Aggregate image crack segmentation method, device, storage medium and program product

Abstract

The present invention discloses a method, device, storage medium and program product for crack segmentation of aggregate images. The method comprises: filtering a cross-sectional grayscale image of the aggregate to reduce noise in the cross-sectional grayscale image to obtain a first grayscale image of the aggregate; performing image enhancement processing on the first grayscale image to highlight the crack region in the first grayscale image to obtain a second grayscale image of the aggregate; performing crack segmentation on the second grayscale image in a manner of edge region growth, wherein the seed point and seed growth range of the edge region growth are obtained in a human-computer interactive manner. The method has good flexibility, controllability and adaptability to different types of aggregates.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method, device, storage medium and program product for segmenting cracks in aggregate images. Background Art

[0002] This section is intended to provide a background or context to the embodiments recited in the claims. No admission is made that the description herein is prior art by inclusion in this section.

[0003] In the related art, most of the crack segmentation methods of rock aggregates are based on the threshold segmentation principle. The prerequisite for segmenting cracks using threshold difference is that the grayscale value in the crack area is significantly different from the grayscale value in the rest of the rock aggregate area.

[0004] For example, in the master's thesis "Research on Cement-stabilized Gravel Damage Based on CT Images" by Lu Yuejing of Wuhan University of Science and Technology, it is mentioned that the best processing method for N1-0500CT grayscale images is: Wiener filtering (9×9 filter) + iterative threshold segmentation; the best processing method for N4-0437CT grayscale images is: reducing brightness + contrast enhancement (0.8) + Wiener filtering (9×9 filter) + iterative threshold segmentation.

[0005] However, for meso-scale cracks, their crack width is small, their depth is shallow, and their grayscale value difference with other areas around the rock aggregate is small, making it difficult to extract them through simple threshold segmentation. Summary of the invention

[0006] The invention provides a method, a device, a storage medium and a program product for segmenting cracks in aggregate images.

[0007] The present invention adopts the following technical solution: a method for segmenting cracks in aggregate images, comprising:

[0008] Performing filtering processing on the cross-sectional grayscale image of the aggregate to reduce noise in the cross-sectional grayscale image to obtain a first grayscale image of the aggregate;

[0009] Performing image enhancement processing on the first grayscale image to highlight the crack region in the first grayscale image, thereby obtaining a second grayscale image of the aggregate;

[0010] The second grayscale image is subjected to crack segmentation in a manner of edge region growth, wherein a seed point and a seed growth range of the edge region growth are obtained in a manner of human-computer interaction.

[0011] Optionally, filtering is performed on the cross-sectional grayscale image of the aggregate, including:

[0012] The cross-sectional grayscale image is subjected to Gaussian filtering and median filtering in sequence.

[0013] Optionally, performing image enhancement processing on the first grayscale image includes:

[0014] Perform a gamma transform on the first grayscale image.

[0015] Optionally, in the step of performing crack segmentation on the second grayscale image in a manner of edge area growing, boundary extraction is performed on the volume data of the aggregate to obtain the boundary of the aggregate in the cross-sectional grayscale image, wherein the cross-sectional grayscale image belongs to the volume data, and the segmented crack area is set to be located inside the area defined by the boundary of the aggregate.

[0016] Optionally, in the step of performing crack segmentation on the second grayscale image in a manner of edge region growing, connected pores are extracted from the volume data of the aggregate to obtain a connected pore region of the aggregate in the cross-sectional grayscale image, wherein the cross-sectional grayscale image belongs to the volume data, and the segmented crack region is set to have no intersection with the connected pore region.

[0017] Optionally, the aggregate is rock aggregate.

[0018] The present invention adopts the following technical solution: a device for segmenting cracks in aggregate images, comprising: a memory and a processor, the memory storing instructions, and the processor running the instructions to execute the aforementioned method.

[0019] The present invention adopts the following technical solution: a device for segmenting cracks in aggregate images, comprising:

[0020] A filtering module, used for filtering the cross-sectional grayscale image of the aggregate to reduce noise in the cross-sectional grayscale image and obtain a first grayscale image of the aggregate;

[0021] an enhancement module, configured to perform image enhancement processing on the first grayscale image to highlight the crack region in the first grayscale image, and obtain a second grayscale image of the aggregate;

[0022] The segmentation module is used to perform crack segmentation on the second grayscale image in a way of edge region growth, wherein the seed point and seed growth range of the edge region growth are obtained in a way of human-computer interaction.

[0023] The present invention adopts the following technical solution: a storage medium stores instructions thereon, and the instructions execute the aforementioned method when executed.

[0024] The present invention adopts the following technical solution: a computer program product, which executes the above method when running.

[0025] Since the seed point and seed growth range in the edge region growing algorithm can be set manually by the operator based on experience, the edge region growing process is more flexible, controllable, and adaptable to different types of aggregates. Furthermore, for the specific application scenario of aggregate crack segmentation, the edge region growing algorithm has a higher quality of crack segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of the aggregate image crack segmentation method according to an embodiment of the present invention.

[0027] Figure 2 It is a schematic structural diagram of an aggregate image crack segmentation device according to an embodiment of the present invention.

[0028] Figure 3 It is a schematic diagram of the structure of the aggregate image crack segmentation device of other embodiments of the present invention.

[0029] Figure 4a It is a grayscale image of the cross section of the aggregate according to the embodiment of the present invention.

[0030] Figure 4b to Figure 4f Yes Figure 4a The first grayscale image is obtained by performing different filtering processes on the end face grayscale image shown, wherein: Figure 4b The mean filter method is used in Figure 4c The box filter method is used in Figure 4d The median filter method is used in Figure 4e The Gaussian filtering method is used in Figure 4f The method of first Gaussian filtering and then median filtering is adopted.

[0031] Figure 5a is the first grayscale image of the embodiment of the present invention.

[0032] Figure 5b to Figure 5d The image enhancement effects of processing the first grayscale image using the histogram mean method, logarithmic transformation method and gamma transformation method are shown in sequence.

[0033] Figures 6a to 6c The crack segmentation results are obtained by using the Otsu method, the interactive threshold segmentation method and the aggregate image crack segmentation method of the embodiment of the present invention in turn. DETAILED DESCRIPTION

[0034] The present invention is further described below in conjunction with specific embodiments, but the protection scope of the present invention is not limited thereto.

[0035] Figure 1 It is a flow chart of the aggregate image crack segmentation method according to an embodiment of the present invention.

[0036] Figure 2It is a schematic structural diagram of an aggregate image crack segmentation device according to an embodiment of the present invention. Figure 3 It is a schematic diagram of the structure of the aggregate image crack segmentation device of other embodiments of the present invention. Figure 4a It is a grayscale image of the cross section of the aggregate according to the embodiment of the present invention. Figure 4b to Figure 4f Yes Figure 4a The first grayscale image is obtained by performing different filtering processes on the end face grayscale image shown, wherein: Figure 4b The mean filter method is used in Figure 4c The box filter method is used in Figure 4d The median filter method is used in Figure 4e The Gaussian filtering method is used in Figure 4f The method of first Gaussian filtering and then median filtering is adopted. Figure 5a is the first grayscale image of the embodiment of the present invention. Figure 5b to Figure 5d The image enhancement effects of processing the first grayscale image using the histogram mean method, logarithmic transformation method and gamma transformation method are shown in sequence. Figures 6a to 6c The crack segmentation results are obtained by using the Otsu method, the interactive threshold segmentation method and the aggregate image crack segmentation method of the embodiment of the present invention in turn.

[0037] refer to Figure 1 , the method for aggregate image crack segmentation provided by the embodiment of the present invention includes:

[0038] Step 101, filtering the cross-sectional grayscale image of the aggregate to reduce noise in the cross-sectional grayscale image to obtain a first grayscale image of the aggregate;

[0039] Step 102: performing image enhancement processing on the first grayscale image to highlight the crack region in the first grayscale image, thereby obtaining a second grayscale image of the aggregate;

[0040] Step 103 : performing crack segmentation on the second grayscale image in a manner of edge region growth, wherein a seed point and a seed growth range of the edge region growth are obtained in a human-computer interaction manner.

[0041] From the perspective of equipment, the execution subject of the above steps can be one or more electronic devices, more specifically, the processing modules therein; from the perspective of program, the execution subject of the above steps can correspondingly be the programs installed on these electronic devices.

[0042] Since the seed point and seed growth range in the edge region growing algorithm can be set manually by the operator based on experience, the edge region growing process is more flexible, controllable, and adaptable to different types of aggregates. Furthermore, for the specific application scenario of aggregate crack segmentation, the edge region growing algorithm has a higher quality of crack segmentation.

[0043] Each step is described in detail below with reference to the accompanying drawings.

[0044] Obtaining grayscale images of aggregate cross sections

[0045] In one embodiment of the present invention, a grayscale image of the surface morphology of the asphalt pavement aggregate during the wear process is obtained by CT scanning, and grayscale image filtering and image enhancement are performed to provide a basis for the subsequent extraction of microcracks on the surface of the road aggregate. Taking limestone pavement as an example, the surface aggregate of the wheel track of the road test piece is taken. CT scanning can obtain cross-sectional images (also called slice images) of the aggregate sample under three views respectively. The method of obtaining the cross-sectional image of the aggregate is not limited to CT scanning.

[0046] The raw image data obtained by CT scanning is usually saved in DICOM format, and the raw data needs to be converted into RAW format data. RAW format data is a raw data format that has not been compressed or encoded, which can retain the integrity and accuracy of the data.

[0047] Filtering

[0048] In the grayscale image of cracks in road aggregates, isolated noise points will interfere with the segmentation and recognition of crack targets. At the same time, unlike the commonly used cracks in asphalt pavement and asphalt mixture, the size of cracks on the surface of road aggregates is often only 2.3mm, which is more easily interfered by noise points, pores and mineral particles on the surface of aggregates in crack segmentation. Therefore, the quality of image denoising is very important for subsequent crack identification and segmentation. In the extraction of cracks on the surface of road aggregates, selecting a suitable filtering method to remove noise in the aggregate grayscale image is a critical step.

[0049] Currently, the commonly used methods for image denoising include mean filtering, median filtering, Gaussian filtering and box filtering.

[0050] The mean filter method, also known as the neighborhood average method, is a relatively simple smoothing method used to remove noise from images. The basic idea of ​​the mean filter is to average the pixel values ​​within a certain range around each pixel point, so that the noise is suppressed and the image is smoothed. The noise reduction effect of the mean filter mainly depends on the size of the convolution kernel during the filtering process. The larger the convolution kernel, the more obvious the smoothing effect of the image. The mean filter algorithm process can be expressed as follows:

[0051]

[0052] Where: I new (x, y) represents the gray value after median filtering; I (x+I, y+j) represents the gray value of the pixel before processing; k represents the number of pixels in the filter.

[0053] As can be seen from formula 1, the principle of mean filtering to smooth the image is that after the gray value of the pixel is averaged, the difference between the gray value of the pixel and the surrounding becomes smaller, thereby achieving the purpose of removing noise. In the mean filtering process, it is assumed that the weight of each pixel in the filter window is the same, and the influence of the position relationship of the pixel is ignored. Therefore, for small defects such as microcracks on the surface of aggregates, mean filtering will blur the crack edge and lose information.

[0054] The principle of the box filter method is similar to that of the mean filter method. Both are linear smoothing filters. The difference is that the grayscale value weight distribution of the pixel points in the box filter method is rectangular. Therefore, compared with the mean filter method, the box filter method will amplify the information of the central pixel point of the rectangle, and can maintain the edge information and details of the selected center point to a certain extent.

[0055] Median filtering is a common nonlinear smoothing filtering algorithm. Its basic idea is to sort the grayscale values ​​of pixels in the neighborhood around a specific pixel, and then take the middle value to cover the grayscale value of the pixel to achieve the denoising effect. The denoising effect of median filtering mainly depends on the size of the selected filter window. The larger the window, the more obvious the smoothing effect, but too large a window will cause the loss of image details. In addition, in order to ensure that the grayscale value of the pixel taken after sorting is an integer, the filter window size is usually set to an odd value, such as 3×3, 5×5, etc. The median filtering algorithm can be summarized as Equation 2:

[0056] I new (x,y)=med{I(x,y),...,I(x+k-1,y),I(x,y+1),...,I(x,y+k-1)} (2)

[0057] Where: I new (x, y) represents the gray value after median filtering; I(x, y) represents the gray value of the pixel before processing; k represents the filter window size; med represents the median function.

[0058] From Formula 2, we can see that median filtering can reduce the difference between the grayscale value of a pixel and the grayscale values ​​of other pixels in the neighborhood, thereby effectively removing salt and pepper noise and speckle noise. However, for linearly distributed noise such as Gaussian noise, the noise reduction effect of median filtering is still limited.

[0059] Gaussian filtering is a commonly used linear smoothing filtering algorithm. Its basic idea is to use a weight function that obeys the Gaussian distribution to perform weighted averaging on the grayscale values ​​of pixels in the image, thereby making the grayscale value distribution smoother. The pixel weights in the Gaussian filter are shown in Equation 3

[0060]

[0061] Where: G i,jrepresents the weight of the pixel in the i-th row and j-th column in the Gaussian filter; k represents the radius of the filter; σ represents the standard deviation of the Gaussian function.

[0062] For each pixel in the image, the Gaussian filter performs a weighted average of the grayscale values ​​of the pixel and the surrounding pixels to obtain a new grayscale value. The new grayscale value of the pixel at (x, y) can be expressed as:

[0063]

[0064] Where: I new (x, y) represents the gray value after median filtering; I (x+i, y+j) represents the gray value of the pixel before processing; k represents the filter radius.

[0065] Gaussian filtering can remove most noises that obey linear distribution, such as Gaussian noise, hole noise, etc. However, it is less effective in removing isolated noise points, and Gaussian filtering will reduce the edge sharpness and details of the image.

[0066] Different filtering methods have their own advantages and disadvantages, and their applicable conditions are also different. For the aggregate microstructure, the imaging is affected by many factors, so there are often multiple noises in the grayscale image. Therefore, if a single filtering method is used, it is impossible to achieve a good noise reduction effect. The above filtering methods must be used in an organic combination. The different filtering effects are shown in Figure b to Figure 4f shown.

[0067] refer to Figure 4b Although the image is relatively smooth after mean filtering, the edges of the surface cracks are blurred, and the fine cracks on the surface are covered during the filtering process, so the filtering effect is not ideal.

[0068] refer to Figure 4c ,The box filter retains some crack edge information, but there are still many isolated white noise points in the image, and the filtering effect is not ideal.

[0069] refer to Figure 4d ,The median filter obviously eliminates most of the isolated white noise points, but some surface crack details are lost in the filtering process.

[0070] refer to Figure 4e , Gaussian filtering better preserves the surface crack details, but has a poor effect on eliminating isolated white noise points.

[0071] refer to Figure 4f ,After first Gaussian filtering and then median filtering, the isolated noise points in the image are ,effectively removed, and the edge information of the crack on the aggregate surface is ,retained relatively completely, highlighting the crack information compared with the ,original image.

[0072] The present invention refers to the filtered image as a first grayscale image.

[0073] During the CT scanning process, noise points and ray threads will inevitably exist on the grayscale images of the cross section, which will affect the analysis of image information. Image noise and ray threads can be eliminated to a certain extent by filtering, but image filtering alone cannot completely eliminate the interference in the image. Therefore, after image filtering, the first grayscale image must be enhanced. Image enhancement means making the image easier to observe, analyze and process visually while keeping the basic information of the image unchanged. Its fundamental purpose is to highlight the part to be analyzed in the grayscale image and blur or simplify the rest of the surrounding parts.

[0074] Note: Ray thread is caused by the uneven radiation of the X-ray tube and the uneven response of the detector during CT scanning. It usually appears as local periodic brightness changes in grayscale images, similar to water waves. Ray thread is a common noise in CT tomography.

[0075] In a preferred embodiment, the cross-sectional grayscale image is subjected to multiple Gaussian filtering and then multiple median filtering to improve the filtering effect.

[0076] Multiple Gaussian filters are cascaded together to enhance the filtering effect. The kernel function size of the previous Gaussian filter is smaller than the kernel function size of the subsequent Gaussian filter.

[0077] For example, the kernel function size of the first-layer Gaussian filter is 3*3, and the standard deviation is 1.5. The kernel function size of the second-layer Gaussian filter is 5*5, and the standard deviation is 2.5.

[0078] In stacked Gaussian filtering, the size and weight of each layer of filters can be set according to specific needs. The first layer of filters uses a smaller kernel function to remove some small noise and details in the image without causing much impact on the entire image. The second layer of filters uses a larger kernel function to smooth the entire image and further remove some noise.

[0079] If the order of the filter sizes is reversed, the image will be smoother after processing, but the crack details required in this case will be difficult to retain.

[0080] When the size of the median filter remains unchanged, the median filter is repeated preferably 2 to 4 times, more preferably 3 times.

[0081] This number is an empirical value. Using median filtering too many times will reduce the local crack details of the image; using median filtering too few times cannot effectively remove isolated noise points. In this case, after comparing the actual filtering effects, it is proposed that the optimal number is 3 times.

[0082] Image Enhancement

[0083] Histogram equalization is a method used for image enhancement, which can make the grayscale value distribution of the image more uniform, thereby enhancing the contrast and brightness of the image. Its basic idea is to transform the pixel values ​​of the image so that the grayscale value of the image is distributed in a wider range, while increasing the visual effect and information content of the image.

[0084] Histogram equalization is a more practical image enhancement method in grayscale image processing, which has the advantages of being fast and easy to operate. However, the essence of the histogram equalization algorithm is to redistribute the global grayscale values ​​of the image, which may cause image noise enhancement and may destroy the local features of the image.

[0085] Grayscale transformation refers to the linear or nonlinear transformation of the grayscale values ​​of pixels in a digital image, thereby changing the image's visual characteristics such as brightness, contrast, and hue. For the grayscale image of road aggregate analyzed by the present invention, the surface cracks are similar to the grayscale values ​​of the surrounding aggregates. If the conventional linear grayscale transformation method is used, it is not possible to significantly improve the contrast between the cracks and the aggregate minerals. Therefore, in order to improve the subsequent crack extraction quality, an appropriate nonlinear grayscale transformation method should be selected. Commonly used nonlinear grayscale transformation methods include logarithmic transformation, gamma transformation, etc.

[0086] Logarithmic transformation is a commonly used nonlinear grayscale transformation method in digital image processing. Logarithmic transformation can compress the high grayscale part of the image and enhance the part with lower grayscale value in the image. The algorithm principle is shown in Formula 5.

[0087] s = c × log x (1+r) (5)

[0088] Where: s represents the gray value of the pixel after logarithmic transformation; r represents the gray value of the pixel before logarithmic transformation; c is the transformation constant; x represents the degree of logarithmic transformation. The larger the x, the more obvious the low gray enhancement effect.

[0089] Logarithmic transformation can perform nonlinear mapping of grayscale values ​​in an image, improve image contrast and clarity, and make the image more consistent with human perception. However, when logarithmic transformation is used to adjust the contrast of grayscale images, it is easy to over-enhance the image, resulting in image distortion. At the same time, when there are isolated white noise points in the image, the white noise points will be more obvious after logarithmic transformation.

[0090] Gamma transform is used to adjust the brightness and contrast of an image. It redistributes the grayscale range of the image by performing nonlinear mapping on the grayscale value of the image, thereby achieving the purpose of adjusting the brightness and contrast of the image. Its mathematical expression is shown in Equation 6:

[0091] s=c×r γ (6)

[0092] Where: s represents the output grayscale value; r represents the input grayscale value, c is a constant, and γ is the gamma value. The larger γ is, the greater the difference between the output pixel value and the input pixel value is, and the stronger the enhancement effect is. In the gamma transform, the value of γ determines the degree of change in the brightness and contrast of the image. When γ is less than 1, the gamma transform can enhance the part of the image with lower grayscale values, making the image brighter; when γ is greater than 1, the gamma transform can enhance the part of the image with higher grayscale values, making the image dimmer. What needs to be enhanced in the present invention is the crack area with lower grayscale values.

[0093] Gamma transform can adjust different grayscale areas to different degrees, so as to better preserve the details in the image. The parameters of the transformation are relatively easy to control and can be adjusted according to different image processing requirements. However, gamma transform is sensitive to noise and artifacts in the image. When there are many noise points or ring artifacts in the enhanced grayscale image, gamma transform will reduce the image quality.

[0094] and Figure 5a Compared with the first grayscale image shown in Figure 1, the three image enhancement methods all highlight the crack part to a certain extent. Figure 5b As shown in the figure, the image noise is significantly enhanced after histogram equalization, which is not conducive to the subsequent extraction of aggregate surface cracks. Therefore, the histogram equalization enhancement effect is not ideal. For the two nonlinear grayscale transformation methods, the surface noise and artifacts are not significantly enhanced after image enhancement. Figure 5c As shown in the figure, after logarithmic transformation, there are still some obvious white noise points in the image, and the low grayscale parts are enhanced indiscriminately, resulting in the grayscale value difference between the surface cracks and the surrounding aggregate minerals is still not obvious, which is not conducive to the subsequent surface crack extraction. In contrast, after gamma transformation, the image clearly highlights the aggregate crack part, and the details related to the surface cracks in the image are relatively complete, as shown in the figure. Figure 5d Therefore, the present invention selects gamma transform to perform grayscale image enhancement to improve the segmentation accuracy and reliability of microcracks on the aggregate surface.

[0095] Crack segmentation

[0096] The threshold-based method is a common method for aggregate surface crack segmentation, among which the Otsu method is one of the commonly used algorithms based on global threshold. The Otsu method is an adaptive method that determines the optimal threshold by calculating the variance of the grayscale of the image, so that the difference between the crack area and the background area is maximized under this threshold. This method has the advantages of simplicity and speed, but it is sensitive to uneven illumination and noise interference of the image, which may lead to unsatisfactory segmentation results.

[0097] The Otsu method (OTSU), also known as the maximum inter-class variance method, is the most representative image segmentation method among the threshold segmentation methods, which can adaptively select the threshold range. The essential idea of ​​the algorithm is to regard the part to be extracted as the target and the rest of the picture as the background. Based on the difference in grayscale values ​​between the target and background areas, the appropriate threshold boundary is found through repeated trial and error, that is, the optimal threshold in the Otsu method. Under this threshold, the inter-class variance of the selected target and background is maximized, while the intra-class variance is minimized. By setting the optimal threshold, the segmentation of the desired target can be completed. The Otsu method can be applied to various types of image segmentation tasks, especially when the grayscale difference between the background and foreground is relatively large, it has a good segmentation effect. However, when the grayscale difference between the background and foreground is not obvious, the segmentation accuracy of the Otsu method will be reduced.

[0098] refer to Figure 6a After segmentation by the Otsu method, the surface crack information of the image is relatively complete, but due to the small size of the crack information, the gray value difference between some cracks and the surrounding pores and minerals is very small. Therefore, it is difficult to directly traverse the optimal threshold in the Otsu method, and the surface crack extraction effect is not ideal.

[0099] The interactive threshold segmentation method is similar to the Otsu method in that both methods are based on the grayscale value difference between the image target and the background for segmentation. However, the difference is that the interactive threshold segmentation method can manually select the optimal threshold for segmentation through human-computer interaction, thereby improving the segmentation accuracy. Compared with the Otsu method, the interactive threshold segmentation method can more accurately extract crack information with small grayscale value differences, especially small cracks.

[0100] refer to Figure 6b ,Although the interactive threshold segmentation method effectively improves the ,segmentation accuracy of micro-cracks on the aggregate surface, there is still the ,problem of over-segmentation for elliptical or flat pores.

[0101] Region growing is an aggregation algorithm based on pixel or sub-region attributes. Its basic idea is to gradually merge adjacent pixels or sub-regions into larger regions starting from certain starting points of pixels or sub-regions according to a set of prior rules. The basic process is to first select some seed points as starting points based on the prior characteristics of the cracks, that is, the depth of the cracks, and then find pixels or sub-regions that meet the conditions in the pixels or sub-regions adjacent to the seed points for growth until all pixels or sub-regions that meet the conditions are merged into a connected crack region. This method can not only extract all cracks of varying depths, but also retain the morphological information of the cracks, thereby improving the accuracy of crack extraction.

[0102] In practical applications, the key to the region growing algorithm lies in how to select a suitable starting point and a suitable termination condition. For the selection of the starting point, it can generally be selected according to the prior characteristics of the crack, such as the depth, or according to the grayscale value of the pixel. For the selection of the termination condition, a common method is to set a threshold and stop the growth when the threshold is reached, or to set a specific growth range and stop the growth when the growth reaches this range. In addition, in order to avoid breakage or overlap during the growth process, some restrictions are generally required, such as pixel connectivity and grayscale value differences. Based on the principle of edge region growth, the present invention develops a human-computer interactive aggregate surface microcrack extraction program, and the specific steps are as follows:

[0103] (1) Importing the CT scan RAW data of road aggregate into the VG-Studio platform, the cross-sectional images of the aggregate in three directions and its 3D rendering model can be obtained. In the subsequent crack segmentation process, the divided cracks can be observed in real time in the 3D model.

[0104] Select the "Surface Determination" module in the "Analysis" menu bar. This module can extract the information of each component in the multiphase material in two-dimensional or three-dimensional space based on the type of scan data, and create the multiphase material boundary. For the CT scanned material analyzed by the present invention, the surface determination method is selected as the equivalue partitioning algorithm based on volume data ("Volume Material Boundaries"), which can count the grayscale value of each pixel in the volume data and output the grayscale distribution histogram. Then, the aggregate boundary is divided by setting the grayscale interval between the "material" and the "background". After the boundary division, the segmented crack area is set to be located inside the area defined by the boundary of the aggregate. This can effectively avoid extracting the material boundary information together when extracting cracks, thereby improving the accuracy of crack segmentation.

[0105] There are often connected pores and high-density inclusions in aggregates. Among them, high-density inclusions have a high grayscale value and usually exist in the form of white spots in the grayscale image. They have little effect on crack segmentation and can be ignored; while pores are close to the crack threshold and are very likely to cause interference in the segmentation process. Therefore, pore analysis must be performed on aggregates before segmenting cracks. Create a "pore / inclusion analysis" calculation module in the volume data, and use the program's built-in measuring tools to measure the pore threshold range and size.

[0106] The grayscale value of the area where the pores are located in the second grayscale image may be set to a higher value, for example, the grayscale is set to 255 (corresponding to white).

[0107] The pore threshold range is the grayscale variation range between the pore center and the pore edge in the grayscale image.

[0108] When performing crack segmentation in volume data, it is necessary to select a region of interest (Region of Interest) in order to accurately perform crack segmentation. The region of interest is the seed growth range, and the seed growth will not exceed the region of interest.

[0109] In the VG-Studio platform, crack segmentation can be performed by creating a new ROI and manually selecting the starting point of the crack. The specific operations include: first, create a new ROI in the imported image data, select the Region of Interest Tool in the toolbar, and then manually select the starting point of the crack in the image (that is, the seed point), and define the shape and size of the ROI (that is, the seed growth range). Next, select the Region Growing Tool and set relevant parameters, such as the direction of growth, the rate of growth, etc., to automatically segment the crack. According to actual needs, the same second grayscale image can be segmented multiple times in sequence to improve the accuracy of segmentation. The advantage of using a region of interest is that it can reduce the segmentation of the entire image, thereby improving the efficiency and accuracy of segmentation, while also avoiding interference with non-crack areas.

[0110] The crack segmentation effect is as follows Figure 6c As shown in the figure, under the edge region growth algorithm, by repeatedly adding grayscale seed growth positions and adjusting the growth range, the aggregate microcracks can be segmented more completely.

[0111] By comparing the three segmentation methods, it can be found that although the Otsu method can initially segment the crack morphology, it is difficult to remove the boundary effect of the aggregate grayscale image. At the same time, the crack segmentation accuracy is easily affected by noise, pores and high-density inclusions. It is difficult to automatically traverse the algorithm to find the appropriate optimal threshold, and the segmentation effect of microcracks on the surface of road aggregate is poor. The interactive threshold segmentation rule can completely remove the boundary effect of the aggregate grayscale image. At the same time, the shape edge sharpness factor is introduced as the segmentation standard to avoid the interference of circular holes and isolated noise points, but it still cannot avoid the interference of flat pores. Therefore, the interactive threshold segmentation method is prone to over-segmentation. For the edge region growth segmentation method, since the grayscale seed growth range and growth mode of the local area can be repeatedly set, it not only effectively avoids the interference of various types of pores and isolated noise points, but also greatly improves the segmentation accuracy of microcracks on the aggregate surface. The human experience judgment is perfectly combined with the automatic operation of the machine. For microcracks on the aggregate surface at a small scale, the edge region growth segmentation method is a relatively excellent segmentation method.

[0112] In addition, the aggregate surface microcrack segmentation method based on the edge area growth principle proposed in this chapter also has the following advantages:

[0113] (1) By using the principle of edge region growing algorithm, compared with the traditional threshold segmentation method, the ability to capture crack detail information can be improved while maintaining high segmentation accuracy, effectively improving the accuracy and precision of crack segmentation. It still has a high segmentation accuracy for cracks with little change in gray value and relatively small.

[0114] (2) Compared with the traditional automatic segmentation method that relies solely on algorithms, human-computer interactive segmentation is a more flexible and controllable segmentation method. During the crack segmentation process, the operator can guide the crack segmentation process through subjective judgment and control, which can effectively avoid the misjudgment and missed judgment in the automatic segmentation algorithm. During the crack segmentation process, the operator can judge and adjust the segmentation process based on his own experience and professional knowledge, such as adjusting the starting point of the segmentation, the direction of the segmentation, the segmentation rate and other parameters, thereby improving the reliability and accuracy of the segmentation. Compared with the automatic segmentation algorithm, the human-computer interactive method has better flexibility and controllability, can better adapt to different materials, different segmentation requirements and segmentation difficulties, and improves the efficiency and quality of segmentation. At the same time, the human-computer interactive crack segmentation method also provides a more reliable and accurate data basis for subsequent crack morphology and statistical analysis.

[0115] (3) Compared with the traditional threshold segmentation method, the edge region growth segmentation method has better adaptability and flexibility in crack segmentation. The traditional threshold segmentation method requires manual setting of the threshold, and different thresholds may be required for different samples and crack types, so there is a certain degree of subjectivity and limitations. The edge region growth segmentation method can optimize the edge region growth algorithm through custom parameters to adapt to different crack segmentation requirements and improve the applicability and flexibility of the algorithm. At the same time, the edge region growth segmentation method can better handle problems such as complex material structures and noise interference, and has a wider range of applicability and higher segmentation accuracy.

[0116] Based on the same inventive concept, continue to refer to Figure 2 An embodiment of the present invention further provides a device for segmenting cracks in aggregate images, comprising: a memory and a processor, wherein the memory stores instructions, and the processor runs the instructions to execute the aforementioned method.

[0117] Based on the same inventive concept, continue to refer to Figure 3 , an embodiment of the present invention further provides a device for segmenting cracks in aggregate images, comprising:

[0118] A filtering module 1 is used to filter the cross-sectional grayscale image of the aggregate to reduce noise in the cross-sectional grayscale image and obtain a first grayscale image of the aggregate;

[0119] Enhancement module 2, used for performing image enhancement processing on the first grayscale image to highlight the crack area in the first grayscale image to obtain a second grayscale image of the aggregate;

[0120] The segmentation module 3 is used to perform crack segmentation on the second grayscale image in a way of edge region growth, wherein the seed point and seed growth range of the edge region growth are obtained in a way of human-computer interaction.

[0121] Specifically, the device may be implemented in the form of hardware such as an application specific integrated circuit (ASIC), a micro control unit (MCU), or in the form of software, or in the form of a combination of software and hardware.

[0122] Based on the same inventive concept, an embodiment of the present invention further provides a storage medium storing instructions, which execute the aforementioned method when executed.

[0123] Based on the same inventive concept, an embodiment of the present invention further provides a computer program product, which executes the aforementioned method when running.

[0124] Each embodiment of the present invention is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, computer-readable storage medium, and computer program product embodiments, since they are basically similar to the method embodiments, their descriptions are simplified, and the relevant parts can be referred to the partial description of the method embodiments.

[0125] The device, computer-readable storage medium, and computer program product provided by the embodiments of the present invention correspond to the method one by one, therefore, the device, computer-readable storage medium, and computer program product also have beneficial technical effects similar to the corresponding method, and since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device, computer-readable storage medium, and computer program product are not repeated here. It should be understood by those skilled in the art that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0126] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatuses, and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0127] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0128] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0129] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory. In some configurations, the device that performs operations may also be a graphics processing unit (GPU).

[0130] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0131] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. In addition, although the operations of the method of the present invention are described in a particular order in the accompanying drawings, this does not require or imply that these operations must be performed in this particular order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step, and / or one step can be decomposed into multiple steps.

[0132] The protection scope of the present invention is not limited to the above-mentioned embodiments. Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the scope and spirit of the present invention. If these changes and modifications fall within the scope of the claims of the present invention and their equivalents, the present invention is intended to include these changes and modifications.

Claims

1. A method for segmenting cracks in aggregate images, characterized in that: include: Performing filtering processing on the cross-sectional grayscale image of the aggregate to reduce noise in the cross-sectional grayscale image, and obtaining a first grayscale image of the aggregate, wherein the cross-sectional grayscale image is subjected to Gaussian filtering and median filtering in sequence, and the aggregate is rock aggregate; Performing image enhancement processing on the first grayscale image to highlight the crack region in the first grayscale image, thereby obtaining a second grayscale image of the aggregate, wherein the first grayscale image is subjected to a gamma transformation; The second grayscale image is segmented for cracks in an edge region growth manner, wherein the seed point and seed growth range of the edge region growth are obtained in a human-computer interaction manner, connected pores are extracted from the volume data of the aggregate to obtain a connected pore region of the aggregate in the cross-sectional grayscale image, wherein the cross-sectional grayscale image belongs to the volume data, and the segmented crack region is set to have no intersection with the connected pore region, and boundary extraction is performed on the volume data of the aggregate to obtain the boundary of the aggregate in the cross-sectional grayscale image, wherein the cross-sectional grayscale image belongs to the volume data, and the segmented crack region is set to be located inside the region defined by the boundary of the aggregate.

2. A device for segmenting cracks in aggregate images, characterized in that: include: A memory and a processor, wherein the memory stores instructions and the processor executes the instructions to perform the method according to claim 1.

3. A device for segmenting cracks in aggregate images, characterized in that: include: A filtering module, used for filtering the cross-sectional grayscale image of the aggregate to reduce noise in the cross-sectional grayscale image and obtain a first grayscale image of the aggregate, wherein the cross-sectional grayscale image is subjected to Gaussian filtering and median filtering in sequence, and the aggregate is rock aggregate; an enhancement module, configured to perform image enhancement processing on the first grayscale image to highlight the crack region in the first grayscale image, and obtain a second grayscale image of the aggregate, wherein the first grayscale image is subjected to a gamma transformation; A segmentation module is used to perform crack segmentation on the second grayscale image in an edge region growth manner, wherein the seed point and seed growth range of the edge region growth are obtained in a human-computer interactive manner, and connected pores are extracted from the volume data of the aggregate to obtain a connected pore region of the aggregate in the cross-sectional grayscale image, wherein the cross-sectional grayscale image belongs to the volume data, and the segmented crack region is set to have no intersection with the connected pore region; boundary extraction is performed on the volume data of the aggregate to obtain the boundary of the aggregate in the cross-sectional grayscale image, wherein the cross-sectional grayscale image belongs to the volume data, and the segmented crack region is set to be located inside the region defined by the boundary of the aggregate.

4. A storage medium, characterized in that: Instructions are stored thereon, and when the instructions are executed, the method according to claim 1 is performed.

5. A computer program product, characterized in that When running, it executes the method according to claim 1.

Images