Aggregate extraction device and aggregate extraction method
The aggregate extraction device and method improve the accuracy of identifying aggregate shape and distribution in concrete structures by employing entropy filtering, k-means clustering, and dynamic contour expansion, overcoming brightness variations and noise in X-ray CT images.
Patent Information
- Application Number
- JP2023074294
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-04-28
- Publication Date
- 2024-11-08
AI Technical Summary
Existing methods using X-ray CT for visualizing concrete structures struggle to accurately separate and specify the shape and distribution of aggregates due to variations in brightness values and noise, leading to incomplete separation of aggregate regions.
An aggregate extraction device and method utilizing X-ray CT images, employing entropy filter processing, k-means clustering, morphological conversion, region release, and dynamic contour expansion to enhance the accuracy of aggregate identification.
Enables precise determination of aggregate shape and distribution within concrete structures, effectively distinguishing aggregates from other materials.
Smart Images

Figure 2024158776000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to an aggregate extraction device and an aggregate extraction method. [Background technology]
[0002] It has been reported that the compressive strength and fracture behavior of concrete are greatly affected by the shape and dimensions of the aggregates present inside the concrete, and the weak zones (transition bodies) formed between the aggregates and the cement paste. Understanding the distribution and shape of aggregates is important for predicting the fracture behavior of concrete structures, so there are hopes for visualization of the inside of concrete structures.
[0003] X-ray CT (Computed Tomography) is known as a method for non-destructively visualizing the inside of an object. X-ray CT not only makes it possible to visualize the inside of an object non-destructively, but also enables quantitative analysis using numerical data, the CT value (brightness value). Using X-ray CT, it is possible to obtain two-dimensional cross-sectional images of an object.
[0004] The following Patent Document 1 describes a non-destructive inspection method in which a test specimen is irradiated with neutron rays, a transmitted image made visible as the neutron rays pass through the test specimen is captured, and the moisture content of the test specimen is evaluated from the transmitted image. The following Patent Document 2 describes an evaluation method in which an X-ray CT method is applied to a cementitious hardened body to obtain an image showing the density distribution inside the cementitious hardened body, and the image is analyzed to evaluate a portion with a relatively high density as a portion with a high possibility of efflorescence generation. [Prior art documents] [Patent documents]
[0005] [Patent Document 1] JP 2019-174255 A [Patent Document 2] JP 2021-165722 A Summary of the Invention [Problem to be solved by the invention]
[0006] In cross-sectional images of concrete structures obtained by X-ray CT, the image regions in which aggregates appear have a relatively higher brightness than other image regions. Therefore, it is conceivable to set a threshold value for the cross-sectional images and use that value as a boundary to binarize and separate the aggregate regions. Then, by sequentially overlapping the cross-sectional images that represent the aggregate regions, the aggregate regions can be observed three-dimensionally.
[0007] However, due to variations in brightness in the cement region and noise called artifacts, simply binarizing the cross-sectional image may not be enough to separate the aggregate region from other regions, resulting in the problem that the shape and distribution of the aggregate cannot be accurately identified.
[0008] The present invention has been made in consideration of the above-mentioned circumstances, and has an object to provide an aggregate extraction device and an aggregate extraction method that are capable of accurately specifying the shape and distribution of aggregates. [Means for solving the problem]
[0009] In order to achieve the above object, the present invention employs, as a first solution relating to an aggregate extraction device, a means including an acquisition unit that acquires an X-ray CT image of a mortar / concrete test specimen containing aggregate and cement, and an image processing unit that generates an aggregate extraction image by performing clustering processing based on the k-means method, morphological transformation processing, region release processing, and active contour extension processing on the X-ray CT image.
[0010] In the present invention, as a second solution related to the bone material extraction device, in the above-mentioned first solution, the image processing unit applies entropy filter processing to the X-ray CT image as preprocessing of the clustering processing.
[0011] The present invention employs, as a third solving means related to the aggregate extraction device, a means in which, in the above-mentioned first or second solving means, the morphological transformation processing is an opening processing and a closing processing.
[0012] In the present invention, as a first solution relating to an aggregate extraction method, a means is adopted which includes a first step of performing a clustering process based on the k-means method on an X-ray CT image of a mortar / concrete test specimen containing aggregate and cement, a second step of performing a morphological transformation process on a result of the first step, a third step of performing a region release process on the result of the second step, and a fourth step of performing an active contour extension process on the result of the third step.
[0013] The present invention employs, as a second solving means related to the aggregate extraction method, a means for performing an entropy filter process on the X-ray CT image as a preprocessing of the clustering process in the above-mentioned first solving means. Effect of the Invention
[0014] According to the present invention, it is possible to provide an aggregate extraction device and an aggregate extraction method that are capable of accurately specifying the shape and distribution of aggregates. [Brief description of the drawings]
[0015] [Figure 1] 1 is a block diagram showing a functional configuration of an aggregate extraction device according to an embodiment of the present invention. [Diagram 2] 1 is a flowchart illustrating an aggregate extraction method according to an embodiment of the present invention. [Diagram 3] 11 is an image showing a result of image processing in one embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0016] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. The aggregate extraction device A according to this embodiment is a general-purpose computer equipped with an aggregate extraction program (application program), and includes a communication unit 1, a memory unit 2, an operation unit 3, a calculation unit 4, and a display unit 5, as shown in FIG.
[0017] Of the communication unit 1, storage unit 2, operation unit 3, calculation unit 4, and display unit 5, the communication unit 1 corresponds to an acquisition unit of the present invention. The storage unit 2, operation unit 3, and calculation unit 4 constitute an image processing unit of the present invention. Although details will be described later, the storage unit 2, operation unit 3, and calculation unit 4 work together to generate an aggregate extraction image Gr by performing clustering processing based on the k-means method, morphological conversion processing, region release processing, and active contour extension processing on the X-ray CT image.
[0018] The communication unit 1 is electrically connected to an external X-ray CT device and acquires an X-ray CT image G0 by communicating with the X-ray CT device. The X-ray CT device generates an X-ray CT image Gx of an object based on a well-known X-ray CT method. The object in this embodiment is a mortar / concrete test specimen containing aggregates such as sand and stones.
[0019] That is, the X-ray CT device in this embodiment is an imaging device that utilizes the fact that the amount of X-ray absorption differs depending on the material (aggregate, cement, etc.) of the mortar / concrete test specimen. Also, the X-ray CT image Gx in this embodiment is a cross-sectional image of the mortar / concrete test specimen captured by the X-ray CT device, and is a black and white image in which the aggregate part and the cement part in the mortar / concrete test specimen have different lightness (brightness). The communication unit 1 acquires the X-ray CT image Gx of the mortar / concrete test specimen from the X-ray CT device based on an image acquisition instruction input from the calculation unit 4, and outputs this X-ray CT image Gx to the calculation unit 4.
[0020] The memory unit 2 is made up of multiple storage devices such as a ROM (Read Only Memory), a RAM (Random Access Memory), and a hard disk, and stores the above-mentioned aggregate extraction program as well as the X-ray CT images Gx of the mortar / concrete test specimens input from the calculation unit 4. The memory unit 2 also temporarily stores various data and images generated in the process of execution of the aggregate extraction program by the calculation unit 4, and provides these data and images to the calculation unit 4 in response to a read request from the calculation unit 4.
[0021] The operation unit 3 is composed of a keyboard, a mouse, etc., and receives instructions (operation instructions) from the operator and outputs them to the calculation unit 4. These operation instructions are, for example, an instruction to start execution of the aggregate extraction program and an instruction to output the execution result of the aggregate extraction program (aggregate extraction image Gr of a mortar / concrete test specimen). The operator inputs his / her own instructions to the operation unit 3, i.e., the aggregate extraction device A, by manually operating the operation unit 3.
[0022] The calculation unit 4 generates an aggregate extraction image Gr of the mortar / concrete specimen by executing an aggregate extraction program (application program) on a predetermined OS (Operating System: basic software). This calculation unit 4 outputs various data and images generated in the process of executing the aggregate extraction program to the storage unit 2 and stores them temporarily. The calculation process of the calculation unit 4 based on the aggregate extraction program will be described later as the operation of the aggregate extraction device A.
[0023] Moreover, the calculation unit 4 outputs an aggregate extraction image Gr of the mortar / concrete test specimen, which is a result of execution of the aggregate extraction program, to the display unit 5. Furthermore, when an instruction to transmit the aggregate extraction image Gr to an external device is input from the operation unit 3, the calculation unit 4 outputs the aggregate extraction image Gr of the mortar / concrete test specimen to the communication unit 1 to transmit it to the external device.
[0024] The display unit 5 displays the aggregate extracted image Gr of the mortar / concrete test specimen input from the calculation unit 4. That is, the display unit 5 displays the aggregate extracted image Gr of the mortar / concrete test specimen input from the calculation unit 4. Such a display unit 5 is, for example, a liquid crystal display device.
[0025] Next, the operation of the aggregate extraction device A according to this embodiment, that is, an aggregate extraction method using the aggregate extraction device A, will be described in detail with reference to FIGS.
[0026] In this aggregate extraction device A, the calculation unit 4 executes a series of processes shown in FIG. 2 based on the aggregate extraction program to generate an aggregate extraction image Gr of the mortar / concrete test specimen, which is the final product.
[0027] First, when an instruction to execute an aggregate extraction program is input from the operation unit 3, the calculation unit 4 outputs an instruction to acquire an X-ray CT image Gx based on the aggregate extraction program to the communication unit 1. That is, the calculation unit 4 acquires an X-ray CT image Gx of a mortar / concrete test specimen by outputting an instruction to acquire an X-ray CT image Gx to the communication unit 1 (step S1).
[0028] When an instruction to acquire an X-ray CT image Gx is input from the calculation unit 4, the communication unit 1 transmits a transmission request for the X-ray CT image Gx to the X-ray CT device. In response to this transmission request, the X-ray CT device transmits the X-ray CT image Gx of the mortar / concrete test specimen to the communication unit 1. Then, the communication unit 1 outputs the X-ray CT image Gx of the mortar / concrete test specimen received from the X-ray CT device to the calculation unit 4.
[0029] The computing unit 4 temporarily stores the X-ray CT image Gx of the mortar / concrete specimen in the memory unit 2, reads it out from the memory unit 2, and sequentially performs the following four image processing steps: the first image processing step is a filter processing step (step S2), the second image processing step is a k-means clustering processing step (step S3), the third image processing step is a morphological transformation processing step (step S4), the fourth image processing step is a region release processing step (step S5), and the fifth or final image processing step is an active contour extension processing step (step S6).
[0030] Here, the k-means clustering process (step S3) corresponds to the first step of the aggregate extraction method according to this embodiment. The morphological transformation process (step S4) corresponds to the second step of the aggregate extraction method according to this embodiment. The region release process (step S5) corresponds to the third step of the aggregate extraction method according to this embodiment. Furthermore, the active contour expansion process (step S5) corresponds to the fourth step of the aggregate extraction method according to this embodiment.
[0031] That is, the k-means clustering process (first step) is an image processing step in which a clustering process based on the k-means method is applied to an X-ray CT image of a mortar / concrete test specimen containing aggregate and cement. The morphological transformation process (second step) is an image processing step in which a morphological transformation process is applied to the processing result of the first step. The region release process (third step) is an image processing step in which a region release process is applied to the processing result of the second step. The active contour extension process (fourth step) is an image processing step in which an active contour extension process is applied to the processing result of the third step.
[0032] The filtering process (step S2) is an entropy filtering process using an entropyfilt function, a high-frequency noise reduction process using a Gaussian filter, and a contrast enhancement process. In this filtering process, the X-ray CT image Gx of the mortar and concrete specimen is subjected to the filtering process (step S2) in the order of the entropy filtering process, the high-frequency noise reduction process, and the contrast enhancement process.
[0033] Here, the above entropy filter processing utilizes the fact that there is little variation in the brightness values of aggregates in the X-ray CT image Gx. When focusing on a certain point in the image, the entropy filter converts the pixel to white if there is a large variation in the brightness values around that pixel, and conversely, if the brightness values have little variation and are homogeneous, the pixel is converted to black. Therefore, when entropy filter processing is performed on the X-ray CT image Gx of a mortar and concrete test specimen, the aggregate part is converted to black because it is relatively homogeneous, and the cement part is converted to white because the brightness value variation is relatively large.
[0034] 3(a) is a first intermediate image G1 showing the result of such filtering (step S2). From this first intermediate image G1, it can be seen that the filtering (step S2) allows extraction of only uniform areas compared to the surroundings, and therefore allows rough extraction of aggregates.
[0035] The k-means clustering process (step S3) is one of the AI image processing methods classified as unsupervised machine learning, and is an image processing method that divides given image data into k clusters specified in advance. This k-means clustering process (step S3) is easy to perform clustering because filtering (step S2), such as entropy filtering, has been applied to the X-ray CT image Gx of the mortar / concrete test specimen as pre-processing.
[0036] The morphological transformation process (step S4) is a general term for shape-based image processing, which is an image processing method that smooths the surface of a structure and fills in the inside of a structure by combining dilation and contraction of an object in an image. The dilation is a process of adding pixels to the boundary of an object in an image to expand the object. Conversely, the contraction is a process of removing pixels from the boundary of an object.
[0037] As image processing that combines the above-mentioned expansion and contraction, morphological opening and morphological closing are known. In the morphological conversion processing (step S4) in this embodiment, the k-means clustering processing (step S3) is followed by an opening processing and a closing processing to remove small objects and fill holes in the objects in the first intermediate image G1.
[0038] Morphological opening is a process that performs dilation after erosion, and can remove small objects from an image. For example, in morphological opening, which performs erosion and dilation twice each (erosion → erosion → dilation → dilation), objects that are 4 pixels or smaller in size disappear in the first two erosion processes, and are completely removed because they cannot be restored in the subsequent two dilation processes.
[0039] Conversely, objects that are 4 pixels or larger will not be removed by opening, but after processing, the fine pixel arrangement information near the boundary will be lost, resulting in a smooth boundary. In other words, morphological closing is a process that performs contraction after dilation, and has the effect of filling holes inside objects and smoothing boundaries.
[0040] 3(b) shows the second intermediate image G2 and the third intermediate image G3, which are results of such a morphological transformation process (step S4). From the second intermediate image G2 and the third intermediate image G3, it can be seen that noise has been effectively removed by the morphological transformation process (step S4).
[0041] Area release processing (step S5) is a general term for image processing that removes objects based on the volume of the objects in the image. Area release processing (step S5) can be performed immediately after k-means clustering processing (step S3), but the calculation time increases as the number of objects to be removed increases, so the calculation is made more efficient by performing morphological transformation processing (step S4) in advance.
[0042] 3(c) is a third intermediate image G3 showing the result of the area release process (step S5). Pixel elements of 2.0×105 or less are removed from the result of the morphological transformation process (step S4). According to this third intermediate image G3, it can be seen that the area release process (step S5) effectively removes noise that could not be removed by the morphological transformation process (step S4).
[0043] The active contour expansion process (step S6) is image processing based on the active contour expansion method. The active contour expansion method, also known as "snakes", is a segmentation algorithm that performs iterative region expansion. In this active contour expansion method, the image to be segmented and the initial contour to be used for segmentation are first specified.
[0044] The initial contour is then expanded to its surroundings, but if there is a large change in brightness in the image to be segmented during the expansion, the expansion is not performed. By repeating this process a specified number of times, the active contour expansion method can segment structures in the image.
[0045] In the active contour expansion process (step S5) in this embodiment, the contour of the aggregate part in the third intermediate image G3 is extracted by using an active contour expansion method. Fig. 3(d) shows an aggregate extraction image Gr obtained as a result of the active contour expansion process (step S5). With this aggregate extraction image Gr, it is possible to extract only the aggregate part in the third intermediate image G3 with high accuracy.
[0046] According to this embodiment, image processing is sequentially performed on the X-ray CT image Gx of the mortar / concrete test specimen in the order of filter processing (step S2), k-means clustering processing (step S3), morphological transformation processing (step S4), and active contour extension processing (step S5), so that aggregates contained in the mortar / concrete test specimen can be extracted more clearly from other materials. Therefore, according to this embodiment, it is possible to provide an aggregate extraction device A and an aggregate extraction method capable of accurately identifying the shape and distribution of aggregates.
[0047] The present invention is not limited to the above-described embodiment, and the following modifications are possible. (1) In the above embodiment, the X-ray CT image Gx of the mortar / concrete test specimen is subjected to the image processing in the order of filtering (step S2), k-means clustering (step S3), morphological transformation (step S4), and active contour extension (step S5), but the present invention is not limited to this. For example, the k-means clustering (step S3) may be omitted as necessary.
[0048] (2) In the above embodiment, the entropy filter process, the high frequency noise reduction process, and the contrast enhancement process are performed on the X-ray CT image Gx of the mortar and concrete test object as the filtering process (step S2), but the present invention is not limited to this. For example, only the entropy filter process may be performed on the X-ray CT image Gx of the mortar and concrete test object as the filtering process (step S2). [Explanation of symbols]
[0049] A Aggregate Extraction Device 1. Communications Department 2 Storage section 3 Control section 4 Arithmetic section 5 Display section
Claims
1. an acquisition unit that acquires an X-ray CT image of a mortar / concrete specimen including aggregate and cement; an image processing unit that generates an aggregate extraction image by performing clustering processing based on a k-means method, morphological transformation processing, region release processing, and active contour extension processing on the X-ray CT image; An aggregate extraction device comprising:
2. The bone material extraction device according to claim 1 , wherein the image processing unit performs entropy filtering on the X-ray CT image as a preprocessing step of the clustering process.
3. 3. The aggregate extraction device according to claim 1, wherein the morphological transformation processing comprises an opening processing and a closing processing.
4. A first step of performing a clustering process based on a k-means method on an X-ray CT image of a mortar / concrete specimen including aggregate and cement; a second step of performing a morphological transformation process on the result of the first step; a third step of performing an area release process on the processing result of the second step; a fourth step of performing an active contour expansion process on the processing result of the third step; The method for extracting aggregates comprising the steps of:
5. The bone material extraction method according to claim 4 , further comprising the step of: subjecting the X-ray CT image to an entropy filter process as a preprocessing step for the clustering process.
Citation Information
Patent Citations
Inspection device and inspection method
JP2009180559A
Image processing method
JP2022025864A
Computer-implemented method for segmenting measurement data of an object
JP2023502451A
Image processing device, image processing method, and program
WO2020066456A1
Nondestructive inspection method
JP2019174255A
