A method and system for evaluating the quality of concrete vibration based on image analysis

Abstract

The present application relates to the technical field of image data processing, and more particularly to a concrete vibration quality evaluation method and system based on image analysis. The method comprises: acquiring a gray image of the surface of the newly vibrated concrete; constructing a region adjacency graph and extracting multi-dimensional features; calculating physical entropy weight based on the multi-dimensional features; performing region merging operation; identifying bubble clusters and calculating the quality evaluation index of the concrete surface according to the statistical characteristics of the bubble clusters; and outputting the concrete vibration quality evaluation result according to the quality evaluation index. The present application divides the concrete surface by the watershed algorithm and region adjacency graph, extracts multi-dimensional features, constructs a physical entropy weight model to evaluate the saliency, uses a boundary gray level priority queue and an adaptive multi-threshold intelligent region merging method to solve the problem of over-segmentation, and combines the Gaussian mixture model and the Bayesian information criterion to identify bubble clusters, so as to comprehensively evaluate the concrete vibration quality and improve the objectivity and accuracy.

Description

A method and system for evaluating the quality of concrete vibration based on image analysis Technical Field

[0001] This invention relates to the field of image data processing technology, and in particular to a method and system for evaluating the quality of concrete vibration based on image analysis. Background Technology

[0002] In the fields of building construction and civil engineering, the quality of concrete vibration is a key factor determining the strength and safety of a structure. Insufficient vibration leads to a large number of air bubbles remaining inside the concrete, reducing the structure's load-bearing capacity, while excessive vibration can cause aggregate settling and slurry rising, resulting in segregation and also impairing the mechanical properties of the concrete. Therefore, accurately assessing the surface quality of the concrete after vibration is a necessary step in ensuring project quality.

[0003] Currently, construction site assessments are typically conducted manually. However, this method is highly subjective and difficult to implement effectively. With the development of computer vision technology, automated assessment based on image analysis is becoming a trend. However, existing image analysis techniques struggle with complex surfaces like concrete. The concrete surface has a complex microstructure, including cement paste matrix, aggregates of varying sizes, and air bubbles generated during vibration. The grayscale distribution of these components is extremely uneven and the boundaries are blurred, making effective analysis impossible.

[0004] In traditional methods, while the standard watershed algorithm can capture topological structures, it is extremely sensitive to minute grayscale changes on complex surfaces. It often misjudges texture fluctuations on the concrete slurry surface as region boundaries, resulting in numerous meaningless oversegmented regions in a single image. In such severe oversegmentation, the true bubble features are overwhelmed by a large number of randomly distributed noise regions. Due to the lack of a model that can effectively measure the essential physical differences between bubbles, aggregates, and noise, existing technologies struggle to establish a reliable mapping relationship between surface image features and internal vibration quality. This makes it impossible to accurately identify and evaluate bubble parameters, thus reducing the reliability of subsequent assessments of concrete vibration quality. Summary of the Invention

[0005] To address the problems of severe oversegmentation, difficulty in distinguishing between air bubbles and aggregate noise, and inability to accurately assess concrete vibration quality when performing image analysis using the standard watershed algorithm, this invention provides a concrete vibration quality assessment method and system based on image analysis.

[0006] In a first aspect, the present invention provides a method for evaluating the quality of concrete vibration based on image analysis, employing the following technical solution:

[0007] A method for evaluating the quality of concrete vibration based on image analysis includes: acquiring a grayscale image of the concrete surface after vibration; performing a watershed algorithm on the grayscale image to obtain a labeled image containing multiple connected regions; constructing a region adjacency graph based on the labeled image and extracting the region area, region texture entropy, and region grayscale difference of each connected region; normalizing the region area, region texture entropy, and region grayscale difference of each connected region; calculating a physical entropy weight that measures the structural saliency of each connected region based on the normalized region area, region texture entropy, and region grayscale difference; and then, based on the region adjacency graph... A priority queue is constructed using the minimum grayscale value of the boundary of adjacent connected regions. An adaptive multi-threshold algorithm is used to determine the merging termination threshold. Adjacent connected regions are merged according to the order of the priority queue until the minimum grayscale value of the boundary exceeds the merging termination threshold, resulting in merged regions. Feature vectors of each merged region are extracted, and bubble clusters are identified using a Gaussian mixture model and Bayesian information criterion. Based on the statistical characteristics of the bubble clusters, quality assessment indicators for the concrete surface are calculated. These quality assessment indicators include at least bubble density, bubble size uniformity, and surface compaction. The concrete vibration quality assessment results are output based on these quality assessment indicators.

[0008] This invention obtains initial connected regions using a watershed algorithm and constructs them using a region adjacency graph, achieving preliminary segmentation of complex concrete surfaces. By extracting multi-dimensional features such as region area, texture entropy, and grayscale difference, a physical entropy weight evaluation model is constructed, effectively measuring the structural saliency of each region. Intelligent merging of adjacent regions is achieved by constructing a priority queue based on the minimum grayscale value at the boundary. An adaptive multi-threshold algorithm ensures precise control of the merging process, effectively solving the problem of over-segmentation in traditional watershed algorithms. Accurate identification of bubble clusters is achieved through a combination of Gaussian mixture model and Bayesian information criterion, fully utilizing the essential differences in physical characteristics between bubbles, aggregates, and noise, establishing a reliable mapping relationship between surface image features and internal vibration quality. A comprehensive evaluation of concrete vibration quality is achieved through multi-dimensional quality assessment indicators such as bubble density, bubble size uniformity, and surface compaction, improving the objectivity and accuracy of concrete surface quality assessment and providing reliable technical support for engineering quality control.

[0009] Furthermore, the grayscale image is a grayscale image of the concrete surface after being smoothed by Gaussian filtering.

[0010] Furthermore, the method for obtaining the region texture entropy is as follows: for each connected region in the label image, calculate the gray-level co-occurrence matrix; compress the gray-level levels of the gray-level co-occurrence matrix and calculate the average value of the gray-level co-occurrence matrix in multiple directions; calculate the region texture entropy of the connected region based on the average gray-level co-occurrence matrix.

[0011] Furthermore, the physical entropy weight satisfies: In the formula, Connected regions in the label image Physical entropy weight, Connected regions in the label image The regional grayscale difference Connected regions in the label image The area of ​​the region, Connected regions in the label image Regional texture entropy, This is a hyperparameter used to prevent the denominator from being zero.

[0012] This invention achieves a scientific evaluation of physical entropy weights by constructing a ratio of the product of the gray-level difference and area of ​​a region to the texture entropy. It comprehensively considers the geometric and textural features of the region. The product of the gray-level difference and area reflects the structural saliency of the region, and the normalization of the texture entropy ensures the comparability of regions with different complexities. This effectively distinguishes between real structural regions and noisy regions, providing a reliable weighting basis for subsequent region merging and bubble identification, and improving the accuracy of concrete surface feature analysis.

[0013] Furthermore, the step of merging adjacent connected regions according to the priority queue order includes: applying the Multi-Otsu algorithm to the grayscale image to divide the grayscale image into dark areas, mid-range areas, and bright areas, and selecting the boundary value between the mid-range and bright areas as the merging termination threshold; when the minimum grayscale value of the boundary of the currently processed adjacent connected regions in the priority queue is not greater than the merging termination threshold, the two adjacent connected regions are merged, the physical entropy weight of the merged region is the average of the physical entropy weights of the two connected regions, and the grayscale difference of the merged region is the maximum value of the grayscale differences of the two connected regions.

[0014] This invention uses the Multi-Otsu algorithm to divide grayscale images into dark, medium, and bright areas, achieving intelligent differentiation of different components on the concrete surface. The boundary between the medium and bright areas is used as the merging termination threshold, ensuring the merging process stops within a reasonable grayscale range. The merging strategy controls the merging process based on the comparison between the minimum boundary grayscale value and the threshold, avoiding over-merging. The calculation method of taking the mean of the physical entropy weight and the maximum grayscale difference after merging maintains the rationality and representativeness of the merged area characteristics, improving the accuracy and reliability of area merging.

[0015] Furthermore, the step of identifying bubble clusters using Gaussian mixture models and Bayesian information criteria includes: constructing a feature vector set containing region area, region texture entropy, and region grayscale difference; clustering the feature vector set using Gaussian mixture models to obtain single-cluster models and dual-cluster models; and calculating the Bayesian information criterion value of the single-cluster model. Bayesian information criterion value of bi-cluster model ; in response to If the sum of grayscale differences and texture entropy of all connected regions in the dual-cluster model is maximized, then the cluster with the largest sum of grayscale differences and the smallest sum of texture entropy is identified as the bubble cluster; otherwise, no bubbles are identified.

[0016] This invention achieves a comprehensive characterization of concrete surface components by constructing a multidimensional feature vector that includes region area, texture entropy, and gray-level difference. It uses a Gaussian mixture model for single-cluster and double-cluster modeling and determines the optimal model by comparing and contrasting the two models using the Bayesian information criterion. By selecting bubble clusters based on the criterion of maximizing the sum of gray-level differences and minimizing the sum of texture entropies, it effectively distinguishes between bubble, aggregate, and noise regions, thereby improving the accuracy and robustness of bubble identification.

[0017] Furthermore, the bubble size uniformity satisfies: In the formula, To ensure uniformity of bubble size on the concrete surface, The standard deviation of the area of ​​all merged regions in the bubble cluster. This is the average area of ​​all merged regions within the bubble cluster.

[0018] Furthermore, the surface density satisfies: In the formula, The surface density of the concrete surface. This is the sum of the areas of all merged regions within the bubble cluster. This represents the total area of ​​the label image.

[0019] Furthermore, the output concrete vibration quality assessment result includes: outputting insufficient concrete vibration in response to the surface compaction being less than a set compaction threshold or the bubble density being greater than a set density threshold; outputting localized insufficient concrete vibration in response to the bubble size uniformity being less than a set uniformity threshold and the bubble density being greater than a set density threshold; and outputting good concrete vibration in response to the surface compaction being not less than a set compaction threshold and the bubble density being not greater than a set density threshold, thereby achieving an image analysis-based assessment of concrete vibration quality.

[0020] This invention achieves precise classification of concrete vibration quality by constructing a multi-index judgment logic based on surface compaction, bubble density, and bubble size uniformity. It can accurately identify different quality defect types such as insufficient vibration and localized insufficient vibration, providing an objective and reliable basis for concrete quality control and improving the accuracy and consistency of quality assessment.

[0021] Secondly, the present invention provides a concrete vibration quality assessment system based on image analysis, which adopts the following technical solution:

[0022] A concrete vibration quality assessment system based on image analysis includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned concrete vibration quality assessment method based on image analysis is implemented.

[0023] By adopting the above technical solution, a computer program for evaluating the quality of concrete vibration based on image analysis is generated and stored in a memory so that it can be loaded and executed by a processor. This allows for the creation of a terminal device based on the memory and processor, making it convenient to use.

[0024] The present invention has the following technical effects:

[0025] (1) To address the oversegmentation problem caused by the sensitivity of traditional watershed algorithms to small gray-scale changes on concrete surfaces, physical entropy weights are calculated by extracting the area of ​​connected regions, the texture entropy of regions, and the gray-scale difference of regions. This allows for a more accurate measurement of the structural saliency of each connected region, such as the physical characteristics of the bubble region, the texture of cement paste, and the noise. A priority queue is then constructed based on the minimum gray-scale value of the boundary of adjacent connected regions. An adaptive multi-threshold algorithm is used to control the merging termination condition, avoiding misjudging the normal texture fluctuations on the surface of concrete paste as the region boundary. This effectively merges meaningless noise-segmented regions while fully preserving the connected region features corresponding to real bubbles, thus solving the problem of real bubble features being submerged by a large number of random noise regions.

[0026] (2) Breaking through the limitations of existing technologies that lack models of the essential physical differences between bubbles, aggregates and noise, based on the feature vectors of the merged region, Gaussian mixture model and Bayesian information criterion are used to achieve accurate classification and identification of bubble clusters. The model evaluates the differences in physical characteristics of different regions, such as the gray distribution pattern of bubbles, the range of texture entropy and the essential differences between aggregates and noise, avoiding misjudging aggregate particles or random noise as bubbles, thereby establishing a reliable mapping relationship between concrete surface image features and internal bubble parameters, providing a more accurate and interference-free data foundation for subsequent vibration quality assessment.

[0027] (3) It replaces the traditional subjective evaluation method of manual observation. By extracting quantitative quality evaluation indicators such as bubble density, bubble size uniformity, and surface compaction, it realizes the automated and standardized evaluation of concrete vibration quality. It does not rely on the experience judgment of operators. The evaluation results have a unified standard, which reduces human subjective error and improves evaluation efficiency. It is suitable for the large-scale and continuous quality control needs of building construction and civil engineering, and avoids the problems of low efficiency and inconsistent results caused by manual evaluation.

[0028] (4) By suppressing over-segmentation and accurately identifying bubble clusters, the quality assessment indicators can be ensured to truly reflect the vibration situation inside the concrete. For example, when vibration is insufficient, the bubble density is high and the size uniformity is poor. When vibration is excessive, the bubble distribution is abnormal and the surface density is not up to standard due to aggregate segregation. Based on the output assessment results, it is possible to accurately determine whether the vibration meets the quality requirements, promptly identify problems of insufficient or excessive vibration, avoid problems such as a large number of residual bubbles and aggregate segregation inside the concrete due to vibration quality defects, and thus ensure the bearing capacity and mechanical properties of the concrete structure, providing key technical support for the strength and safety of the engineering structure. Attached Figure Description

[0029] Figure 1 is a flowchart of a method for evaluating the quality of concrete vibration based on image analysis according to an embodiment of the present invention.

[0030] Figure 2 is an original image of the concrete surface in an image analysis-based concrete vibration quality assessment method according to an embodiment of the present invention.

[0031] Figure 3 is a labeled image of the concrete surface segmented by the watershed algorithm in an image analysis-based concrete vibration quality assessment method according to an embodiment of the present invention.

[0032] Figure 4 is a label image of the concrete surface after performing a region merging operation in a concrete vibration quality assessment method based on image analysis according to an embodiment of the present invention.

[0033] Figure 5 is a binary image of the merged region containing bubble clusters after cluster analysis in a concrete vibration quality assessment method based on image analysis according to an embodiment of the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] This invention discloses a method for evaluating the quality of concrete vibration based on image analysis, referring to Figure 1, including steps S001-S006:

[0036] S001: Obtain a grayscale image of the newly vibrated concrete surface.

[0037] It should be noted that concrete surfaces have complex and multi-scale texture features, including microbubbles, aggregates of different sizes, and cement paste matrix. The initial grayscale image is processed using the standard watershed algorithm to obtain an initial segmentation region containing accurate topological boundaries and local minima, capturing all the microscopic details of the concrete surface and providing complete basic data for subsequent processing.

[0038] Specifically, as shown in Figure 2, under the condition of ensuring uniform illumination during the acquisition, the original image of the newly vibrated concrete surface is acquired by an industrial camera, and the original image is converted into an initial grayscale image.

[0039] Apply a Gaussian filter with a standard deviation of 1, such as a Gaussian distribution, to the initial input grayscale image to remove high-frequency noise and obtain a smoothed grayscale image.

[0040] As shown in Figure 3, the standard watershed algorithm is applied to the initial grayscale image, and the label image is obtained by immersion simulation process.

[0041] Assign a unique numeric identifier to each connected region in the labeled image, such as from to This forms a set of connected regions.

[0042] At this point, the label image is segmented into a large number of small regions, preserving the local minimum structure of the surface.

[0043] S002: Construct a region adjacency graph and extract multidimensional features.

[0044] It is important to note that after obtaining the segmented regions, the core issue becomes how to distinguish the different physical components of the concrete surface, including air bubbles, aggregates, and cement paste. This is because different physical components manifest as complex grayscale and texture patterns in the image, making them impossible to identify using a single feature. This step constructs a region adjacency map and extracts multi-dimensional features to establish a mapping relationship with physical properties. Air bubbles are represented by regions with large grayscale differences and uniform textures, while aggregates are represented by complex textures and gradual grayscale differences. Multi-dimensional feature fusion provides a basis for calculating physical entropy weights, thus solving the problem that a single feature cannot distinguish complex surface components.

[0045] Specifically, create a region adjacency graph. ,in, For a set of nodes, Let it be the set of edges.

[0046] In the label image, create a corresponding node for each connected region. If two connected components are adjacent on the label image, then at the corresponding nodes of the two connected components... and Add edges between .

[0047] Calculate the area of ​​each connected region, which is equal to the total number of pixels within the connected region. Then, find the minimum gray value inside the connected region that is not on the region boundary. .

[0048] Calculate the region texture entropy for each connected region. On the grayscale image, calculate the gray-level co-occurrence matrix (GLCM) for each connected region. The parameters of the GLCM are set as follows: gray levels compressed to 32 levels, spatial distance... The average GLCM in four directions (0°, 45°, 90°, and 135°) is calculated, and then the regional texture entropy is obtained by calculating the Shannon entropy in information theory. The reason for using GLCM to calculate texture entropy is that GLCM can accurately assess the texture complexity of different regions of the concrete surface. For example, low entropy indicates that the inside of the bubble is uniform, while high entropy indicates that the aggregate surface is rough. As a second-order statistical method, GLCM captures the spatial relationship between pixels, so that the subsequent physical entropy weight can effectively distinguish between real features and noise, while also possessing robustness and computational efficiency.

[0049] Calculate the grayscale difference for each connected region. Since air bubbles on the concrete surface typically appear as recessed areas with large grayscale differences, the regional grayscale difference is key to distinguishing air bubbles from ordinary textures; traverse the connected regions. All edges Find connected components Minimum gray value on the shared boundary The minimum gray value at the boundary represents the gray-level barrier height between two connected regions. The gray-level difference between regions is obtained as follows: .

[0050] S003: Calculate physical entropy weights based on multidimensional features.

[0051] It should be noted that the oversegmentation results contain a mixture of real physical features and random noise. The physical entropy weight is designed in this step because it comprehensively considers the gray-level saliency and internal texture order of the connected region, and can measure the structural saliency of the region. Real bubbles have the characteristics of large gray-level difference, large area and low entropy, so they are given high weights, while noise regions have small areas and complex textures, so they are given low weights. This allows the subsequent merging process to retain real physical features and eliminate irrelevant details, thereby solving the oversegmentation problem.

[0052] Specifically, the multidimensional features are normalized by dividing the region gray-level difference, region area, and region texture entropy of each connected region by the maximum value among the corresponding dimensional features of all connected regions, thus obtaining the normalized features. .

[0053] If the area of ​​a connected region is less than 3 pixels, the connected region is considered as minimal noise, and its physical entropy weight is recorded as 0. Otherwise, the physical entropy weight is calculated, and the physical entropy weight satisfies the following:

[0054] ;

[0055] In the formula, Connected regions in the label image Physical entropy weight, Connected regions in the label image The regional grayscale difference Connected regions in the label image The area of ​​the region, Connected regions in the label image Regional texture entropy, For example, to prevent hyperparameters with a denominator of 0, .

[0056] in, The grayscale terrain capacity refers to the theoretical amount of water required to fill a connected region in a grayscale terrain model. This index is derived from the terrain flooding theory of the watershed algorithm (Meyer, 1994) and can effectively assess the salience of a connected region as an independent physical structure. The larger the value, the more salience the region is and the more likely it is to be the target bubble. This introduces a texture penalty. The more complex the texture, such as the aggregate region, the larger the value and the lower the physical entropy weight. By calculating the physical entropy weight, connected regions with high physical entropy weights are marked as salient structures, while connected regions with low physical entropy weights are marked as background or noise to be merged.

[0057] S004: Perform a region merge operation.

[0058] It should be noted that the core challenge of image segmentation is to eliminate oversegmentation while preserving the true physical features and avoiding the erroneous merging of bubble regions and aggregate regions. This step adopts a hierarchical merging strategy based on physical entropy weights because this strategy follows the principle of gray-scale topography: connected regions with high physical entropy weights and true bubble features can absorb connected regions with low physical entropy weights caused by noise. The adaptive threshold dynamically determines the natural boundaries of different materials, and the pre-termination condition ensures that only naturally connected regions on the gray-scale topography are merged. Merging is terminated when the boundary height first exceeds the threshold, stopping precisely when it connects bubble regions but does not cross into aggregate regions. This preserves the true distribution structure of concrete surface features and lays an accurate foundation for quality assessment.

[0059] Specifically, construct a priority queue. Arranging all edges in ascending order of their minimum grayscale value means that the region with the lowest grayscale barrier will be processed first.

[0060] Determine the merging threshold and apply it to the smoothed image. The Multi-Otsu algorithm is used to obtain the dark / middle boundary threshold. And mid / brightness threshold ,because This corresponds to the boundary between the background matrix and significant protrusions. Exceeding this value indicates crossing into the aggregate region, and merging should be stopped. Therefore, a merging termination threshold is set. .

[0061] The iterative region merging operation is shown in Figure 4. Take out the edges in order If the side corresponding If the condition is met, the merging process will terminate; otherwise, the connected regions will be merged. Connected regions The physical entropy weight of the merged region is the average of the physical entropy weights of the two connected regions, and the grayscale difference of the merged region is the maximum grayscale difference between the two connected regions. The method for obtaining the region area and region texture entropy of the merged region is the same as the method for calculating the region area and region texture entropy of the connected regions, and will not be repeated here.

[0062] For a given connected region, if it has the same minimum gray value across multiple contributing boundaries, then they are merged in the order in which each connected region in the label image is assigned a unique numerical identifier.

[0063] For the merged region, all adjacency relationships of the two connected regions before the merger are transferred to the merged region. When the boundaries of the two connected regions before the merger are involved in subsequent processes, they are treated as edges of the merged region.

[0064] S005: Identify bubble clusters and calculate the quality assessment index of the concrete surface based on the statistical characteristics of the bubble clusters.

[0065] It should be noted that traditional vibration quality assessment relies on a single indicator or linear judgment process, which cannot identify multiple simultaneous defect types, such as insufficient vibration and aggregate segregation, leading to potential quality risks. This step involves multi-dimensional diagnosis because concrete vibration quality manifests on the surface as multiple independent yet interrelated physical characteristics: insufficient vibration is characterized by high air bubble density and low surface compaction; excessive vibration is characterized by uneven aggregate distribution. By setting independent status markers for parallel evaluation, multiple defect states can be detected simultaneously, especially identifying compound defects. This overcomes the limitations of traditional linear judgment, provides construction personnel with process improvement suggestions, and achieves a mapping from surface characteristics to internal quality.

[0066] Specifically, the feature vector of the merged region is extracted. The feature vector includes the region area, region texture entropy, and region grayscale difference after Z-score normalization.

[0067] Gaussian mixture models were used to fit a monocluster model assuming no bubbles and a bicluster model assuming bubbles and background, respectively, and the Bayesian information criterion value was calculated. and According to Raftery's (1995) Bayesian information criterion, the strength of evidence between models is defined by the |ΔBIC| value: when Time is considered weak evidence, indicating that the model differences are not significant. Provide positive evidence to support the superior model. This constitutes strong evidence, and This represents very strong evidence and is the standard for judging whether a bicluster model is superior to a monocluster model in engineering practice. Therefore, if If the data distribution shows two distinct characteristics, then bubbles are identified, and the cluster with the highest grayscale difference and lowest texture entropy is selected as the bubble cluster. If no cluster satisfies both conditions, the average physical entropy weights of all regions in the two clusters are compared, and the cluster with the larger average physical entropy weight is selected as the bubble cluster. The number and total area of ​​the merged regions within the bubble cluster are then calculated. Otherwise, no bubbles are identified, and the concrete vibration quality is output as good. As shown in Figure 5, the Bayesian information criterion is used to strictly compare single and double cluster models, ensuring that bubbles are only identified when the data distribution significantly exhibits two types of characteristics. This avoids misjudging surface texture noise as a defect. At the same time, based on the physical characteristics of bubbles, the real bubble regions are accurately screened, providing a reliable basis for subsequent vibration quality diagnosis. This solves the problems of strong subjectivity and susceptibility to interference from complex surface textures in traditional methods.

[0068] Calculate quality assessment indicators for concrete surfaces, including: bubble density, bubble size uniformity, and surface compactness.

[0069] The bubble density is the ratio of the number of merged regions within a bubble cluster to the total area.

[0070] The bubble size uniformity satisfies:

[0071] ;

[0072] In the formula, To ensure uniformity of bubble size on the concrete surface, The standard deviation of the area of ​​all merged regions in the bubble cluster. This represents the average area of ​​all merged regions within the bubble cluster. The closer the value is to 1, the more uniform the bubble size and the better the process.

[0073] The surface density satisfies:

[0074] ;

[0075] In the formula, The surface density of the concrete surface. This is the sum of the areas of all merged regions within the bubble cluster. The total area of ​​the label image. The closer to 1, the smaller the proportion of bubbles and the denser the surface.

[0076] S006: Output the concrete vibration quality assessment results based on the aforementioned quality assessment indicators.

[0077] Specifically, in response to the surface compaction being less than a set compaction threshold or the bubble density being greater than a set density threshold, the output is "insufficient concrete vibration".

[0078] When the bubble size uniformity is less than a set uniformity threshold and the bubble density is greater than a set density threshold, the system outputs that the concrete is locally under-vibrated.

[0079] When the surface compaction is not less than a set compaction threshold and the bubble density is not greater than a set density threshold, the system outputs that the concrete is well compacted, thus achieving an image analysis-based assessment of the concrete compaction quality.

[0080] The compaction threshold, density threshold, and uniformity threshold are all obtained through image calibration of the concrete surface under standard vibration conditions. The core logic behind not including the comparison of bubble size uniformity with the corresponding threshold in the criteria for good vibration quality is as follows: when the bubble density is not greater than the set threshold, it indicates that the number of bubbles on the concrete surface is at an extremely low level. At this point, regardless of whether the bubble size is uniform, its negative impact on the overall quality of the concrete is already constrained by the dual conditions of meeting the surface compaction standard and the bubble density compliance. Even if a small number of bubbles have size differences, their sparse distribution and extremely low volume proportion will not have a substantial impact on the core mechanical and durability properties of the concrete, such as structural integrity, load-bearing strength, impermeability, and frost resistance, and are insufficient to constitute a quality hazard. Therefore, provided that both surface compaction and bubble density meet the set requirements, there is no need to additionally use bubble size uniformity as a criterion for good vibration.

[0081] This invention also discloses an image analysis-based concrete vibration quality assessment system, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an image analysis-based concrete vibration quality assessment method according to the present invention.

[0082] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0083] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for evaluating the quality of concrete vibration based on image analysis, characterized in that, include: Obtain a grayscale image of the newly vibrated concrete surface; The watershed algorithm is applied to the grayscale image to obtain a label image containing multiple connected regions. A region adjacency graph is constructed based on the label image, and the region area, region texture entropy, and region grayscale difference of each connected region are extracted. The area, texture entropy, and grayscale difference of each connected region are normalized. Based on the normalized area, texture entropy, and grayscale difference of each connected region, the physical entropy weight that measures the structural salience of each connected region is calculated. A priority queue is constructed based on the minimum gray value of the boundary of adjacent connected regions in the regional adjacency graph. An adaptive multi-threshold algorithm is used to determine the merging termination threshold. Adjacent connected regions are merged according to the order of the priority queue until the minimum gray value of the boundary exceeds the merging termination threshold, resulting in merged regions. Feature vectors of each merged region are extracted, and bubble clusters are identified using Gaussian mixture model and Bayesian information criterion. Based on the statistical characteristics of the bubble clusters, quality assessment indicators of the concrete surface are calculated. The quality assessment indicators include at least bubble density, bubble size uniformity, and surface compactness. Output the concrete vibration quality assessment results based on the quality assessment indicators; According to the priority queue order, adjacent connected regions are merged, including: applying the Multi-Otsu algorithm to the grayscale image to divide the grayscale image into dark areas, middle areas and bright areas, and selecting the boundary value between the middle area and the bright area as the merging termination threshold. When the minimum gray value of the boundary of the currently processed adjacent connected regions in the priority queue is not greater than the merging termination threshold, the two adjacent connected regions are merged. The physical entropy weight of the merged region is the average of the physical entropy weights of the two connected regions, and the gray value difference of the merged region is the maximum value of the gray value differences of the two connected regions.

2. The method for evaluating the quality of concrete vibration based on image analysis according to claim 1, characterized in that, The grayscale image is a grayscale image of the concrete surface after being smoothed by Gaussian filtering.

3. The method for evaluating the quality of concrete vibration based on image analysis according to claim 1, characterized in that, The method for obtaining the region texture entropy is as follows: for each connected region in the label image, calculate the gray-level co-occurrence matrix; compress the gray-level levels of the gray-level co-occurrence matrix, and calculate the average value of the gray-level co-occurrence matrix in multiple directions; The region texture entropy of the connected region is calculated based on the average gray-level co-occurrence matrix.

4. The method for evaluating the quality of concrete vibration based on image analysis according to claim 1, characterized in that, The physical entropy weight satisfies: In the formula, Connected regions in the label image Physical entropy weight, Connected regions in the label image The regional grayscale difference Connected regions in the label image The area of ​​the region, Connected regions in the label image The region texture entropy, This is a hyperparameter used to prevent the denominator from being zero.

5. The method for evaluating the quality of concrete vibration based on image analysis according to claim 1, characterized in that, The method of identifying bubble clusters using Gaussian mixture models and Bayesian information criteria includes: constructing a set of feature vectors containing region area, region texture entropy, and region grayscale difference; clustering the feature vector set using Gaussian mixture models to obtain single-cluster models and dual-cluster models; and calculating the Bayesian information criterion value of the single-cluster model. Bayesian information criterion value of bi-cluster model ; in response If the sum of grayscale differences and texture entropy of all connected regions in the dual-cluster model is maximized, then the cluster with the largest sum of grayscale differences and the smallest sum of texture entropy is identified as the bubble cluster; otherwise, no bubbles are identified.

6. The method for evaluating the quality of concrete vibration based on image analysis according to claim 1, characterized in that, The bubble size uniformity satisfies: In the formula, To ensure uniformity of bubble size on the concrete surface, The standard deviation of the area of ​​all merged regions in the bubble cluster. This is the average area of ​​all merged regions within the bubble cluster.

7. The method for evaluating the quality of concrete vibration based on image analysis according to claim 1, characterized in that, The surface density satisfies: In the formula, The surface density of the concrete surface. This is the sum of the areas of all merged regions within the bubble cluster. This represents the total area of ​​the label image.

8. The method for evaluating the quality of concrete vibration based on image analysis according to claim 1, characterized in that, The output concrete vibration quality assessment results include: inadequate concrete vibration when the surface density is less than a set density threshold or the bubble density is greater than a set density threshold; localized insufficient concrete vibration when the bubble size uniformity is less than a set uniformity threshold and the bubble density is greater than a set density threshold; and good concrete vibration when the surface density is not less than a set density threshold and the bubble density is not greater than a set density threshold, thus achieving an image analysis-based assessment of concrete vibration quality.

9. A concrete vibration quality assessment system based on image analysis, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a method for evaluating the quality of concrete vibration based on image analysis according to any one of claims 1-8.

Images