Concrete appearance quality comprehensive evaluation method and system
Through computer vision and numerical analysis technology, a concrete appearance quality evaluation system was established, which solved the problems of incomplete information and strong subjectivity in the existing technology, achieved rapid and accurate detection of concrete appearance defects, improved detection efficiency and accuracy, and supported Quick decisions by engineers.
Patent Information
- Application Number
- CN202411050889.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-01
AI Technical Summary
The prior art information is not comprehensive enough in the comprehensive evaluation of concrete appearance quality, lacks mature systems to support rapid inspection and adjustment by engineers, and there are subjectivity and complexity problems in the inspection process.
Using a method based on computer vision and numerical analysis, the network and machine learning models are segmented through deep learning examples, concrete appearance image data is collected, defect area annotation and feature extraction is carried out, and concrete appearance quality evaluation system is established, including data acquisition, model training and inference. module, output comprehensive evaluation results.
It realizes fast and accurate concrete appearance defect detection, reduces the experience requirements and equipment needs of inspectors, improves inspection efficiency and accuracy, reduces subjectivity, provides a data management and technical application platform, and supports the rapid decision-making of engineers.
Smart Images

Figure CN119006392A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of evaluation of a comprehensive evaluation method for concrete appearance quality, and more specifically, to a comprehensive evaluation method and system for concrete appearance quality based on computer vision and numerical analysis. Background Art
[0002] With the development of society, the existence of concrete structures is no longer just to meet the requirements of space use. In terms of design, more attention is paid to the pursuit of landscape effects. It not only pursues practicality, but also requires it to meet the requirements of ornamental value. From ordinary buildings to bridges across mountains and seas, no matter what form of structure, the requirements for the appearance of concrete are relatively high. The beauty of concrete structures requires not only the wonderful creativity of designers, but also the appearance quality of concrete to reflect the quality of the entity. The appearance evaluation of concrete structures is a key project of quality inspection, but due to the influence of different regions and construction levels, the appearance quality of concrete is still a difficult point to control during the construction process of structures. Relevant departments have also made certain requirements for this. For example, the Chinese standard JGJ 169-2009 "Technical Specifications for the Application of Plain Concrete" and the American standard ACI 347.3R-13 Guide to Formed Concrete Surfaces both propose that the concrete surface should be flat and smooth, uniform in color, without exposed reinforcement, honeycombs, holes, slag inclusions and other defects, and stipulate parameters such as the pore area of some specific types of concrete.
[0003] Although relevant technical personnel have been continuously researching and optimizing, due to the many factors affecting the appearance quality of concrete and the complex and changeable control methods, some problems have not been effectively solved, causing many troubles to the relevant units. In some current studies, researchers have tried to identify the appearance quality problems of concrete through computer vision and numerical analysis algorithms. It can perform high-dimensional quantitative analysis of the appearance quality of concrete based on the digital features of the collected images, with high accuracy and reliability, so it has been applied in some engineering cases. However, the information obtained by this type of technology in the comprehensive evaluation of concrete appearance quality is often not comprehensive enough, and secondary integration is required, and the corresponding evaluation information should be given to engineering personnel in a timely manner. In addition, there is a lack of a mature system for engineering personnel to use and make subsequent adjustments. Summary of the invention
[0004] In order to achieve these purposes and other advantages according to the present invention, on the one hand, a preferred embodiment of the present invention provides a method for comprehensive evaluation of concrete appearance quality, comprising the following steps:
[0005] S1. Collect concrete appearance image data and mark the data of quality defect areas;
[0006] S2. Based on the concrete appearance image data obtained in S1, a deep learning instance segmentation network is constructed and trained to obtain a concrete appearance comprehensive defect segmentation model; subsequently, the concrete appearance image is input into the concrete appearance comprehensive defect segmentation model, and defect area parameters are obtained through image segmentation model reasoning and image analysis, and a data table is prepared;
[0007] S3. Score the appearance quality of each concrete appearance photo as the target value, and use the defect area parameters obtained in S2 as feature values to establish a machine learning model for concrete appearance quality evaluation and output the evaluation results.
[0008] Preferably, in S1, the image data is collected by using a camera device, the distance between the camera device and the concrete appearance is determined during the collection process, and the original image captured by the camera device is pre-adjusted according to the calibration situation;
[0009] The marking types of the quality defect areas include: color difference areas, crack areas, honeycomb areas, and pockmarked areas, and all of them are marked using a polygonal marking method.
[0010] Preferably, in S2, after the concrete appearance picture is input into the concrete appearance comprehensive defect segmentation model, a target mask is obtained, and image processing is performed on it to extract quantitative numerical information; wherein,
[0011] For the crack region, the length and maximum width of the crack are obtained using a skeleton-based slender numerical extraction method;
[0012] For the color difference area, a conventional numerical extraction method based on the circumscribed polygon is used to extract quantitative numerical information.
[0013] Preferably, for the crack region, a skeleton-based elongated numerical extraction method is used to obtain the length and maximum width of the crack, comprising the following steps:
[0014] S21, using an image skeleton extraction method, thinning the connected area of the crack mask into a pixel width, and thereby establishing an undirected cyclic skeleton line graph;
[0015] S22. The length information of the crack is directly obtained according to the undirected cyclic skeleton line graph, and each point of the crack can be traversed with a circle along the undirected cyclic skeleton line graph to obtain the maximum inscribed circle at the point, which is the maximum width at the point.
[0016] Preferably, for the color difference area, honeycomb area and pockmarked area, a conventional numerical extraction method based on a circumscribed polygon is used to extract quantitative numerical information. The length, width and area of the area are obtained by finding the minimum circumscribed rectangle or ellipse of the irregular area. For the color difference area, the CIE average difference between the color of the concrete and the normal color is additionally calculated.
[0017] Preferably, for the crack region, a skeleton-based elongated numerical extraction method is used to obtain the length and maximum width of the crack; and,
[0018] For the color difference area, a conventional numerical extraction method based on a circumscribed polygon is used to extract quantitative numerical information. The distance between the camera and the concrete appearance recorded by the S1 acquisition data is converted into the corresponding real value according to the ratio, and the value is calculated by the following formula:
[0019]
[0020] Where: For the length of any line segment in the image,
[0021] P real , represents the distance in pixels between the two endpoints in the image in real situations;
[0022] D real , represents the distance between the camera and the concrete surface in real conditions;
[0023] L base , represents the distance in pixels between the two endpoints in reality under calibration conditions;
[0024] P base , represents the distance in pixels between the two endpoints in the image under calibration;
[0025] D base , which represents the distance between the camera and the concrete surface under calibration conditions.
[0026] Preferably, in S3, the appearance quality score is divided into five levels from good to bad, namely, A, B, C, D, E, and the converted value is 12345.
[0027] The characteristic values obtained by S2 include: the number, average length, and average maximum width of crack areas; the number, average area, and CIE average difference of the color difference area with the concrete appearance color; the number and average area of honeycomb areas and pitted areas, wherein the calculation formula of CIE color difference is as follows:
[0028]
[0029] Where:
[0030] L a ,L b -Brightness values of the normal color of the concrete surface and the color of the concrete color difference area in the CIE color space;
[0031] A a,A b - Red and green axis values of the normal color of the concrete surface and the color of the concrete color difference area in the CIE color space;
[0032] B a ,B b -The blue-yellow axis values of the normal color of the concrete surface and the color of the concrete color difference area in the CIE color space.
[0033] Preferably, it is characterized in that, in step S3, establishing a machine learning model for evaluating the appearance quality of concrete comprises the following steps:
[0034] S31, dividing the concrete appearance feature value data obtained in S2 into a training set, a validation set and a test set;
[0035] S32. Select six machine learning algorithms including CatBoost, XgbBoost, LightGBM, ExtraTrees, RandomTree, and DecisionTree. Train the above six algorithms on the training set using default hyperparameters, and select the three algorithms with the highest scores on the validation set to enter the next stage.
[0036] S33. Set the optimized hyperparameter range, use the 5-fold cross-validation method and the Bayesian hyperparameter optimization method to make adjustments on the sub-datasets consisting of the training set and the validation set, and obtain the best hyperparameter combination for each of the three algorithms. Finally, introduce the optimized hyperparameter combination settings into the three algorithms obtained in S32, and select the model with the highest prediction score on the test set as the machine learning model for concrete appearance quality evaluation.
[0037] On the other hand, another technical solution of the present application provides an evaluation system for the comprehensive evaluation method for concrete appearance quality according to any one of claims 1 to 8, comprising:
[0038] Data collection, addition and preprocessing module: used to collect concrete appearance image data and annotate the data of quality defect areas;
[0039] Model training and reasoning module: used to construct a deep learning instance segmentation network based on concrete appearance image data and perform training to obtain a concrete appearance comprehensive defect segmentation model; then, the concrete appearance picture is input into the concrete appearance comprehensive defect segmentation model, and the defect area parameters are obtained through image segmentation model reasoning and image analysis, and a data table is prepared: the appearance quality score of each concrete appearance photo is used as the target value, and based on the defect area parameters as the feature value, a machine learning model for concrete appearance quality evaluation is established, and the evaluation results are output;
[0040] Adjustment and management module: used to add, delete and view the images and quantitative data required for the model, and adjust the output type of expert basic score and concrete appearance quality quantitative numerical results.
[0041] The present invention includes at least the following beneficial effects: the present invention can quickly extract and integrate various types of concrete appearance defect areas for various types of concrete appearance pictures, reduce the experience requirements, equipment requirements and repeated inspections for inspection personnel, improve efficiency, and effectively avoid subjective problems in the inspection process. In addition, it has the advantages of short inspection time, high inspection accuracy, and effectively improves the ability to detect concrete appearance defects. At the same time, the comprehensive quality level of concrete appearance is quickly evaluated according to the expert scoring database, and the mask results obtained through image analysis provide accurate appearance defect quantification and evaluation information. In addition, according to the above technology, a data management and technology use platform is provided to promote engineering personnel to quickly apply corresponding technologies and make decisions in a timely manner.
[0042] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] Figure 1 This is a flow chart of a method for comprehensive evaluation of concrete appearance quality based on computer vision and numerical analysis according to an embodiment of the present invention;
[0044] Figure 2 Schematic diagram of a common target analysis method based on a circumscribed polygon and a slender target analysis method based on a skeleton model according to an embodiment of the present invention;
[0045] Figure 3 Schematic diagram of loss curve of the training process of the comprehensive defect segmentation model for concrete appearance according to an embodiment of the present invention;
[0046] Figure 4 A diagram showing the prediction results obtained by the concrete appearance comprehensive defect segmentation model according to an embodiment of the present invention;
[0047] Figure 5 Schematic diagram of comparison of prediction results obtained by the machine learning model for evaluating concrete appearance quality according to an embodiment of the present invention. DETAILED DESCRIPTION
[0048] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.
[0049] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations. The basic principles of the present invention defined in the following description can be applied to other embodiments, variations, improvements, equivalents, and other technical solutions that do not deviate from the spirit and scope of the present invention.
[0050] Those skilled in the art should understand that, in the disclosure of the present invention, the orientation or position relationship indicated by the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc. are based on the orientation or position relationship shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the above terms should not be understood as limiting the present invention.
[0051] It is to be understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element may be one, while in another embodiment, the number of the element may be multiple, and the term "one" should not be understood as a limitation on the quantity.
[0052] like Figure 1-5 As shown, on the one hand, a preferred embodiment of the present invention provides a method for comprehensive evaluation of concrete appearance quality, comprising the following steps:
[0053] S1. Collecting concrete appearance image data and annotating data of quality defect areas; the annotated types of quality defect areas include: color difference areas, crack areas, honeycomb areas, and pitted areas, and all of them are annotated using a polygonal annotation method.
[0054] In the S1, the acquisition process includes manual shooting and drone shooting. During the shooting process, the distance between the shooting equipment and the concrete appearance is recorded by manual measurement or infrared light. For the same concrete surface area, two pictures are taken. The first one is the original surface area, called K1; the second one uses a calibration plate to stick to the concrete surface and shoot to measure the shooting angle, called K2; that is, the original image is pre-adjusted to a standard parallel positive image based on this information, called K3. In this embodiment, if a depth camera is used to shoot directly, there is no need to record the distance and calibration information, and the image can be directly adjusted through the depth matrix obtained by the depth camera, which is equivalent to K3. In this embodiment, the monocular camera is tested and the HikVision series camera is selected, and the binocular depth camera is tested and the Intel RealSense series camera is selected. The shooting working distance is recommended to be between 50-350cm.
[0055] After shooting, all K3 images and distance information are collected to build a concrete appearance dataset. The annotation types of image quality defect areas in the dataset include the following four types: color difference area, crack area and other defect areas, all of which are polygonal annotation methods.
[0056] At the same time, due to the small number of concrete appearance quality images, the quality is greatly affected by environmental conditions and shooting level, and the difference from natural images is large. Therefore, before officially starting the model training, additional high-quality concrete appearance images are taken under laboratory conditions and a pre-trained model is obtained. Then the above concrete appearance dataset is introduced for model fine-tuning. Finally, all processed image experimental data are divided into three datasets, namely training set, validation set and prediction set. All data in the prediction set are from the concrete appearance dataset.
[0057] S2. Based on the concrete appearance image data obtained in S1, a deep learning instance segmentation network is constructed and trained to obtain a concrete appearance comprehensive defect segmentation model; subsequently, the concrete appearance image is input into the concrete appearance comprehensive defect segmentation model, and defect area parameters, such as defect area area, length and other quantitative results, are obtained through image segmentation model reasoning and image analysis, and a data table is prepared;
[0058] In this embodiment, the network used in the concrete appearance comprehensive defect segmentation model uses an instance segmentation network based on the improved ConvNeXt-V2 as the backbone network. The backbone network uses the ConvNeXt-V2-Tiny version, the number of FPNs in the Neck part is set to 3, and the Head part uses a conventional one-stage segmentation head.
[0059] Specifically, in this embodiment, the optimizer selects SGD, the number of training times is 200, and the initial learning rate is set to 10 -3 , which decreases linearly with the increase of training times, and finally decreases to 10 -6 ; Focal Loss is selected as the loss function to solve the long-tail imbalance problem. The loss curve during training is shown below Figure 3 .
[0060] After obtaining the comprehensive defect segmentation model for concrete appearance, input the concrete appearance picture, and the model will output the defect area mask and category judged by the picture. The specific result diagram is shown in Figure 4 After obtaining the output mask, it is necessary to extract quantitative numerical information such as area, length, width, etc. through image processing. For crack targets, a skeleton-based slender numerical extraction method is used to obtain the length and maximum width of the crack. For other defect areas including chromatic aberration areas, a conventional numerical extraction method based on circumscribed polygons is used. See the schematic diagram. Figure 2 .
[0061] S3. Score the appearance quality of each concrete appearance photo as the target value, and use the defect area parameters obtained in S2 as feature values to establish a machine learning model for concrete appearance quality evaluation and output the evaluation results.
[0062] In this embodiment, in order to establish a machine learning model for concrete appearance quality evaluation, it is necessary to obtain a feature set and a target value. Based on the appearance quality of concrete, concrete designers and project site engineers are invited in advance to score the pictures in the concrete appearance image data set for the appearance quality of concrete. The scores are divided into five levels from good to bad, ABCDE, which are converted into numerical values of 12345. The maximum and minimum values of these values are discarded, and the appearance quality score of each picture is obtained as the target value by averaging. The feature set is obtained from the quantitative values obtained from S2, including: the number, average length, and average maximum width of cracks; the number, average area, and CIE average difference of the color difference area with the appearance color of the concrete itself; the number and average area of other defective areas.
[0063] In another technical solution, in S2, after the concrete appearance picture is input into the concrete appearance comprehensive defect segmentation model, a target mask is obtained, and image processing is performed on it to extract quantitative numerical information; wherein,
[0064] For the crack region, the length and maximum width of the crack are obtained using a skeleton-based slender numerical extraction method;
[0065] For the color difference area, a conventional numerical extraction method based on the circumscribed polygon is used to extract quantitative numerical information.
[0066] in,
[0067] For the crack area, the length and maximum width of the crack are obtained using a skeleton-based slender numerical extraction method, which includes the following steps:
[0068] S21, using an image skeleton extraction method, thinning the connected area of the crack mask into a pixel width, and thereby establishing an undirected cyclic skeleton line graph;
[0069] S22. The length information of the crack is directly obtained according to the undirected cyclic skeleton line graph, and each point of the crack can be traversed with a circle along the undirected cyclic skeleton line graph to obtain the maximum inscribed circle at the point, which is the maximum width at the point.
[0070] In another technical solution, for color difference areas, honeycomb areas and pockmarked areas, a conventional numerical extraction method based on circumscribed polygons is used to extract quantitative numerical information. The length, width and area of the area are obtained by finding the minimum circumscribed rectangle or ellipse of the irregular area. For the color difference area, the CIE average difference between it and the normal color of the concrete appearance is additionally calculated.
[0071] In another technical solution, for the crack region, a skeleton-based slender numerical extraction method is used to obtain the length and maximum width of the crack; and,
[0072] For the color difference area, a conventional numerical extraction method based on a circumscribed polygon is used to extract quantitative numerical information. The distance between the camera and the concrete appearance recorded by the S1 acquisition data is converted into the corresponding real value according to the ratio, and the value is calculated by the following formula:
[0073]
[0074] Where: For the length of any line segment in the image,
[0075] P real , represents the distance in pixels between the two endpoints in the image in real situations;
[0076] D real , represents the distance between the camera and the concrete surface in real conditions;
[0077] L base , represents the distance in pixels between the two endpoints in reality under calibration conditions;
[0078] P base , represents the distance in pixels between the two endpoints in the image under calibration;
[0079] D base , which represents the distance between the camera and the concrete surface under calibration conditions.
[0080] The results of comparing the numerical information of concrete appearance quality defects obtained by image analysis with the corresponding manual measurement information are shown in Table 1, which verifies the accuracy of the image analysis method.
[0081] Table 1 Comparison between numerical information analysis values and measured values of concrete appearance quality defects
[0082]
[0083] In another technical solution, in S3, the appearance quality score is divided into five levels from good to bad, A, B, C, D, E, and is converted into a numerical value of 12345.
[0084] The characteristic values obtained by S2 include: the number, average length, and average maximum width of crack areas; the number, average area, and CIE average difference of the color difference area with the concrete appearance color; the number and average area of honeycomb areas and pitted areas, wherein the calculation formula of CIE color difference is as follows:
[0085]
[0086] Where:
[0087] L a ,L b -Brightness values of the normal color of the concrete surface and the color of the concrete color difference area in the CIE color space;
[0088] A a ,A b -Red-green axis values of the normal color of the concrete surface and the color of the concrete color difference area in the CIE color space;
[0089] B a ,B b -The blue-yellow axis values of the normal color of the concrete surface and the color of the concrete color difference area in the CIE color space.
[0090] In another technical solution, S3 specifically includes the following steps:
[0091] S31. Divide the concrete appearance feature value data obtained from S2 into a training set, a validation set, and a test set; the ratio is approximately 7:2:1.
[0092] S32. Since it is not known in advance which machine learning algorithm is suitable for establishing a machine learning model for selecting concrete appearance quality evaluation, six machine learning algorithms including CatBoost, XgbBoost, LightGBM, ExtraTrees, RandomTree, and DecisionTree are introduced at the same time. The above six algorithms are trained on the training set using default hyperparameters, and the three algorithms with the highest scores on the validation set are selected to enter the next stage;
[0093] S33. Set the optimized hyperparameter range, use the 5-fold cross-validation method and the Bayesian hyperparameter optimization method to make adjustments on the sub-datasets consisting of the training set and the validation set, and obtain the best hyperparameter combination for each of the three algorithms. Finally, introduce the optimized hyperparameter combination settings into the three algorithms obtained in S32, and select the model with the highest prediction score on the test set as the machine learning model for concrete appearance quality evaluation.
[0094] In this embodiment, a total of 310 sets of data were used to construct the concrete appearance quality evaluation machine learning model. Finally, the CatBoost algorithm was selected as the best algorithm with an accuracy of 0.86. Figure 5 The test results of 31 sets of test data are shown.
[0095] On the other hand, another technical solution of the present application provides an evaluation system for a comprehensive evaluation method of concrete appearance quality, comprising:
[0096] Data collection, addition and preprocessing module: used to collect concrete appearance image data and annotate the data of quality defect areas;
[0097] Model training and reasoning module: used to construct a deep learning instance segmentation network based on concrete appearance image data and perform training to obtain a concrete appearance comprehensive defect segmentation model; then, the concrete appearance picture is input into the concrete appearance comprehensive defect segmentation model, and the defect area parameters are obtained through image segmentation model reasoning and image analysis, and a data table is prepared: the appearance quality score of each concrete appearance photo is used as the target value, and based on the defect area parameters as the feature value, a machine learning model for concrete appearance quality evaluation is established, and the evaluation results are output;
[0098] Adjustment and management module: used to add, delete and view the images and quantitative data required for the model, and adjust the output type of expert basic score and concrete appearance quality quantitative numerical results.
[0099] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.
Claims
1. A method for comprehensive evaluation of concrete appearance quality, characterized in that: The following steps are involved: S1. Collect concrete appearance image data and mark the data of quality defect areas; S2. Based on the concrete appearance image data obtained in S1, a deep learning instance segmentation network is constructed and trained to obtain a concrete appearance comprehensive defect segmentation model; subsequently, the concrete appearance image is input into the concrete appearance comprehensive defect segmentation model, and defect area parameters are obtained through image segmentation model reasoning and image analysis, and a data table is prepared; S3. Score the appearance quality of each concrete appearance photo as the target value, and use the defect area parameters obtained in S2 as feature values to establish a machine learning model for concrete appearance quality evaluation and output the evaluation results.
2. The method for comprehensive evaluation of concrete appearance quality according to claim 1, characterized in that: In S1, the image data is collected by using a camera device, the distance between the camera device and the concrete appearance is determined during the collection process, and the original image captured by the camera device is pre-adjusted according to the calibration situation; The marking types of the quality defect areas include: color difference areas, crack areas, honeycomb areas, and pockmarked areas, and all of them are marked using a polygonal marking method.
3. The method for comprehensive evaluation of concrete appearance quality according to claim 2, characterized in that: In S2, after the concrete appearance picture is input into the concrete appearance comprehensive defect segmentation model, a target mask is obtained, and image processing is performed on it to extract quantitative numerical information; wherein, For the crack region, the length and maximum width of the crack are obtained using a skeleton-based slender numerical extraction method; For the color difference area, a conventional numerical extraction method based on the circumscribed polygon is used to extract quantitative numerical information.
4. The method for comprehensive evaluation of concrete appearance quality according to claim 3, characterized in that: For the crack area, the length and maximum width of the crack are obtained using a skeleton-based slender numerical extraction method, which includes the following steps: S21, using an image skeleton extraction method, thinning the connected area of the crack mask into a pixel width, and thereby establishing an undirected cyclic skeleton line graph; S22. The length information of the crack is directly obtained according to the undirected cyclic skeleton line graph, and each point of the crack can be traversed with a circle along the undirected cyclic skeleton line graph to obtain the maximum inscribed circle at the point, which is the maximum width at the point.
5. The method for comprehensive evaluation of concrete appearance quality according to claim 3, characterized in that: For the color difference area, honeycomb area and pockmarked area, a conventional numerical extraction method based on the circumscribed polygon is used to extract quantitative numerical information. The length, width and area of the area are obtained by finding the minimum circumscribed rectangle or ellipse of the irregular area. For the color difference area, the CIE average difference between it and the normal color of the concrete appearance is additionally calculated.
6. The method for comprehensive evaluation of concrete appearance quality according to claim 1, characterized in that: For the crack region, the length and maximum width of the crack are obtained using a skeleton-based slender numerical extraction method; and, For the color difference area, a conventional numerical extraction method based on a circumscribed polygon is used to extract quantitative numerical information. The distance between the camera and the concrete appearance recorded by the S1 acquisition data is converted into the corresponding real value according to the ratio, and the value is calculated by the following formula: Where: For the length of any line segment in the image, P real , represents the distance in pixels between the two endpoints in the image in real situations; D real , represents the distance between the camera and the concrete surface in real conditions; L base , represents the distance in pixels between the two endpoints in reality under calibration conditions; P base , represents the distance in pixels between the two endpoints in the image under calibration; D base , which represents the distance between the camera and the concrete surface under calibration conditions.
7. The method for comprehensive evaluation of concrete appearance quality according to claim 1, characterized in that: In S3, the appearance quality score is divided into five levels from good to bad, A, B, C, D, E, and the converted value is 12345. The characteristic values obtained by S2 include: the number, average length, and average maximum width of crack areas; the number, average area, and CIE average difference of color difference areas with the appearance color of the concrete itself; the number and average area of honeycomb areas and pitted areas, wherein the calculation formula of CIE color difference is as follows: Where: L a ,L b -Brightness values of the normal color of the concrete surface and the color of the concrete color difference area in the CIE color space; A a ,A b - Red and green axis values of the normal color of the concrete surface and the color of the concrete color difference area in the CIE color space; B a ,B b -The blue-yellow axis values of the normal color of the concrete surface and the color of the concrete color difference area in the CIE color space.
8. The method for comprehensive evaluation of concrete appearance quality based on computer vision and numerical analysis according to claim 1, characterized in that: In step S3, a machine learning model for evaluating the appearance quality of concrete is established, including the following steps: S31, dividing the concrete appearance feature value data obtained in S2 into a training set, a validation set and a test set; S32. Select six machine learning algorithms including CatBoost, XgbBoost, LightGBM, ExtraTrees, RandomTree, and DecisionTree. Train the above six algorithms on the training set using default hyperparameters, and select the three algorithms with the highest scores on the validation set to enter the next stage. S33. Set the optimized hyperparameter range, use the 5-fold cross-validation method and the Bayesian hyperparameter optimization method to make adjustments on the sub-datasets consisting of the training set and the validation set, and obtain the best hyperparameter combination for each of the three algorithms. Finally, introduce the optimized hyperparameter combination settings into the three algorithms obtained in S32, and select the model with the highest prediction score on the test set as the machine learning model for concrete appearance quality evaluation.
9. An evaluation system for the comprehensive evaluation method for concrete appearance quality according to any one of claims 1 to 8, characterized in that: include: Data collection, addition and preprocessing module: used to collect concrete appearance image data and annotate the data of quality defect areas; Model training and reasoning module: used to construct a deep learning instance segmentation network based on concrete appearance image data and perform training to obtain a concrete appearance comprehensive defect segmentation model; then, the concrete appearance picture is input into the concrete appearance comprehensive defect segmentation model, and the defect area parameters are obtained through image segmentation model reasoning and image analysis, and a data table is prepared: the appearance quality score of each concrete appearance photo is used as the target value, and based on the defect area parameters as the feature value, a machine learning model for concrete appearance quality evaluation is established, and the evaluation results are output; Adjustment and management module: used to add, delete and view the images and quantitative data required for the model, and adjust the output type of expert basic score and concrete appearance quality quantitative numerical results.
Citation Information
Patent Citations
Fair-faced concrete surface color difference analysis method based on convolutional neural network
CN111311695A
Vehicle-mounted front-view image crack detection method
CN114596551A
Nuclear power plant concrete defect identification method based on YOLO algorithm
CN117152073A
Fair-faced concrete color difference detection and identification method, system and equipment and storage medium
CN117853475A
Cited By
Engineering project intelligent construction data analysis system and method based on artificial intelligence
CN120782596A
Concrete apparent quality intelligent detection and evaluation method and system based on unmanned aerial vehicle and deep learning
CN120894295A
