concrete core sample end face, and to identify the type and severity of defects in the concrete core sample end face acquired in real time. The cylindrical surface feature analysis module is used to extract the texture features, color features, and geometric features of concrete core samples based on multi-view cylindrical surface images and time-coded registered 3D point cloud data, and to construct a joint feature matrix. The cylindrical surface defect recognition module is used to construct a core sample cylindrical surface defect recognition model based on the joint feature matrix of the cylindrical surface and the drilling and cutting process parameter records, combined with the core sample cylindrical surface defect type annotation information, to identify the defect type of the concrete core sample cylindrical surface; The core sample screening and process allocation module is used to allocate the core sample rejection, repair and testing process based on the concrete core sample diameter, end face deformation, cylindrical surface deformation, combined with the severity of end face defects and human defect judgment information. The core sample compressive strength testing module is used to determine the compressive strength of concrete core samples based on the re-inspection results of the end face and cylindrical surface of the repaired concrete core sample, combined with the loading information of the automatic centering and servo loading device.

Images