Aggregate grading analysis method based on image segmentation

Abstract

The invention discloses an aggregate grading analysis method based on image segmentation, which comprises the following steps: acquiring a two-dimensional image of cluster aggregate on site, preprocessing the image, correcting image distortion by adopting perspective transformation, segmenting the corrected image, and obtaining mask and edge information of each aggregate. After key parameters of each aggregate are calculated, gradation analysis of the aggregate is carried out, gradation curves are drawn, and errors of the gradation curves are calculated. According to the method, the aggregate grading analysis efficiency is improved, the operation process is simplified, and a scientific basis is provided for selection and application of aggregates in constructional engineering.

Description

Technical Field

[0001] The invention can be applied in technical fields such as engineering and building materials, and can be specifically used for aggregate gradation analysis. Background Art

[0002] Aggregates are key components of building materials such as concrete and asphalt, and are mainly divided into two categories: fine aggregate and coarse aggregate. The particle size and gradation of aggregates have a significant impact on the strength, durability and performance of building materials. The gradation parameters of aggregates are important factors affecting their performance. The size, shape and distribution of aggregates have a direct impact on the density, deformation resistance and durability of the mixture.

[0003] The most common method for obtaining traditional aggregate grading parameters is the screening method. The screening method uses a set of standard sieves to grade aggregates. The aggregate samples are screened into sieves of different particle sizes step by step through mechanical vibration or manual operation, and then the mass of aggregate remaining on each layer of sieve is weighed separately. Based on these data, the mass percentage of aggregate of each particle size is calculated, and the particle size distribution curve is drawn. This method has been widely used in engineering and construction fields to evaluate and control the quality of aggregates. Although the screening method is simple to operate, it has problems such as low efficiency and insufficient screening accuracy for irregular particles.

[0004] With the increasing requirements for aggregate performance in the construction field and the pursuit of construction efficiency and precision, the traditional aggregate grading analysis method can no longer meet the needs of modern construction projects. The traditional method is inefficient and the screening accuracy is insufficient when dealing with irregular particles. In addition, the construction field is paying more and more attention to environmental impact and cost control during the construction process, which requires the design and construction process of the mixture to be more scientific, precise and efficient. Summary of the invention

[0005] The present invention relates to a new method for two-dimensional image processing and modeling of aggregates. The method implements the segmentation and parameter calculation of aggregates through a series of image processing steps for clustered aggregate images collected on site, and finally performs gradation analysis and evaluation.

[0006] The present invention comprises the following steps:

[0007] A collects a two-dimensional image of the aggregate;

[0008] B performs perspective transformation correction on the two-dimensional image, and proportionally adjusts the size of the two-dimensional image by bicubic interpolation;

[0009] C. segmenting the image to obtain mask and edge information of the aggregate;

[0010] D. calculating aggregate parameters through the mask and edge information, wherein the aggregate parameters include minimum convex polygon width, aggregate area, equivalent ellipse major semi-axis, equivalent ellipse minor semi-axis and equivalent ellipse volume;

[0011] E. Perform grading analysis based on the aggregate parameters and draw a grading curve.

[0012] Furthermore, the perspective transformation correction step includes aligning four vertices of the image with four vertices of the square bounding box.

[0013] Furthermore, a Segment Anything model is used to segment the two-dimensional image.

[0014] Furthermore, proportionally adjusting the size of the two-dimensional image by bicubic interpolation method specifically includes scaling the input image to a specified size by using cv2.res i ze function, and adjusting the size of the input image by bicubic interpolation method 4 to correct image distortion.

[0015] Furthermore, the image segmentation step includes using a Segment Anything model to obtain mask and edge information of the aggregate, and processing the edge information of each aggregate through an ellipse fitting algorithm to obtain a fitting ellipse.

[0016] Furthermore, the step of calculating the parameters includes:

[0017] Calculate the minimum convex polygon width using the provided formula;

[0018] Use the mask_generator.generate method in SAM to calculate the aggregate area using the mask and edge information;

[0019] Use the cv2.fitElli pse method in OpenCV to calculate the major and minor semi-axes of the equivalent ellipse using the ellipse fitting algorithm;

[0020] Calculate the equivalent ellipse volume using the formula provided.

[0021] Furthermore, the gradation analysis step includes: performing gradation calculation using particle size parameters and quality parameters; drawing a gradation curve and calculating an error according to a provided formula.

[0022] Specifically, the present invention first preprocesses three groups of clustered aggregate images collected on site to remove noise and interference and enhance image quality. Subsequently, the preprocessed images are segmented using an advanced Segment Anything Model (SAM) to identify and separate individual aggregates in the images.

[0023] After completing the segmentation, the present invention further calculates the parameter information of each aggregate in the image. Based on the calculated parameter information, the present invention performs gradation analysis on the three groups of aggregates. By statistically calculating and calculating the proportions of aggregates with different particle sizes, a gradation curve of the aggregates is drawn. The gradation curve can intuitively reflect the gradation of the aggregates and provide a scientific basis for the design and construction of the mixture.

[0024] The present invention not only improves the efficiency of aggregate gradation analysis, but also greatly simplifies the analysis process and reduces manpower and time costs. Through the present invention, rapid and comprehensive analysis and evaluation of aggregates can be achieved, providing strong support for aggregate selection and application in engineering practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a comparison diagram of the first group of two-dimensional image gradation curves of the implementation case of the present invention;

[0026] Figure 2 A comparison diagram of the second group of two-dimensional image gradation curves of the implementation case of the present invention;

[0027] Figure 3 This is a comparison diagram of the third group of two-dimensional image gradation curves of the implementation case of the present invention. DETAILED DESCRIPTION

[0028] The present invention collects two-dimensional images of aggregates, performs image preprocessing and segmentation, and finally realizes analysis and evaluation of aggregate gradation.

[0029] First, two-dimensional images of clustered aggregates are collected on site. The so-called clustered aggregate image refers to a two-dimensional image obtained by photographing multiple aggregates through an image acquisition device. During the acquisition process, the aggregates are spread out and placed in a square boundary box. The data is divided into three groups: the first group contains 1645 aggregates, the second group contains 1728 aggregates, and the third group contains 1665 aggregates, totaling 5038 aggregates. The distribution of each group of data in different gradations is shown in the table.

[0030]

[0031] Next, the collected two-dimensional images are preprocessed. Due to reasons such as the shooting angle, the cluster aggregate images collected on site may have image distortion problems. Distortion refers to the deviation of the shape, size or position of the object in the image from the actual situation due to the characteristics of the camera lens or the camera itself. Therefore, the image is corrected for distortion using perspective transformation so that the four vertices of the transformed image coincide with the four vertices of the square bounding box. Perspective transformation is an important image processing technology used to transform an image from one perspective viewpoint to another, correct the perspective distortion in the image, and make the objects in the image have a more natural shape and proportion visually.

[0032] The perspective-transformed aggregate images were then segmented using the Segment Anything Model (SAM) developed by Meta AI. SAM was trained on a dataset of 11 million images and 1.1 billion masks and has strong performance, accurately segmenting each aggregate in the image. Image segmentation is a technique in computer vision that is used to divide an image into multiple meaningful parts or regions, separating objects or areas of interest in an image from the background for further analysis or processing.

[0033] After the segmentation is completed, the best and time-saving parameter combination is found to obtain the mask of each aggregate. The mask is a binary image in image processing, which is used to mark the area or object of interest. Each pixel of the mask image is either 0 (representing the background) or 1 (representing the foreground).

[0034] After obtaining the mask and edge information of each aggregate, two key parameters can be obtained or calculated: the minimum convex polygon width (mw) and the aggregate area (S).

[0035] The width of a convex polygon refers to the width of the circumscribed rectangle, which is the smallest rectangle that encloses the entire convex polygon, whose boundary is tangent to the boundary of the polygon, and whose sides are parallel to the coordinate axis. In the application, first take the convex hull of the edge of the aggregate (i.e. a convex polygon), and then calculate the minimum value of the width of the convex polygon (circumscribed rectangle), which is the minimum convex polygon width of the aggregate. The calculation formula is as follows:

[0036]

[0037] Among them, mw is the minimum convex polygon width, w is the width of the circumscribed rectangle after taking the convex hull of the aggregate edge, and w i It is the width of the i-th circumscribed rectangle, and min means taking the minimum value.

[0038] The aggregate area (S) can be directly obtained from the aggregate mask and edge information by using the mask_generator.generate method in SAM, whose output dictionary contains the area information.

[0039] The edge information of each aggregate is processed by the ellipse fitting algorithm to obtain the fitted ellipse. Using the least squares method to fit an ellipse is a common image processing technique used to estimate the parameters of an ellipse from a set of discrete points. The least squares method finds the parameters of the best fitting curve by minimizing the sum of squares of the residuals between the data points and the fitting curve. The specific method is to use the cv2.fitEllipse method in OpenCV.

[0040] According to the fitted ellipse information, three key parameters of each ellipse can be obtained or calculated: the equivalent ellipse major semi-axis (a), the equivalent ellipse minor semi-axis (b) and the equivalent ellipse volume (V).

[0041] The major and minor axes are two important parameters of an ellipse, used to describe the shape and size of an ellipse. As important parameters of an ellipse, the equivalent ellipse major axis (a) and the equivalent ellipse minor axis (b) have been included in the fitted ellipse information and can be directly obtained by using the cv2.fitEllipse method in OpenCV, whose output dictionary contains the major and minor axis information.

[0042] The calculation formula of equivalent ellipse volume (V) is as follows:

[0043] V=4 / 3*π*0.125*a*b*0.7*b

[0044] Where V is the equivalent ellipse volume, a is the equivalent ellipse major semi-axis, and b is the equivalent ellipse minor semi-axis.

[0045] In summary, by processing the two-dimensional images of aggregates obtained by cluster scanning, five parameters of each aggregate were obtained: mw, S, a, b, and V, in preparation for the subsequent grading analysis.

[0046] like Figure 1-3 As shown in the figure, these parameter information is used for gradation analysis. In traditional gradation analysis, data is usually plotted into a curve, namely the gradation curve, with the horizontal axis representing the particle size and the vertical axis representing the passing percentage. The particle size is used to determine whether the aggregate can pass through the current sieve, so as to screen the aggregate into different gradations. The two parameters mw and b are used to represent the particle size, and they are called particle size parameters. The passing percentage is used to determine the proportion of all aggregates that can pass through the current sieve. Since it is impossible to directly obtain the aggregate mass from the two-dimensional image, the two parameters S and V are used instead of the mass, and they are called mass parameters.

[0047] After selecting the particle size parameters and quality parameters, the aggregate is screened into different gradations according to the particle size parameters, and then the sub-screening residue, sub-screening residue percentage, and cumulative screening residue percentage of each gradation are calculated according to the quality parameters, and then the passing percentage is calculated. The grading curves calculated for the three sets of data are shown in the attached Figure 1 , 2 、3.

[0048] From the qualitative analysis of the grading curve, it can be concluded that the selection of particle size parameters has little effect on the shape of the curve. In order to simplify the calculation, b can be used as the particle size parameter to draw the curve without additionally calculating mw. b can be simply obtained from the fitting ellipse information, while the calculation of mw is more complicated.

[0049] In order to quantitatively analyze the errors between gradations, for the above three groups of data, the manual screening grading curve of each group is used as the standard to calculate the errors of the remaining grading curves. The error calculation process is as follows: first calculate the difference between the passing percentage of the manual screening grading curve and the passing percentage of the grading curve on each gradation, then take the absolute value of the difference (a total of 7 values), and then sum it up to get the error. The calculation formula is as follows:

[0050]

[0051] Where E is the error, i is the gradation (from 1 to 7), x i is the passing percentage of the i-th gradation of the manual screening grading curve, and y is the passing percentage of the i-th gradation of the grading curve for which the error needs to be calculated.

[0052] The errors of each grade distribution curve at each grade distribution are shown in the table.

[0053]

[0054] According to the error calculation results, it can be concluded that the selection of particle size parameters has little effect on the results; the selection of mass parameters has a greater impact on the results; and the results of selecting V as the mass parameter are better.

[0055] In summary, the present invention provides a new method based on image processing and modeling, which realizes efficient aggregate gradation analysis by segmenting and calculating parameters of two-dimensional images of aggregates.

[0056] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which are all covered by the protection scope of the present invention.

Claims

1. An aggregate gradation analysis method based on image processing and modeling, comprising the following steps: A collects a two-dimensional image of the aggregate; B performs perspective transformation correction on the two-dimensional image, and proportionally adjusts the size of the two-dimensional image by bicubic interpolation; C. segmenting the image to obtain mask and edge information of the aggregate; D. calculating aggregate parameters through the mask and edge information, wherein the aggregate parameters include minimum convex polygon width, aggregate area, equivalent ellipse major semi-axis, equivalent ellipse minor semi-axis and equivalent ellipse volume; E. Perform grading analysis based on the aggregate parameters and draw a grading curve.

2. The aggregate gradation analysis method based on image processing and modeling according to claim 1 is characterized in that: The perspective transformation correction step includes aligning four vertices of the image with four vertices of the square bounding box.

3. The aggregate gradation analysis method based on image processing and modeling according to claim 1 is characterized in that: The two-dimensional image is segmented using a Segment Anything model.

4. The aggregate gradation analysis method based on image processing and modeling according to claim 1 is characterized in that: Proportionally adjusting the size of the two-dimensional image by bicubic interpolation specifically includes scaling the input image to a specified size by using a cv2.resize function, and adjusting the size of the input image by bicubic interpolation 4 to correct image distortion.

5. The aggregate gradation analysis method based on image processing and modeling according to claim 1 is characterized in that: The image segmentation step includes using the Segment Anything model to obtain the mask and edge information of the aggregate, and processing the edge information of each aggregate through an ellipse fitting algorithm to obtain a fitting ellipse.

6. The aggregate gradation analysis method based on image processing and modeling according to claim 1 is characterized in that: The steps of calculating the parameters include: Calculate the minimum convex polygon width using the provided formula; Use the mask_generator.generate method in SAM to calculate the aggregate area using the mask and edge information; Use the cv2.fitEllipse method in opencv to calculate the equivalent ellipse major and minor axes using the ellipse fitting algorithm; Calculate the equivalent ellipse volume using the formula provided.

7. The aggregate gradation analysis method based on image processing and modeling according to claim 1 is characterized in that: The gradation analysis step includes: using particle size parameters and quality parameters to perform gradation calculation; drawing a gradation curve and calculating errors.

Images