Pre-processing method for asphalt concrete fault image modeling suitable for aggregate adhesion removal
Through Gauss filtering and morphological processing technology, combined with the similarity fit of adjacent tomographic images, the noise and aggregate adhesion problems in the tomographic images are solved, and efficient three-dimensional reconstruction model generation is achieved, suitable for the mesoscopic structure analysis of asphalt concrete and Mechanical simulation.
Patent Information
- Application Number
- CN202110600135.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-05-31
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2041-05-31
AI Technical Summary
In the existing three-dimensional reconstruction technology of tomographic images, noise and aggregate adhesion problems caused by scanning equipment accuracy limitations, scanning angle errors, uneven exposure and uneven aggregate density, resulting in a large number of false adhesions, holes and sheets in the three-dimensional reconstruction model. aggregates affect the morphological characteristics and accuracy of the model.
Gauss filtering, morphological expansion corrosion treatment, and adjacent tomographic image similarity fitting technology are used to pre-process the tomographic images, false adhesions and holes are removed, good interlayer connections are constructed, and three-dimensional reconstruction is completed with the combined body drawing technology. Specific steps include tomographic image scanning, high pixel interpolation processing, grayscale processing, filtering smoothing processing, morphological expansion corrosion processing, binary processing, pixel reduction processing, adjacent tomographic image similarity fitting processing and volume drawing construction mold.
Effectively remove erroneous adhesions and holes in tomographic images, improve the boundary contrast of aggregate-asphalt glue, and generate a three-dimensional model with good morphological characteristics. It is suitable for mesoscopic structural feature analysis and numerical simulation of mechanical behavior, improving modeling efficiency.
Smart Images

Figure CN113256794B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of three-dimensional reconstruction of tomographic images, and more specifically to an asphalt concrete tomographic image modeling pre-processing method suitable for removing aggregate adhesion. Background Art
[0002] The tomographic image is collected by rotating the probe of computed tomography (CT) around the surface of the asphalt concrete specimen, or by using magnetic resonance (MR) equipment to collect three-dimensional information around the asphalt concrete specimen at the same time, and then arranged in the order of collection after processing. The tomographic image can accurately show the morphological characteristics, density and contour boundaries of the aggregate and asphalt mortar inside the specimen. The processing results can be used for subsequent micro-structural feature analysis and mechanical behavior numerical simulation after three-dimensional reconstruction.
[0003] Asphalt concrete three-dimensional reconstruction technology can use binarization processing technology to binarize the continuous tomographic images obtained by scanning to obtain a binary tomographic image, and then use volume rendering technology to establish a three-dimensional micro-structure model of the specimen based on the obtained binary tomographic image.
[0004] The pixel information of the image in the fault binary image is represented only by the binary code "0" and "1". The pixel information "0" means that the pixel is dark, and the pixel information "1" means that the pixel is bright. The dark area composed of a plurality of dark pixels arranged in a plane represents asphalt mortar, and the bright area composed of a plurality of bright pixels arranged in a plane represents aggregate.
[0005] Volume rendering technology is to convert the binary information of pixels in layered images into Expanded along the initial acquisition direction z to voxel binary information Then the volume pixel information A technology that generates three-dimensional models by combining three-dimensional coordinates.
[0006] Gauss filter processing technology can use the set Gauss operator to process image noise through the algorithm and smooth the image.
[0007] The existing three-dimensional reconstruction technology can generally obtain a binary image by filtering, morphological processing and image binarization of the scanned tomographic image. The purpose of this step is to filter out the impurity information in the tomographic image that is irrelevant to the data required for three-dimensional reconstruction, so as to improve the targeted modeling and simplify the calculation amount of three-dimensional reconstruction. The binary image is then reconstructed by volume rendering or surface rendering technology. However, since the tomographic image is affected by the accuracy limitation of the scanning equipment, the scanning angle error, the uneven image exposure and the uneven density of the asphalt mixture aggregate, the layered image generated by the existing tomographic imaging technology usually has a large number of noise points and aggregate adhesion. In the area with insufficient exposure and the area with concentrated fine aggregates, the aggregate-asphalt mortar boundary is easily blurred, resulting in discontinuity between adjacent layered images and reducing the aggregate separation effect of the layered image binarization processing. This effect will cause a large number of aggregate error adhesions, aggregate error holes and flaky aggregates in the three-dimensional reconstructed model obtained by volume rendering.
[0008] Patent document with application number CN201210172375.7 discloses a method for reconstructing aggregate micro-physical models based on X-ray CT images of asphalt pavement test pieces. The document identifies the contours of independent aggregates in the tomographic images of asphalt pavement test pieces based on a target recognition algorithm and establishes an aggregate edge bounding box array to determine the upper and lower thresholds of aggregate particle size, and then screens out effective edges that meet the particle size requirements based on the thresholds, and then establishes an asphalt mixture micro-physical model. While this method screens out invalid edges based on a loop program, it will cause the generated model to lose a large number of morphological features, which is manifested in the generated aggregate surface having a large number of regular edges and corners, and being unable to screen out layered independent facets with effective particle size.
[0009] Patent document with application number CN201610553499.8 discloses a method for 3D reconstruction of coke microstructure images. The document uses 3D Slicer software based on volume rendering technology to perform 3D reconstruction of 2D microstructure images of coke fault sequences. Although the modeling results generated by this modeling method of directly reconstructing the 3D tomographic images without good interlayer relationships retain good morphological features, there are a large number of problems such as erroneous structural adhesion, holes and independent facets in the modeling results.
[0010] The above problems can be manually cleared and processed through 3D reconstruction software later, but it will consume a lot of manpower, material resources and time. Summary of the invention
[0011] The present invention aims at the deficiencies of the prior art, overcomes the shortcomings of the prior art, and proposes a pre-processing method for asphalt concrete fault image modeling suitable for removing aggregate adhesion. The method pre-processes the fault image based on Gauss filter processing technology, image morphology processing technology and adjacent fault image similarity fitting technology, constructs the inter-layer connection of the fault image along the acquisition direction, can remove the wrong adhesion, wrong holes and independent facets that appear in the fault image scanning process and the binarization process, and then combines the volume rendering technology to complete the three-dimensional reconstruction of the asphalt concrete specimen model.
[0012] To achieve the above object, the technical solution of the present invention is as follows:
[0013] A pre-processing method for asphalt concrete fault image modeling suitable for aggregate adhesion removal includes fault image scanning, high pixel interpolation processing, grayscale processing, filtering and smoothing processing, morphological expansion and corrosion processing, binarization processing, pixel restoration processing, adjacent fault image similarity fitting processing, volume rendering modeling and fault image regeneration processing, all of which can be completed using Matlab loop program or related modeling software; the specific steps are as follows:
[0014] The first step is tomographic image scanning: CT scanning of the asphalt concrete specimen, to perform tomographic image scanning of the specimen in its initial state to obtain a tomographic image, and the acquisition direction is set to the positive direction of the z-axis.
[0015] The second step is high-pixel interpolation processing: based on the interpolation method, the obtained tomographic image is subjected to high-pixel interpolation processing to obtain a high-pixel tomographic image whose length and width pixel values are more than twice the original length and width pixel values.
[0016] The third step is grayscale processing and filtering and smoothing processing: according to the acquisition order, the high-pixel tomographic image is grayscale processed to obtain a high-pixel tomographic grayscale image; after the grayscale is completed, the high-pixel tomographic grayscale image is smoothed using the Gauss filtering algorithm to obtain a smoothed high-pixel tomographic grayscale image.
[0017] The fourth step is morphological dilation and erosion processing and binarization processing: Since the exposure of a single tomographic image is unevenly distributed in concentric rings, the single tomographic image is divided into a circle and several circular ring areas from the inside to the outside by annular segmentation according to the size of the image. If the outermost circle image is not in the shape of a ring, it is segmented as a ring until all valid parts of the layered image are segmented, and the morphological processing is performed on the regions and then assembled and restored to complete the morphological processing of the single tomographic image; then the binarization processing of all tomographic images is completed in batches according to the acquisition order to obtain a high-pixel tomographic binary image.
[0018] Step 5: Pixel restoration processing: Perform pixel restoration processing on the obtained high-pixel tomographic binary image to obtain a tomographic binary image.
[0019] Step 6: Similarity fitting processing of adjacent tomographic images: Adopting the similarity fitting technology of adjacent tomographic images to construct the inter-layer connection along the acquisition direction of the obtained tomographic binary image, the steps are as follows;
[0020] (1) Choose between cross expansion treatment and corrosion treatment according to actual needs
[0021] Intersection expansion processing: add the pixel information values at the same position in a continuous fault binary images, and import the addition results into a blank new image of the same pixel size as the original fault image according to the pixel position. The label of the blank new image is the smaller value of the two adjacent fault images; the pixel information at the position where the addition result of the new image pixel information value is "0" remains unchanged and displays dark; the pixel information at the position where the addition result of the new image pixel information value is "1" is corrected to "0" and displays dark; the pixel information at the position where the addition result of the new image pixel information value is "≥2" is corrected to "1" and displays white; after the correction is completed, the white and bright aggregate area of the new image is expanded, and the asphalt mortar holes inside the aggregate are deleted.
[0022] Take and erode processing: add the pixel information values at the same position in b consecutive fault binary images, and import the addition results into a blank new image of the same pixel size as the original fault image according to the pixel position. The label of the blank new image is the smaller value of the two adjacent fault images; the pixel information at the position where the addition result of the new image pixel information value is "0" remains unchanged and displays dark; the pixel information at the position where the addition result of the new image pixel information value is "1" remains unchanged and displays white and bright; the pixel information at the position where the addition result of the new image pixel information value is "≥2" is corrected to "1" and displays white and bright; after the correction is completed, the white and bright aggregate area of the new image is corroded, and the asphalt mortar holes inside the aggregate are deleted.
[0023] (2) The similarity fitting processing of subsequent adjacent tomographic images is completed in sequence, and new images are generated in sequence. The sequential set of new images constitutes a new tomographic binary image atlas.
[0024] Step 7: Volume rendering modeling: Based on volume rendering technology, the new fault binary image atlas obtained in the sixth step is reconstructed in three dimensions to obtain an asphalt concrete model.
[0025] In the eighth step, the model obtained in the seventh step is subjected to tomographic image regeneration along any radial direction, and a tomographic binary image is obtained through binarization processing.
[0026] Step 9, tomographic image regeneration processing: repeat steps 6 to 8 for the tomographic binary image set regenerated in step 8 to generate a new tomographic binary image set.
[0027] In the tenth step, the volume rendering technology is used to reconstruct the asphalt concrete specimen model in three dimensions based on the new fault binary image set obtained in the ninth step.
[0028] Preferably, the asphalt concrete specimens include cuboids, cylinders and specimens having other shapes and size characteristics.
[0029] Preferably, the tomographic image types include RGB color images, grayscale images and binary images.
[0030] Preferably, the length and width of the tomographic image are required to completely include the tomographic image of the asphalt concrete specimen, and the excess part is filled with darkness.
[0031] Preferably, the high-pixel interpolation processing requires converting the original pixel image into a high-pixel image in equal proportion.
[0032] Preferably, the tomographic grayscale image includes a uint8-type grayscale image with a grayscale eigenvalue range of 0-255 and a uint16-type grayscale image with a grayscale eigenvalue range of 0-65535.
[0033] Preferably, the morphological dilation and corrosion processing uses a star-shaped kernel function operator.
[0034] Preferably, the pixel restoration process requires that the high-pixel image be restored to the original pixel image in equal proportion.
[0035] Preferably, the values of a and b in the adjacent tomographic image similarity fitting process are required to be ≥2 and ≤10, and four categories are proposed according to the order and processing method: top-down intersection dilation, top-down union erosion, bottom-up intersection dilation and bottom-up union erosion, but it does not exclude the existence of other categories on the basis of the four categories, and the above categories can be used in combination according to the processing purpose.
[0036] Preferably, in the tomographic image regeneration step, any radial direction is required to be perpendicular to the initial scanning direction z in three-dimensional space.
[0037] The present invention has the following beneficial effects: based on the Gauss filter algorithm, the pixel grayscale value of the fault grayscale image is smoothed, which can improve the contrast of the aggregate-asphalt mortar boundary and the stability of the grayscale value of the aggregate area of the fault grayscale image; based on the annular distribution law of the CT scanning exposure, each fault grayscale image is annularly partitioned and reorganized, which can improve the pertinence and applicability of the threshold operator generated by the algorithm program; based on the similarity fitting processing technology of adjacent fault images, the fault binary image is constructed along the acquisition direction. This method can remove the erroneous adhesion, erroneous holes and independent facets that appear in the fault image scanning process and the binarization process, and then use the volume rendering method to establish a three-dimensional asphalt concrete model for the fault binary image with good interlayer connection. The model can be suitable for subsequent micro-structure feature analysis and mechanical behavior numerical simulation. The above process and method can complete the batch processing of fault images by editing the loop structure through Matlab software, and the modeling efficiency is high. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 It is a flowchart of the present invention;
[0039] Figure 2 This is a flow chart of the similarity fitting technology for adjacent tomographic images (intersection expansion from top to bottom);
[0040] Figure 3 This is a flow chart of the similarity fitting technology for adjacent tomographic images (intersection expansion from bottom to top);
[0041] Figure 4 It is the flow chart of the similarity fitting technology of adjacent tomographic images (taking and eroding from top to bottom);
[0042] Figure 5 It is a flow chart of the similarity fitting technology of adjacent tomographic images (taking and eroding from bottom to top);
[0043] Figure 6 This is a schematic diagram of the coordinates of each point and the scanning results after CT scanning (initial scanning);
[0044] Figure 7 This is a schematic diagram of the coordinates of each point after high-pixel interpolation processing;
[0045] Figure 8 Schematic diagram of morphological processing based on ring segmentation;
[0046] Fig. 9 This is a schematic diagram of the coordinates of each point after pixel restoration processing;
[0047] Fig.10 It is a schematic diagram of the direction and results of regenerating an arbitrary radial tomographic image;
[0048] Fig.11This is the principle diagram of similarity fitting processing and intersection expansion of two adjacent tomographic images;
[0049] Fig.12 This is the principle diagram of the similarity fitting processing - merge and erode of two adjacent tomographic images. DETAILED DESCRIPTION
[0050] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and technical solutions.
[0051] The process of the present invention is as follows Figure 1 The detailed steps of this embodiment are as follows:
[0052] The first step is to scan the asphalt concrete specimen, refer to Figure 1 and Figure 6 For a cylindrical specimen with a height of 75 mm and a diameter of 65 mm, a tomographic image scan of the initial state is performed along the initial scanning direction (positive direction of the z axis), and the interval between scanning layers is 0.1 mm. A total of 750 tomographic color images can be obtained, and a single image is a rectangular image with a length of 1507 pixels and a width of 914 pixels;
[0053] The second step is to perform high-pixel interpolation processing on the obtained tomographic image, referring to Figure 7 , a high-pixel color tomographic image with a single image pixel of 3840*2329 is obtained, and the upper left corner vertex of each image is used as the coordinate origin O(1, 1), and a two-dimensional Cartesian coordinate system (x, y) is set;
[0054] The third step is to import the high-pixel tomographic color images into Matlab in sequence, perform grayscale processing on the high-pixel tomographic color images to obtain high-pixel tomographic grayscale images, and after the grayscale processing is completed, use the Gauss filter algorithm to smooth the high-pixel tomographic grayscale images to obtain smoothed high-pixel tomographic grayscale images;
[0055] The fourth step is to perform morphological processing on the smoothed high-pixel grayscale tomographic images in sequence, referring to Figure 8 , the smoothed tomographic grayscale image is imported into Matlab in the order of acquisition. Since the exposure of a single tomographic image is unevenly distributed in concentric rings, the single tomographic image is divided into 1 circular area and 3 circular ring areas from point A from the inside to the outside using the ring segmentation method, and the morphological processing is performed on each area and then assembled and restored to complete the morphological processing of the single tomographic image. After that, the loop program is used to batch complete the binarization processing of all tomographic images in the order of acquisition to obtain a high-pixel tomographic binary image;
[0056] The fifth step is to perform pixel restoration processing on the obtained high-pixel tomographic binary image: Fig. 9, and obtain the tomographic binary image, whose single image pixel is 1507*914;
[0057] Step 6: The obtained tomographic binary image is subjected to similarity fitting processing of adjacent tomographic images. (1) Intersection dilation processing is performed and the binary tomographic image is imported into Matlab in the order of acquisition. Figure 2 , Figure 3 and Fig.11 , add the pixel information values at the same position in two consecutive binary tomographic images, and import the addition results into a blank new image of the same pixel size as the original tomographic image according to the pixel position. The label of the blank new image is the smaller value of the two adjacent tomographic images. The pixel information of the new image whose pixel information value addition result is "0" remains unchanged and displays dark. The pixel information of the new image whose pixel information value addition result is "1" is corrected to "0" and displays dark. The pixel information of the new image whose pixel information value addition result is "2" is corrected to "1" and displays white. After the correction is completed, the white and bright aggregate area of the new image is expanded to delete the asphalt mortar holes inside the aggregate. (2) Take and erode, import the binary tomographic images into Matlab in the order of acquisition, refer to Figure 4 , Figure 5 and Fig.12 , add the pixel information values at the same position in two consecutive binary tomographic images, and import the addition result into a blank new image of the same pixel size as the original tomographic image according to the pixel position. The label of the blank new image is the smaller value of the two adjacent tomographic images. The pixel information of the new image whose pixel information value addition result is "0" remains unchanged and displays dark, the pixel information of the new image whose pixel information value addition result is "1" remains unchanged and displays white and bright, and the pixel information of the new image whose pixel information value addition result is "2" is corrected to "1" and displays white and bright. After the correction is completed, the white and bright aggregate area of the new image is corroded, and the asphalt mortar holes inside the aggregate are deleted. The subsequent adjacent tomographic images are similarly fitted in sequence, and a new image is generated. The new images are arranged in sequence to form a new tomographic binary image atlas 1;
[0058] Step 7: Perform three-dimensional reconstruction on the obtained new atlas 1 based on volume rendering technology to obtain a cylindrical model 1;
[0059] Step 8, Reference Fig.10 , regenerate the tomographic image of the obtained cylindrical model 1 along any radial direction, and obtain the tomographic binary image 2 through binarization processing;
[0060] Step 9, repeating steps 5 to 8 for the regenerated tomographic binary image 2 to generate a new tomographic binary image;
[0061] The tenth step is to use volume rendering technology to complete the three-dimensional reconstruction of the cylindrical asphalt concrete specimen .stl model on the obtained fault binary image.
[0062] The innovations and advantages of the present invention are:
[0063] 1. High-pixel interpolation processing and pixel restoration processing are introduced to expand the selection range of kernel function operators in the morphological expansion and corrosion processing process, and reduce the change of the shape of the original image aggregate and the aggregate loss caused by the morphological expansion and corrosion processing, while ensuring the reasonable amount of calculation.
[0064] 2. The image filtering and smoothing processing based on Gauss algorithm is introduced. This method can sharpen the boundary between aggregate and mortar and improve the boundary effect of subsequent morphological expansion and corrosion processing and binarization processing.
[0065] 3. A morphological processing method based on annular cutting was adopted. After determining the center A of the image, the annular segmentation method was used to divide the single tomographic image from point A from the inside to the outside into a circular area and several ring areas. The morphological processing was performed on each area and then assembled and restored to complete the morphological processing of the single tomographic image. This method can effectively reduce the impact of uneven exposure of a single tomographic image on the accuracy of the processing results during CT scanning.
[0066] 4. A similarity fitting process for adjacent tomographic images is proposed. This method establishes the interlayer connection between adjacent tomographic images along the acquisition direction, and can effectively remove the erroneous adhesion, erroneous holes and flaky aggregates between aggregates by interlayer comparison. The processing method can be divided into four basic categories according to the processing order and processing method: top-down intersection and expansion, top-down merge and erosion, bottom-up intersection and expansion, and bottom-up merge and erosion.
Claims
1. A pre-processing method for asphalt concrete fault image modeling suitable for aggregate adhesion removal, wherein all steps can be completed using Matlab loop program or related modeling software; characterized in that: The specific steps are as follows: The first step is tomographic image scanning: CT scanning of the asphalt concrete specimen, to scan the specimen in its initial state to obtain a tomographic image, with the acquisition direction set to the positive direction of the z-axis; The second step is high-pixel interpolation processing: the obtained tomographic image is subjected to high-pixel interpolation processing based on the interpolation method to obtain a high-pixel tomographic image whose length and width pixel values are more than twice the original length and width pixel values; The third step is grayscale processing and filtering and smoothing processing: grayscale processing is performed on the high-pixel tomographic image in the order of acquisition to obtain a high-pixel tomographic grayscale image; after the grayscale processing is completed, the high-pixel tomographic grayscale image is smoothed using the Gauss filtering algorithm to obtain a smoothed high-pixel tomographic grayscale image; The fourth step is morphological expansion and corrosion processing and binarization processing: Since the exposure of a single tomographic image is unevenly distributed in the shape of concentric rings, the single tomographic image is divided into a circle and several ring areas from the inside to the outside by annular segmentation according to the size of the image. If the outermost circle image is not in the shape of a ring, it is segmented in the same way as a ring until all valid parts of the layered image are segmented, and the morphological processing is performed on each area and then assembled and restored to complete the morphological processing of the single tomographic image; then the binarization processing of all tomographic images is completed in batches according to the acquisition order to obtain a high-pixel tomographic binary image; Step 5: pixel restoration processing: performing pixel restoration processing on the obtained high-pixel fault binary image to obtain a fault binary image; Step 6: Similarity fitting processing of adjacent tomographic images: Adopting the similarity fitting technology of adjacent tomographic images to construct the inter-layer connection along the acquisition direction of the obtained tomographic binary image, the steps are as follows; (1) Choose between cross expansion treatment and corrosion treatment according to actual needs Intersection expansion processing: add the pixel information values at the same position in a continuous fault binary images, and import the addition result into a blank new image of the same pixel size as the original fault image according to the pixel position. The label of the blank new image is the smaller value of the two adjacent fault images. The pixel information of the position where the addition result of the new image pixel information value is "0" remains unchanged and displays dark; the pixel information of the position where the addition result of the new image pixel information value is "1" is corrected to "0" and displays dark; the pixel information of the position where the addition result of the new image pixel information value is "≥2" is corrected to "1" and displays white and bright; after the correction is completed, the white and bright aggregate area of the new image is expanded, and the asphalt mortar holes inside the aggregate are deleted; Take and erode processing: add the pixel information values at the same position in b consecutive fault binary images, and import the addition result into a blank new image of the same pixel size as the original fault image according to the pixel position. The label of the blank new image is the smaller value of the two adjacent fault images. The pixel information at the position where the addition result of the new image pixel information value is "0" remains unchanged and is displayed dark; the pixel information at the position where the addition result of the new image pixel information value is "1" remains unchanged and is displayed white and bright; the pixel information at the position where the addition result of the new image pixel information value is "≥2" is corrected to "1" and displayed white and bright; after the correction is completed, the white and bright aggregate area of the new image is corroded, and the asphalt mortar holes inside the aggregate are deleted; (2) completing the similarity fitting processing of subsequent adjacent tomographic images in sequence, and generating new images in sequence, and the sequence set of the new images constitutes a new tomographic binary image atlas; Step 7: Volume rendering modeling: Based on the volume rendering technology, the new fault binary image atlas obtained in the sixth step is 3D reconstructed to obtain the asphalt concrete model; Step 8, regenerating the tomographic image of the model obtained in step 7 along any radial direction, and obtaining a tomographic binary image through binarization processing; Step 9, tomographic image regeneration processing: repeating steps 6 to 8 for the tomographic binary image atlas obtained by regeneration in step 8 to generate a new tomographic binary image set; In the tenth step, the volume rendering technology is used to reconstruct the asphalt concrete specimen model in three dimensions based on the new fault binary image set obtained in the ninth step.
2. The asphalt concrete fault image modeling pre-processing method suitable for aggregate adhesion removal according to claim 1 is characterized in that: The asphalt concrete test pieces include cuboids, cylinders and test pieces with other shapes and size characteristics.
3. The asphalt concrete fault image modeling pre-processing method suitable for aggregate adhesion removal according to claim 1 is characterized in that: The tomographic image types include RGB color images, grayscale images and binary images; the length and width of the tomographic image are required to completely include the tomographic image of the asphalt concrete specimen, and the excess part is filled with darkness; the tomographic grayscale image includes a uint8 grayscale image with a grayscale eigenvalue range of 0-255 and a uint16 grayscale image with a grayscale eigenvalue range of 0-65535.
4. The asphalt concrete fault image modeling pre-processing method suitable for aggregate adhesion removal according to claim 1 is characterized in that: The high-pixel interpolation process requires that the original pixel image be converted into a high-pixel image in equal proportion.
5. The asphalt concrete fault image modeling pre-processing method suitable for aggregate adhesion removal according to claim 1 is characterized in that: The morphological dilation and corrosion process uses a kernel function operator with a star shape.
6. The asphalt concrete fault image modeling pre-processing method suitable for aggregate adhesion removal according to claim 1 is characterized in that: The pixel restoration process requires that the high-pixel image be restored to the original pixel image in equal proportion.
7. The asphalt concrete fault image modeling pre-processing method suitable for aggregate adhesion removal according to claim 1 is characterized in that: In the similarity fitting process of adjacent tomographic images, the values of a and b are required to be ≥2 and ≤10. Four categories are proposed according to the order and processing method: top-down intersection expansion processing, top-down union corrosion processing, bottom-up intersection expansion processing and bottom-up union corrosion processing, which can be used in combination according to the processing purpose.
8. The asphalt concrete fault image modeling pre-processing method suitable for aggregate adhesion removal according to claim 1 is characterized in that: In the tomographic image regeneration step, any radial direction is required to be perpendicular to the initial scanning direction z in the three-dimensional space.
