Concrete aggregate quality assessment method and system based on image recognition technology

Abstract

The present invention relates to the technical fields of image recognition and building material testing, and more specifically to a method and system for assessing the quality of concrete aggregates based on image recognition technology. The method comprises the following steps: S1: acquiring a top-view grayscale image of a concrete aggregate pile to generate a contact-enhanced image; S2: binarizing the contact-enhanced image and counting the number of contact points per unit area as the contact point density; S3: constructing a skeleton topological network of the aggregate pile; S4: calculating the degree of match between the actual and restored forms as the form restoration degree; and S5: synthesizing the two determination results to output an aggregate quality grade. By integrating the image recognition and assessment mechanism of contact point density and particle shape restoration degree, the present invention achieves high-precision, standardized, and automated assessment of concrete aggregate quality.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition and building material detection, and in particular to a concrete aggregate quality assessment method and system based on image recognition technology. Background Art

[0002] In the process of concrete production and quality control, the grading structure and particle shape characteristics of aggregates are key factors that directly affect the working performance and durability of concrete. Currently, aggregate quality assessment mainly relies on traditional methods such as manual screening and image manual measurement. These methods are not only inefficient and highly subjective, but also difficult to accurately reflect the inter-particle contact behavior and true particle shape contours in the stacked state.

[0003] At the same time, existing automated image analysis methods mostly focus on single-particle detection or size measurement, failing to effectively integrate key processes such as contact relationship identification, skeleton modeling, and morphological restoration. This results in significant deficiencies in the accuracy of particle shape restoration and gradation determination. Therefore, a concrete aggregate quality assessment method and system based on image recognition technology is urgently needed to address these issues. Summary of the Invention

[0004] Based on the above objectives, the present invention provides a concrete aggregate quality assessment method and system based on image recognition technology.

[0005] The concrete aggregate quality assessment method based on image recognition technology includes the following steps:

[0006] S1: Obtain a top-down grayscale image of the concrete aggregate pile, highlight the contact area between aggregates through shadow enhancement processing, and generate a contact-enhanced image;

[0007] S2: Binarize the contact enhancement image, extract the contact point groups between aggregates, and count the number of contact points per unit area as the contact point density;

[0008] S3: Constructing the skeleton topology network of the aggregate pile based on the spatial distribution of contact points;

[0009] S4: Reversely restore the original geometric shape of the single aggregate particle according to the skeleton topology network, and calculate the matching degree between the actual shape and the restored shape as the shape restoration degree;

[0010] S5: When the contact point density exceeds the preset gradation threshold, the gradation is judged to be unqualified. When the morphological restoration degree is lower than the preset morphological threshold, the particle shape is judged to be unqualified. The aggregate quality grade is output based on the two judgment results.

[0011] Optionally, the S1 specifically includes:

[0012] S11: In a darkroom with controlled lighting conditions, a concrete aggregate sample was laid flat on a pure black background. An industrial camera was positioned 80 cm above the sample and photographed vertically. A ring-shaped LED cold light source was used to provide uniform illumination to acquire a raw RGB image.

[0013] S12: converting the original RGB image into a grayscale image, and performing binary segmentation by setting a fixed threshold to extract the aggregate area, removing pure black background pixels, and generating an aggregate area mask image;

[0014] S13: performing homomorphic filtering on the grayscale image to separate the illumination component and the reflection component, wherein the illumination component is smoothed using a Gaussian low-pass filter to generate an illumination-balanced image;

[0015] S14: performing a morphological black hat operation on the illumination balanced image based on the aggregate region mask to extract the contact concave regions between the aggregates;

[0016] S15: Multiply the black hat image by the gain coefficient and perform pixel-level weighted superposition with the illumination balanced image to form a final contact enhanced image.

[0017] Optionally, the S2 specifically includes:

[0018] S21: The contact enhancement image is subjected to pixel-by-pixel binarization using a local adaptive threshold method. Specifically, the size is A sliding window of pixels is used to scan the image pixel by pixel; then a local threshold is calculated within each window. ; If the pixel value is less than , then the pixel is set as the foreground; otherwise it is set as the background to generate a binary image;

[0019] S22: Performing a morphological opening operation on the binary image, using a circular structure element with a radius of 2 pixels, first corroding and then dilating, and deleting noise areas with an area smaller than 5 pixels to obtain a denoised target contact point candidate image;

[0020] S23: identifying all connected regions in the candidate image after morphological processing, and screening connected regions that meet predetermined conditions as valid contact points;

[0021] S24: Count the number of all effective contact points, and calculate the actual coverage area of ​​the aggregate according to the aggregate area mask image, and then obtain the contact point density , whose expression is: ,in, is the total number of effective contact points; is the actual area of ​​the aggregate region in the image.

[0022] Optionally, the predetermined condition includes:

[0023] Condition 1, the area range is between 8 pixels and 50 pixels;

[0024] Condition 2: The tightness index is greater than 0.7. The tightness calculation formula is: ,in, for tightness; is the area of ​​the connected domain; is the perimeter of the connected domain;

[0025] Condition 3: The centroid of the connected domain is located within the valid area of ​​the aggregate region mask image.

[0026] Optionally, the S3 specifically includes:

[0027] S31: All valid contact points extracted in S2 are used as a two-dimensional vertex set, and the Delaunay triangulation algorithm is executed to generate an initial triangular mesh structure covering the entire aggregate area;

[0028] S32: For the generated initial triangular mesh structure, the circumscribed circle radius of each triangular unit is calculated, and all boundary triangle edges whose circumscribed circle radius is greater than 0.8 times the average aggregate particle size are removed, and only the remaining edges are retained to form a preliminary skeleton branch set;

[0029] S33: Perform medial axis transformation on the skeleton branch set, iteratively erode the boundary area through the morphological thinning algorithm until the skeleton presents a single pixel width, and generate an initial aggregate skeleton network;

[0030] S34: In the initial skeleton network, two types of key nodes are detected, including branch intersections and endpoints. Specifically, nodes with a connectivity greater than or equal to 3 are defined as branch intersections; nodes with a connectivity equal to 1 are defined as endpoints.

[0031] S35: Using the detected key nodes as the vertices of the graph and the skeleton branches as the edges of the graph, an adjacency matrix of the aggregate skeleton topological network is established, wherein the matrix element values ​​store the Euclidean distances between adjacent nodes.

[0032] Optionally, the S4 specifically includes:

[0033] S41: constructing a corresponding polygonal control area with each key node in the skeleton topology network as the center, wherein the boundary of the polygon is formed by the perpendicular bisector of the line connecting the corresponding key node and its adjacent nodes;

[0034] S42: In each polygonal control area, with the corresponding key node as the circle center, the average value of the skeleton branch length set from the corresponding key node to all connection endpoints is calculated as the circle radius, and the minimum enclosing circle is constructed to simulate the ideal original geometric shape of the aggregate in the area;

[0035] S43: extracting the actual contour area corresponding to the polygon in the aggregate area mask image, and simplifying the contour to a polyline polygon composed of 20 key points;

[0036] S44: Based on the distance between the minimum enclosing circle and the actual contour, the morphological restoration degree of the region is calculated by normalization. The formula is: ,in, Morphological restoration, is the distance between the minimum enclosing circle and the actual contour; The diameter of the minimum enclosing circle.

[0037] Optionally, the S42 specifically includes:

[0038] S421: In each polygonal control area, extract all skeleton branches directly connected to the current key node, identify and record the pixel-level coordinate positions between the key node and each connection endpoint;

[0039] S422: Based on the coordinate pairs of the key node and each endpoint, the Euclidean distance of each skeleton branch is calculated as the length of the skeleton branch;

[0040] S423: All skeleton branch length values ​​are combined into a set, and the average value of the set is calculated as the minimum enclosing circle radius in the current control area;

[0041] S424: With the coordinates of the key nodes as the center and the average radius as the radius, a minimum enclosing circle is constructed as the ideal original geometric shape of the aggregate in the simulation area.

[0042] Optionally, the S5 specifically includes:

[0043] S51, convert the gradation determination result of the contact point density in S2 and the particle shape determination result of the morphological reduction degree in S4 into binary determination codes respectively, and record the determination codes as ordered pairs ,in, It is the grade matching flag, 1 means qualified, 0 means unqualified; The particle shape qualification flag, 1 means qualified, 0 means unqualified;

[0044] S52: Encoding according to determination The aggregate quality grade is determined according to the following rules: , the output quality level is level I; if or , then the output quality level is level II; if , then the output quality level is level III;

[0045] S53: Output a structured report including quality grade, measured values ​​of contact point density, and measured values ​​of morphological restoration.

[0046] Optionally, the gradation threshold is defined as 90% of the maximum contact point density of a standard continuously graded aggregate sample; and the morphology threshold is defined as 85% of the minimum morphology restoration degree of a standard cubic aggregate sample.

[0047] The concrete aggregate quality assessment system based on image recognition technology is used to implement the above-mentioned concrete aggregate quality assessment method based on image recognition technology, and includes the following modules:

[0048] Image acquisition and preprocessing module: used to acquire images of concrete aggregate samples laid on a black background plate in a darkroom environment, obtain RGB images, and output contact-enhanced images through morphological black-hat enhancement operations;

[0049] Contact point extraction and density calculation module: used to perform local adaptive binarization and morphological denoising on the contact-enhanced image, extract the set of valid contact points that meet the area, density, and position constraints, calculate the number of contact points per unit area based on the area of ​​the aggregate mask region, and output the contact point density value;

[0050] Skeleton topology reconstruction module: used to perform triangulation with valid contact points as vertices, delete redundant edges to form a skeleton branch set, and construct a skeleton topology adjacency matrix based on the Euclidean distance between nodes;

[0051] Grain shape modeling and matching module: used to divide the control area around each key node, calculate the average length of the skeleton branches to each endpoint to generate the minimum enclosing circle, extract the actual contour of the aggregate in the corresponding area, and simplify it into a broken line contour with a fixed number of key points. Then, the distance between the contour and the circle model is calculated and normalized to generate the degree of morphological restoration.

[0052] Quality grade determination module: used to compare the contact point density value and the morphological restoration value with the preset threshold value, generate a binary determination code, output the aggregate quality grade according to the preset mapping rules, and generate a structured evaluation report containing the grade results and the actual measured values ​​of the indicators.

[0053] Beneficial effects of the present invention:

[0054] The present invention realizes the full-process digital evaluation of concrete aggregate grading and morphological quality by integrating core steps such as standardized image acquisition, automatic contact point identification, aggregate skeleton topology reconstruction and quantitative analysis of particle shape restoration.

[0055] The present invention significantly improves the scientific nature and repeatability of aggregate grading and particle shape quality determination by adopting a joint grading and determination mechanism based on contact point density and particle shape restoration, providing efficient, objective and standardized technical support for raw material screening and quality control in the concrete production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0057] Figure 1 Schematic diagram of a method for assessing the quality of concrete aggregate according to an embodiment of the present invention;

[0058] Figure 2 Schematic diagram of a concrete aggregate quality assessment system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0060] It should be noted that references in the specification to "one embodiment," "an embodiment," "exemplary embodiments," "some embodiments," etc. indicate that the described embodiments may include specific features, structures, or characteristics, but not necessarily every embodiment will include such specific features, structures, or characteristics. Furthermore, when specific features, structures, or characteristics are described in conjunction with an embodiment, it is within the knowledge of persons skilled in the relevant art to implement such features, structures, or characteristics in conjunction with other embodiments (whether or not explicitly described).

[0061] In general, terms can be understood, at least in part, from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending at least in part on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0062] like Figure 1As shown, the concrete aggregate quality assessment method based on image recognition technology includes the following steps:

[0063] S1: Obtain a top-down grayscale image of the concrete aggregate pile, highlight the contact area between aggregates through shadow enhancement processing, and generate a contact-enhanced image;

[0064] S2: Binarize the contact enhancement image, extract the contact point groups between aggregates, and count the number of contact points per unit area as the contact point density;

[0065] S3: Constructing the skeleton topology network of the aggregate pile based on the spatial distribution of contact points;

[0066] S4: Reversely restore the original geometric shape of the single aggregate particle according to the skeleton topology network, and calculate the matching degree between the actual shape and the restored shape as the shape restoration degree;

[0067] S5: When the contact point density exceeds the preset gradation threshold, the gradation is judged to be unqualified. When the morphological restoration degree is lower than the preset morphological threshold, the particle shape is judged to be unqualified. The aggregate quality grade is output based on the two judgment results.

[0068] S1 specifically includes:

[0069] S11: In a darkroom with controlled lighting conditions, a concrete aggregate sample was laid flat on a pure black background. An industrial camera was positioned 80 cm above the sample and photographed vertically. A ring-shaped LED cold light source was used to provide uniform illumination to acquire a raw RGB image.

[0070] S12: converting the original RGB image into a grayscale image, and performing binary segmentation by setting a fixed threshold to extract the aggregate area, remove pure black background pixels, and generate an aggregate area mask image;

[0071] S13: Perform homomorphic filtering on the grayscale image to separate the illumination component and the reflection component, wherein the illumination component is smoothed using a Gaussian low-pass filter (the filter parameter is 5 pixels) to generate an illumination-balanced image;

[0072] S14: Based on the aggregate region mask, a morphological black hat operation is performed on the illumination balanced image to extract the contact concave region between the aggregates. The black hat image calculation formula is: in, To perform the dilation and then erosion operation on the original image, the structural element uses a circular template with a radius of 3 pixels;

[0073] S15: Multiply the black hat image by the gain coefficient and perform pixel-level weighted superposition with the illumination balanced image to form the final contact enhanced image. The superposition formula is: ,in, Enhanced images for contact; For lighting balanced images; The image is a black hat image; the gain coefficient is set to 2.0; the weighting ratios are 0.7 and 0.3 respectively; the above steps, through standardized acquisition, illumination compensation and concave area enhancement processing, can significantly improve the image contrast of the aggregate contact area while eliminating background interference, providing a high-fidelity image basis for subsequent contact point extraction and skeleton topology analysis.

[0074] S2 specifically includes:

[0075] S21: The contact enhancement image is subjected to pixel-by-pixel binarization using a local adaptive threshold method. Specifically, the size is A sliding window of pixels is used to scan the image pixel by pixel; then a local threshold is calculated within each window. , the formula is: ,in: is the local threshold of the current window; is the mean gray value in the window; is the standard deviation of the grayscale value in the window; if the pixel value is less than , the pixel is set as foreground (value is 255); otherwise it is set as background (value is 0) to generate a binary image;

[0076] S22: Performing a morphological opening operation on the binary image, using a circular structure element with a radius of 2 pixels, first corroding and then dilating, and deleting noise areas with an area smaller than 5 pixels to obtain a denoised target contact point candidate image;

[0077] S23: identifying all connected regions in the candidate image after morphological processing, and screening connected regions that meet predetermined conditions as valid contact points;

[0078] S24: Count the number of all effective contact points, and calculate the actual coverage area of ​​the aggregate according to the aggregate area mask image, and then obtain the contact point density , whose expression is: ,in, is the total number of effective contact points; is the actual area of ​​the aggregate region in the image.

[0079] Booking conditions include:

[0080] Condition 1, the area range is between 8 pixels and 50 pixels;

[0081] Condition 2: The tightness index is greater than 0.7. The tightness calculation formula is: ,in, for tightness; is the area of ​​the connected domain; is the perimeter of the connected domain;

[0082] Condition 3: The centroid of the connected domain is located within the valid area of ​​the aggregate region mask image. The above steps effectively extract the true aggregate contact points and exclude non-target noise areas through the adaptive local threshold method and precise morphological screening process. At the same time, the compactness and position constraints are introduced to ensure the physical rationality of the proposed contact points, thereby significantly improving the accuracy and repeatability of the contact point density assessment and providing a high-confidence basis for the gradation structure determination.

[0083] S3 specifically includes:

[0084] S31: All valid contact points extracted in S2 are used as a two-dimensional vertex set, and the Delaunay triangulation algorithm is executed to generate an initial triangular mesh structure covering the entire aggregate area;

[0085] S32: For the generated initial triangular mesh structure, calculate the circumscribed circle radius of each triangular unit, remove all boundary triangle edges whose circumscribed circle radius is greater than 0.8 times the average aggregate particle size, and retain only the remaining edges to form a preliminary skeleton branch set. The circumscribed circle radius calculation formula is: ,in, is the radius of the circumscribed circle of the triangular unit; are the lengths of the three sides of the triangle; is the area of ​​the triangle;

[0086] S33: Perform medial axis transformation on the skeleton branch set, iteratively erode the boundary area through the morphological thinning algorithm until the skeleton presents a single pixel width, and generate an initial aggregate skeleton network;

[0087] S34: In the initial skeleton network, two types of key nodes are detected, including branch intersections and endpoints. Specifically, nodes with a connectivity greater than or equal to 3 are defined as branch intersections; nodes with a connectivity equal to 1 are defined as endpoints.

[0088] S35: Using the detected key nodes as the vertices of the graph and the skeleton branches as the edges of the graph, an adjacency matrix of the aggregate skeleton topology network is established, where the matrix element values ​​store the Euclidean distances between adjacent nodes. The distance calculation formula is: ,in, For nodes With node The Euclidean distance between Node and The coordinate position in the image plane; the above steps, by combining the Delaunay triangulation and medial axis transformation methods, can effectively restore the structural connection relationship between aggregates, and use key nodes to construct a topological network reflecting the geometric distribution characteristics of aggregates, while ensuring computational stability and significantly improving the accuracy of inter-granular connection structure analysis and engineering interpretability.

[0089] S4 specifically includes:

[0090] S41: With each key node in the skeleton topology network as the center, a corresponding polygonal control area is constructed. The boundary of the polygon is formed by the perpendicular bisector of the line connecting the corresponding key node and its adjacent nodes, which is used to define the local single-grain aggregate area;

[0091] S42: In each polygonal control area, with the corresponding key node as the circle center, the average value of the skeleton branch length set from the corresponding key node to all connection endpoints is calculated as the circle radius, and the minimum enclosing circle is constructed to simulate the ideal original geometric shape of the aggregate in the area;

[0092] S43: In the aggregate region mask image, extract the actual contour region corresponding to the polygon, and simplify the contour to a polyline polygon consisting of 20 key points to reduce the contour complexity and retain the main shape features;

[0093] S44: Based on the distance between the minimum enclosing circle and the actual contour, the morphological restoration degree of the region is calculated by normalization. The formula is: ,in, Morphological restoration degree, range [0,1], is the distance between the minimum enclosing circle and the actual contour; The diameter of the minimum enclosing circle; the above steps establish a polygonal control area with the key node as the center and construct a local circular model. Combined with the distance measurement geometric deviation, it can accurately restore the ideal shape of a single particle aggregate and compare its true contour, thereby achieving an objective quantitative evaluation of particle shape quality.

[0094] S42 specifically includes:

[0095] S421: In each polygonal control area, extract all skeleton branches directly connected to the current key node, identify and record the pixel-level coordinate positions between the key node and each connection endpoint;

[0096] S422: Based on the coordinate pairs of the key nodes and each endpoint, the Euclidean distance of each skeleton branch is calculated as the length of the skeleton branch. The calculation formula is: ,in, For the The length of the skeleton branches; is the coordinate of the current key node; Indicates the first Endpoint coordinates;

[0097] S423: All skeleton branch length values ​​are combined into a set, and the average value of the set is calculated as the minimum enclosing circle radius in the current control area. The average value calculation formula is: ,in, is the radius of the minimum enclosing circle; Indicates the number of endpoints connected to the current key node;

[0098] S424: With the coordinates of the key nodes as the center and the average radius as the radius, a minimum enclosing circle is constructed as the ideal original geometric shape of the aggregate in the simulation area. The generated circle is used for subsequent morphological restoration matching. The above steps establish a minimum enclosing circle model based on the mean value of the endpoint branch length of the skeleton structure, which can effectively avoid interference from local boundary distortion, achieve stable geometric reconstruction of the local aggregate particle shape, and improve the structural consistency and credibility of subsequent morphological matching and quality judgment.

[0099] S5 specifically includes:

[0100] S51, convert the gradation determination result of the contact point density in S2 and the particle shape determination result of the morphological reduction degree in S4 into binary determination codes respectively, and record the determination codes as ordered pairs ,in, It is the grade matching flag, 1 means qualified, 0 means unqualified; The particle shape qualification flag, 1 means qualified, 0 means unqualified;

[0101] If the contact point density value is not less than the gradation threshold, the grade matching grid is determined and coded as 1; otherwise, it is coded as 0;

[0102] If the morphological restoration value is not less than the morphological threshold, the particle shape is determined to be qualified and coded as 1; otherwise, it is coded as 0;

[0103] S52: Encoding according to determination The aggregate quality grade is determined according to the following rules: , the output quality level is level I; if or , then the output quality level is level II; if , then the output quality level is level III;

[0104] Table 1 Quantification of judgment rules

[0105]

[0106] S53: Output a structured report containing quality grade, measured values ​​of contact point density, and measured values ​​of morphological restoration degree; the above steps, by constructing a coding and mapping mechanism based on two-dimensional judgment results, can achieve the combined quality grading output of aggregate grading structure and particle shape characteristics, thereby improving the judgment consistency and objectivity of the quality assessment system.

[0107] The gradation threshold is defined as 90% of the maximum contact point density of the standard continuously graded aggregate sample; the morphology threshold is defined as 85% of the minimum morphology reduction of the standard cubic aggregate sample.

[0108] An example of threshold calibration is as follows:

[0109] Standard continuously graded sample (5-25 mm granite): Maximum contact point density = 0.72 / mm², gradation threshold = 0.9 × 0.72 = 0.648 / mm²;

[0110] Standard cube sample (diabase): minimum morphological restoration value = 0.94, morphological threshold = 0.85 × 0.94 = 0.799;

[0111] Table 2 Judgment cases

[0112]

[0113] The technical results are as follows: after verification by 500 sets of engineering samples, the consistency rate of quality grade determination with laboratory screening + manual shape analysis reached 96.3%; the processing time of a single sample was ≤3 seconds, while the traditional method was ≥30 minutes.

[0114] like Figure 2 As shown, the concrete aggregate quality assessment system based on image recognition technology is used to implement the above-mentioned concrete aggregate quality assessment method based on image recognition technology, and includes the following modules:

[0115] Image acquisition and preprocessing module: used to acquire images of concrete aggregate samples laid on a black background plate in a darkroom environment, obtain RGB images, and output contact-enhanced images through morphological black-hat enhancement operations;

[0116] Contact point extraction and density calculation module: used to perform local adaptive binarization and morphological denoising on the contact-enhanced image, extract the set of valid contact points that meet the area, density, and position constraints, calculate the number of contact points per unit area based on the area of ​​the aggregate mask region, and output the contact point density value;

[0117] Skeleton topology reconstruction module: used to perform triangulation with valid contact points as vertices, delete redundant edges to form a skeleton branch set, and construct a skeleton topology adjacency matrix based on the Euclidean distance between nodes;

[0118] Grain shape modeling and matching module: used to divide the control area around each key node, calculate the average length of the skeleton branches to each endpoint to generate the minimum enclosing circle, extract the actual contour of the aggregate in the corresponding area, and simplify it into a broken line contour with a fixed number of key points. Then, the distance between the contour and the circle model is calculated and normalized to generate the degree of morphological restoration.

[0119] Quality grade determination module: used to compare the contact point density value and the morphological restoration value with the preset threshold value, generate a binary determination code, output the aggregate quality grade according to the preset mapping rules, and generate a structured evaluation report containing the grade results and the actual measured values ​​of the indicators.

[0120] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0121] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for assessing the quality of concrete aggregates based on image recognition technology, characterized in that: The following steps are involved: S1: Obtain a top-down grayscale image of the concrete aggregate pile, highlight the contact area between aggregates through shadow enhancement processing, and generate a contact-enhanced image; S2: Binarize the contact enhancement image, extract the contact point groups between aggregates, and count the number of contact points per unit area as the contact point density; The S2 specifically includes: S21: The contact enhancement image is binarized pixel by pixel using the local adaptive threshold method. Specifically, the size is A sliding window of pixels is used to scan the image pixel by pixel; then a local threshold is calculated within each window. ; If the pixel value is less than , then the pixel is set as the foreground; otherwise it is set as the background to generate a binary image; S22: Performing a morphological opening operation on the binary image, using a circular structure element with a radius of 2 pixels, first corroding and then dilating, and deleting noise areas with an area smaller than 5 pixels to obtain a denoised target contact point candidate image; S23: identifying all connected regions in the candidate image after morphological processing, and screening connected regions that meet predetermined conditions as valid contact points; S24: Count the number of all effective contact points, and calculate the actual coverage area of ​​the aggregate according to the aggregate area mask image, and then obtain the contact point density , whose expression is: ,in, is the total number of effective contact points; is the actual area corresponding to the aggregate area in the image; The predetermined conditions include: Condition 1, the area range is between 8 pixels and 50 pixels; Condition 2: The tightness index is greater than 0.

7. The tightness calculation formula is: ,in, for tightness; is the area of ​​the connected domain; is the perimeter of the connected domain; Condition 3: The centroid of the connected domain is located within the valid area of ​​the aggregate region mask image; S3: Constructing the skeleton topology network of the aggregate pile based on the spatial distribution of contact points; S4: Reversely restore the original geometric shape of the single aggregate particle according to the skeleton topology network, and calculate the matching degree between the actual shape and the restored shape as the shape restoration degree; S5: When the contact point density exceeds the preset gradation threshold, the gradation is judged to be unqualified. When the morphological restoration degree is lower than the preset morphological threshold, the particle shape is judged to be unqualified. The aggregate quality grade is output based on the two judgment results.

2. The concrete aggregate quality assessment method based on image recognition technology according to claim 1, characterized in that: Said S1 specifically includes: S11: In a darkroom with controlled lighting conditions, a concrete aggregate sample was laid flat on a pure black background. An industrial camera was positioned 80 cm above the sample and photographed vertically. A ring-shaped LED cold light source was used to provide uniform illumination to acquire a raw RGB image. S12: converting the original RGB image into a grayscale image, and performing binary segmentation by setting a fixed threshold to extract the aggregate area, eliminating pure black background pixels, and generating an aggregate area mask image; S13: performing homomorphic filtering on the grayscale image to separate the illumination component and the reflection component, wherein the illumination component is smoothed using a Gaussian low-pass filter to generate an illumination-balanced image; S14: performing a morphological black hat operation on the illumination balanced image based on the aggregate region mask to extract the contact concave regions between the aggregates; S15: Multiply the black hat image by the gain coefficient and perform pixel-level weighted superposition with the illumination balanced image to form a final contact enhanced image.

3. The concrete aggregate quality assessment method based on image recognition technology according to claim 1, characterized in that: The S3 specifically includes: S31: All valid contact points extracted in S2 are used as a two-dimensional vertex set, and the Delaunay triangulation algorithm is executed to generate an initial triangular mesh structure covering the entire aggregate area; S32: For the generated initial triangular mesh structure, the circumscribed circle radius of each triangular unit is calculated, and all boundary triangle edges whose circumscribed circle radius is greater than 0.8 times the average aggregate particle size are removed, and only the remaining edges are retained to form a preliminary skeleton branch set; S33: Perform medial axis transformation on the skeleton branch set, iteratively erode the boundary area through the morphological thinning algorithm until the skeleton presents a single pixel width, and generate an initial aggregate skeleton network; S34: In the initial skeleton network, two types of key nodes are detected, including branch intersections and endpoints. Specifically, nodes with a connectivity greater than or equal to 3 are defined as branch intersections; nodes with a connectivity equal to 1 are defined as endpoints. S35: Using the detected key nodes as the vertices of the graph and the skeleton branches as the edges of the graph, an adjacency matrix of the aggregate skeleton topological network is established, wherein the matrix element values ​​store the Euclidean distances between adjacent nodes.

4. The concrete aggregate quality assessment method based on image recognition technology according to claim 1, characterized in that: The S4 specifically includes: S41: constructing a corresponding polygonal control area with each key node in the skeleton topology network as the center, wherein the boundary of the polygon is formed by the perpendicular bisector of the line connecting the corresponding key node and its adjacent nodes; S42: In each polygonal control area, with the corresponding key node as the circle center, the average value of the skeleton branch length set from the corresponding key node to all connection endpoints is calculated as the circle radius, and the minimum enclosing circle is constructed to simulate the ideal original geometric shape of the aggregate in the area; S43: extracting the actual contour area corresponding to the polygon in the aggregate area mask image, and simplifying the contour to a polyline polygon composed of 20 key points; S44: Based on the distance between the minimum enclosing circle and the actual contour, the morphological restoration degree of the region is calculated by normalization. The formula is: ,in, Morphological restoration, is the distance between the minimum enclosing circle and the actual contour; The diameter of the minimum enclosing circle.

5. The concrete aggregate quality assessment method based on image recognition technology according to claim 4, characterized in that: The S42 specifically includes: S421: In each polygonal control area, extract all skeleton branches directly connected to the current key node, identify and record the pixel-level coordinate positions between the key node and each connection endpoint; S422: Based on the coordinate pairs of the key node and each endpoint, the Euclidean distance of each skeleton branch is calculated as the length of the skeleton branch; S423: All skeleton branch length values ​​are combined into a set, and the average value of the set is calculated as the minimum enclosing circle radius in the current control area; S424: With the coordinates of the key nodes as the center and the average radius as the radius, a minimum enclosing circle is constructed as the ideal original geometric shape of the aggregate in the simulation area.

6. The concrete aggregate quality assessment method based on image recognition technology according to claim 1, characterized in that: The S5 specifically includes: S51, convert the gradation determination result of the contact point density in S2 and the particle shape determination result of the morphological reduction degree in S4 into binary determination codes respectively, and record the determination codes as ordered pairs ,in, It is the grade matching flag, 1 means qualified, 0 means unqualified; The particle shape qualification flag, 1 means qualified, 0 means unqualified; S52: Encoding according to determination The aggregate quality grade is determined according to the following rules: , the output quality level is level I; if or , then the output quality level is level II; if , then the output quality level is level III; S53: Output a structured report including quality grade, measured values ​​of contact point density, and measured values ​​of morphological restoration.

7. The concrete aggregate quality assessment method based on image recognition technology according to claim 1, characterized in that: The gradation threshold is defined as 90% of the maximum contact point density of the standard continuously graded aggregate sample; the morphology threshold is defined as 85% of the minimum morphology restoration degree of the standard cubic aggregate sample.

8. A concrete aggregate quality assessment system based on image recognition technology, for implementing the concrete aggregate quality assessment method based on image recognition technology according to any one of claims 1 to 7, characterized in that: Includes the following modules: Image acquisition and preprocessing module: used to acquire images of concrete aggregate samples laid on a black background plate in a darkroom environment, obtain RGB images, and output contact-enhanced images through morphological black-hat enhancement operations; Contact point extraction and density calculation module: used to perform local adaptive binarization and morphological denoising on the contact-enhanced image, extract the set of valid contact points that meet the area, density, and position constraints, calculate the number of contact points per unit area based on the area of ​​the aggregate mask region, and output the contact point density value; Skeleton topology reconstruction module: used to perform triangulation with valid contact points as vertices, delete redundant edges to form a skeleton branch set, and construct a skeleton topology adjacency matrix based on the Euclidean distance between nodes; Grain shape modeling and matching module: used to divide the control area around each key node, calculate the average length of the skeleton branches to each endpoint to generate the minimum enclosing circle, extract the actual contour of the aggregate in the corresponding area, and simplify it into a broken line contour with a fixed number of key points. Then, the distance between the contour and the circle model is calculated and normalized to generate the degree of morphological restoration. Quality grade determination module: used to compare the contact point density value and the morphological restoration value with the preset threshold value, generate a binary determination code, output the aggregate quality grade according to the preset mapping rules, and generate a structured evaluation report containing the grade results and the actual measured values ​​of the indicators.

Images