Software imagej-based identification method for processing pore parameters in magnesium phosphate cement cementing material images in batches

Abstract

The invention discloses an identification method for batch processing of pore parameters in a magnesium phosphate cement cementing material image based on software imagej, and relates to the field of magnesium phosphate cement microstructure image identification, in particular to an identification method for batch processing of pore parameters in a magnesium phosphate cement cementing material image based on software imagej. The invention aims to solve the technical problems that the SEM image data volume of MPC is large, and the actual demand is difficult to meet by purely depending on manual analysis. According to the method, image recognition of the pore parameters of the magnesium phosphate cement cementing materials in batches is realized by utilizing ImageJ software, and the processing efficiency is greatly improved. According to the method, subjectivity and errors of manual judgment in a traditional method are avoided, an image processing technology and a machine learning algorithm are combined, and a new thought and method are provided for analysis of the pore structure of the porous material. With continuous development of the technology, the method is expected to be applied and popularized in more fields, and innovation and development of related technologies are promoted.

Description

Technical Field

[0001] The invention relates to the field of magnesium phosphate cement microstructure image recognition, and in particular to a method for identifying pore parameters in magnesium phosphate cement gelling material images by batch processing based on software ImageJ. Background Art

[0002] Magnesium phosphate cement (MPC) is a new type of cementitious material. It is widely used in repair engineering, building structures and refractory materials due to its excellent properties such as rapid solidification, high early strength and good durability. In these applications, pore structure is one of the key factors affecting the performance of MPC. Parameters such as pore size, shape, distribution and connectivity will directly affect the mechanical properties, durability and impermeability of the material. Therefore, it is particularly important to accurately, quickly and efficiently identify and analyze the pore structure of MPC.

[0003] Traditional pore structure analysis methods mainly rely on manual observation and measurement of microscope images. This method is not only time-consuming and labor-intensive, but also highly subjective, making it difficult to obtain accurate and consistent results. In addition, numerical simulation methods such as finite element analysis (FEA) usually simplify porous materials into homogeneous materials or simulate pore distribution through random pore models. This simplified treatment ignores the impact of actual pore distribution on material properties, resulting in a large deviation between the analysis results and the actual situation. With the development of scanning electron microscopy (SEM) technology, high-resolution MPC microstructure images can be obtained to clearly show the morphology and distribution of pores. However, the amount of SEM image data is large, and it is difficult to meet actual needs by relying solely on manual analysis. For this reason, the introduction of image recognition technology for automated analysis has become an inevitable choice. Summary of the invention

[0004] The present invention aims to solve the technical problem that the amount of SEM image data of MPC is large and it is difficult to meet the actual needs by simply relying on manual analysis, and provides a method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on software ImageJ.

[0005] The method for identifying pore parameters in magnesium phosphate cement binder images based on software ImageJ batch processing of the present invention is carried out in the following steps:

[0006] 1. Use scanning electron microscope to obtain SEM original images of magnesium phosphate cement in batches;

[0007] 2. Use the software ImageJ to set the scale of the image obtained in step 1 and select the required area;

[0008] 3. Import 1 / 3 of the images processed in step 2 into the trainable wekasegmentation plug-in in batches to train the classifier model. At the same time, click record in Macros in plugins to record the model code after training and get the corresponding training images. Open record again to record the running code behind 8-bit, closing operation, filling holes, denoising and calculation operations. Integrate the model code after training and the running code behind the operation process. Open ProcessFolder (ImageJ Macro) in Batch in ImageJ 1.x in Templates in record, embed these codes into the batch processing code edited in ImageJ Macros, and then train the remaining 2 / 3 images to get the corresponding training images and results.

[0009] 4. Run run("8-bit") in ImageJ Macros on all the training images obtained in step 3 to perform 8-bit operations to obtain grayscale images;

[0010] 5. Binarize the grayscale image obtained in step 4 by running run("Convert to Mask"); run("Options...","iterations=1count=6black do=Close"); run("FillHoles") in ImageJ Macros;

[0011] 6. De-noise the image by running run("Remove Outliers...","radius=3threshold=50which=Bright") in ImageJ Macros;

[0012] 7. Run the command "run("Analyze Particles...","size=10-Infinity display clear summarize overlay add composite")" in ImageJ Macros to calculate and obtain the pore parameter information;

[0013] 8. Import the pore parameter information obtained in step 7 into Box-Counting, calculate the fractal dimension, describe the complexity and irregularity of the pore structure through fractal geometry, and calculate the pore size distribution.

[0014] Box-Counting is an existing method: overlay the image at different scales, count the number of boxes containing pores N(∈) at each scale, and plot log(N(∈)) vs. log(1 / ∈). The slope is calculated as the fractal dimension D, which is given by: ∈ is the side length of the grid. For each scale ∈, the number of grids required to cover the object is counted and recorded as N(∈).

[0015] The present invention uses scanning electron microscopy technology to obtain the microscopic pore structure of magnesium phosphate cement. First, the scale is set, and then the original image is batch imported into trainable weka segmentation. Next, the classifier is trained: the classifier is trained by drawing different foregrounds and backgrounds on the image to obtain the result graph of the classifier; 8-bit operation is performed; the image is binarized: the image is closed in morphological processing; the image is filled with holes; the final binary image is obtained, and the binary image is denoised; the recognition results are compared; the pore structure analysis operation is performed to obtain the image porosity, the average pore size of the pores, the limiting pore size, the median pore size, the effective pore size, the maximum pore size area and other information; the complexity and irregularity of the pore structure are described by fractal geometry, and the pore size distribution of the pores is calculated. The present invention can simply, quickly, batch process cement pores similar to those in electron microscope images, derive the required pore parameters, and can save the trainable weka segmentation training model, which is convenient for subsequent graphic recognition.

[0016] The method of the present invention can identify the pore distribution of cement microstructure in a simple, fast and batch manner, and can provide strong support for research and application in related fields by deeply understanding the pore structure characteristics of the material.

[0017] The method of the present invention utilizes ImageJ software and its plug-in Trainable Weka Segmentation to realize image recognition of pore parameters of batch magnesium phosphate cement binder, greatly improving processing efficiency. Through the batch processing function, multiple images can be processed simultaneously, further shortening the processing time. Trainable Weka Segmentation is based on a machine learning algorithm and can accurately identify the pore structure in the image, avoiding the subjectivity and error of human judgment in traditional methods. The Box-Counting method is a commonly used method for calculating fractal dimension with high accuracy and reliability. The method is not only applicable to the image recognition of pore parameters of magnesium phosphate cement binder, but also can be applied to the pore structure analysis of other porous materials. At the same time, the method also has good adaptability to different types of images (such as microscope images, scanning electron microscope images, etc.).

[0018] The present invention can significantly shorten the research cycle and improve research efficiency through automation and batch processing functions. Researchers can obtain pore parameter data more quickly, providing strong support for subsequent analysis and research. Accurate pore parameter data helps researchers understand the relationship between the microstructure and performance of materials. By optimizing the pore structure, key performance indicators such as the mechanical properties and thermal properties of the material can be improved, providing a theoretical basis for the application and development of the material. This method combines image processing technology and machine learning algorithms to provide new ideas and methods for the pore structure analysis of porous materials. With the continuous development of technology, this method is expected to be applied and promoted in more fields, promoting the innovation and development of related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is the interface diagram of the display when setting the scale in step 2 of experiment 1;

[0020] Figure 2 The interface diagram of the display when the required area is selected in step 2 of experiment 1;

[0021] Figure 3 This is the interface diagram of the display when training the classifier model in step 3 of experiment 1;

[0022] Figure 4 It is the training image obtained in step 3 of experiment 1;

[0023] Figure 5 This is the interface diagram of the display when 8-bit is performed in step 4 of experiment 1;

[0024] Figure 6 This is the interface of the display when making binary in step 5 of experiment 1;

[0025] Figure 7 This is the interface diagram of the display when the closing operation is performed in step 5 of experiment 1;

[0026] Figure 8 This is the interface of the display when filling holes by clicking fill holes in step 5 of experiment 1;

[0027] Fig. 9 This is a diagram of the display interface when the noise reduction process is performed in step 6 of experiment 1;

[0028] Fig.10 This is the calculation result diagram obtained in step 7 of experiment 1;

[0029] Fig.11 for Fig.10 and Figure 2 The comparison diagram after superposition of pores;

[0030] Fig.12 This is the fractal dimension data graph obtained in step eight of experiment one. DETAILED DESCRIPTION

[0031] Specific implementation method 1: This implementation method is a method for identifying pore parameters in magnesium phosphate cementitious material images based on imagej software batch processing, which is specifically carried out in the following steps:

[0032] 1. Use scanning electron microscope to obtain SEM original images of magnesium phosphate cement in batches;

[0033] 2. Use the software ImageJ to set the scale of the image obtained in step 1 and select the required area;

[0034] 3. Import 1 / 3 of the images processed in step 2 into the trainable wekasegmentation plug-in in batches to train the classifier model. At the same time, click record in Macros in plugins to record the model code after training and get the corresponding training images. Open record again to record the running code behind 8-bit, closing operation, filling holes, denoising and calculation operations. Integrate the model code after training and the running code behind the operation process. Open ProcessFolder (ImageJ Macro) in Batch in ImageJ 1.x in Templates in record, embed these codes into the batch processing code edited in ImageJ Macros, and then train the remaining 2 / 3 images to get the corresponding training images and results.

[0035] 4. Run run("8-bit") in ImageJ Macros on all the training images obtained in step 3 to perform 8-bit operations to obtain grayscale images;

[0036] 5. Binarize the grayscale image obtained in step 4 by running run("Convert to Mask"); run("Options...","iterations=1count=6black do=Close"); run("FillHoles") in ImageJ Macros;

[0037] 6. De-noise the image by running run("Remove Outliers...","radius=3threshold=50which=Bright") in ImageJ Macros;

[0038] 7. Run the command "run("Analyze Particles...","size=10-Infinity display clear summarize overlay add composite")" in ImageJ Macros to calculate and obtain the pore parameter information;

[0039] 8. Import the pore parameter information obtained in step 7 into Box-Counting, calculate the fractal dimension, describe the complexity and irregularity of the pore structure through fractal geometry, and calculate the pore size distribution.

[0040] Specific implementation method 2: This implementation method is different from specific implementation method 1 in that: in step 1, a scanning electron microscope is used to obtain a batch of SEM original images of magnesium phosphate cement, which are saved in jpg or tif format, and the image format is a grayscale image. The rest is the same as specific implementation method 1.

[0041] Specific implementation method 3: This implementation method is different from specific implementation method 1 or 2 in that: in step 2, the image obtained in step 1 is imported into the software imagej, and according to the calibration on the image, click Set scale in Analyze, set the known distance and unit of length, and then select the required area. Others are the same as specific implementation method 1 or 2.

[0042] Specific implementation method 4: This implementation method is different from specific implementation methods 1 to 3 in that the method for training the classifier model described in step 3 is to use the magic wand tool to draw the foreground and background on the image to train the model. The rest is the same as specific implementation methods 1 to 3.

[0043] Specific implementation example 5: This implementation example is different from specific implementation example 4 in that the binarization process in step 5 is to perform a closing operation and fill holes. The rest is the same as specific implementation example 4.

[0044] Specific implementation example 6: This implementation example is different from specific implementation example 5 in that the specific operation of binarization in step 5 is: make binary for the grayscale image, then click close in options in binary to close the image, and then click fill holes in binary to fill the holes. The rest is the same as specific implementation example 5.

[0045] Specific implementation example 7: This implementation example is different from specific implementation example 6 in that: the noise reduction processing method described in step 6 is: click remove outliers in noise to perform noise reduction processing on the image. Other aspects are the same as specific implementation example 6.

[0046] Specific embodiment eight: This embodiment differs from specific embodiment seven in that the pore parameters described in step seven include porosity, average pore size, limiting pore size, median pore size, effective pore size, maximum pore size area, pore shape, pore distribution and connectivity. The rest is the same as specific embodiment seven.

[0047] Specific implementation method 9: This implementation method is different from specific implementation method 8 in that the calculation method of fractal dimension in step 8 is: first set the minimum and maximum box sizes and the step size of the box size, start the calculation, and FracLac will generate a report and chart containing the fractal dimension. The rest is the same as specific implementation method 8.

[0048] Specific implementation example 10: This implementation example is different from specific implementation example 8 in that the Box-Counting described in step 8 uses square grids to cover the image at different scales, counts the number of grids containing pores, and calculates the fractal dimension as the grid size changes. The rest is the same as specific implementation example 9.

[0049] The present invention is verified by the following tests:

[0050] Experiment 1: This experiment is a method for identifying pore parameters in magnesium phosphate cementitious material images based on imagej software batch processing, such as Figure 1-Figure 12 As shown, the specific steps are as follows:

[0051] 1. Use a scanning electron microscope to obtain the original SEM images of magnesium phosphate cement in batches, save them as jpg, and the image format is a grayscale image;

[0052] 2. Use the imagej software to set the scale for 1 / 3 of the number of images obtained in step 1 (such as Figure 1 ) and select the desired area (as shown Figure 2 shown);

[0053] To set the scale, click Set scale in Analyze, set the known distance and unit of length, and then select the desired area.

[0054] 3. Import 1 / 3 of the images processed in step 2 into the trainable wekasegmentation plugin in batches to train the classifier model (such as Figure 3 At the same time, by clicking the record in Macros in plugins, the model code after training is recorded and the corresponding training picture is obtained (as shown in Figure 4); open record again, record the running codes behind the 8-bit, closing operation, filling holes, denoising and calculation operations, integrate the model code after training and the running codes behind the operation process, open ProcessFolder (ImageJ Macro) in Batch in ImageJ1.x in Templates in record, embed these codes into the batch processing code edited in ImageJMacros, and then train the remaining 2 / 3 images to obtain the corresponding training images and results;

[0055] 4. Run run("8-bit") in ImageJ Macros on all the training images obtained in step 3 to perform 8-bit operations and obtain grayscale images (such as Figure 5 shown);

[0056] 5. Binarize the grayscale image obtained in step 4: Run the run("Convert to Mask") in ImageJ Macros (such as Figure 6 as shown); run("Options...","iterations=1count=6black do=Close")(as shown Figure 7 As shown, perform closing operation); run("FillHoles") (as shown Figure 8 As shown, fill the holes);

[0057] 6. Perform noise reduction on the image by running run("Remove Outliers...","radius=3threshold=50which=Bright") in ImageJ Macros (e.g. Fig. 9 shown);

[0058] 7. Run the run("Analyze Particles...","size=10-Infinity display clear summarize overlay add composite") in ImageJ Macros to calculate the pore parameter information (such as Fig.10 and Table 1); after the calculation is completed, copy the initially selected valid area ( Figure 2 ), the pore area selected for model calculation ( Fig.10 ) are superimposed and compared with the pores on the replicated effective area, such as Fig.11 As shown in the figure, from the comparison results in the yellow area, the model calculation results are relatively accurate and reliable, and have certain practical value;

[0059] 8. Import the pore parameter information obtained in step 7 into Box-Counting, calculate the fractal dimension, describe the complexity and irregularity of the pore structure through fractal geometry, and calculate the pore size distribution (such as Fig.12 and Table 1);

[0060] The Box-Counting method uses a square grid to cover the image at different scales, counts the number of grids containing pores, and calculates the fractal dimension as the grid size changes. The Box-Counting method is an existing method: covering the image at different scales, counting the number of boxes N(∈) containing pores at each scale, and plotting the logarithmic coordinates log(N(∈)) and log(1 / ∈). The slope is calculated to be the fractal dimension D, and the formula is ∈ is the side length of the grid. For each scale ∈, the number of grids required to cover the object is counted and recorded as N(∈).

[0061] Table 1

[0062]

Claims

1. A method for identifying pore parameters in magnesium phosphate cement binder images by batch processing based on imagej software, characterized in that The identification method of pore parameters in magnesium phosphate cement binder images based on imagej software batch processing is carried out in the following steps:

1. Use scanning electron microscope to obtain SEM original images of magnesium phosphate cement in batches; 2. Use the software ImageJ to set the scale of the image obtained in step 1 and select the required area; 3. Import 1 / 3 of the images processed in step 2 into the trainable wekasegmentation plug-in in batches to train the classifier model. At the same time, click record in Macros in plugins to record the model code after training and get the corresponding training images. Open record again to record the running code behind the 8-bit, closing operation, filling holes, denoising and calculation operations. Integrate the model code after training and the running code behind the operation process. Open ImageJ Macro in Batch in ImageJ 1.x in Templates in record. Embed these codes into the batch processing code edited in ImageJ Macros. Then train the remaining 2 / 3 images to get the corresponding training images and results. Fourth, run ImageJ Macros to perform 8-bit operations on all training images obtained in step 3 to obtain grayscale images; 5. Binarize the grayscale image obtained in step 4 by running ImageJ Macros; 6. Reduce the noise of the image by running ImageJ Macros; 7. Obtain pore parameter information by running ImageJ Macros calculations; 8. Import the pore parameter information obtained in step 7 into Box-Counting, calculate the fractal dimension, describe the complexity and irregularity of the pore structure through fractal geometry, and calculate the pore size distribution.

2. A method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on software ImageJ according to claim 1, characterized in that In step 1, a scanning electron microscope is used to obtain a batch of SEM original images of magnesium phosphate cement, which are saved in jpg or tif format, and the image format is a grayscale image.

3. A method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on software ImageJ according to claim 1, characterized in that In step 2, import the image obtained in step 1 into the software ImageJ. According to the calibration on the image, click Set scale in Analyze, set the known distance and unit of length, and then select the required area.

4. The method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on the software ImageJ according to claim 1, characterized in that The method for training the classifier model described in step 3 is to use the magic wand tool to draw the foreground and background on the image to train the model.

5. The method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on the software ImageJ according to claim 1, characterized in that The binarization process in step 5 includes closing operation and filling holes.

6. A method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on software ImageJ according to claim 5, characterized in that The specific operation of the binarization processing in step 5 is: by executing the binarization operation code in ImageJMacros, the grayscale image can be made binary, closed and filled holes.

7. The method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on the software ImageJ according to claim 1, characterized in that The noise reduction method described in step six is: by executing the denoising code in ImageJ Macros, the remove outliers operation can be implemented to denoise the image.

8. The method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on the software ImageJ according to claim 1, characterized in that The pore parameters described in step seven include porosity, average pore size, limiting pore size, median pore size, effective pore size, maximum pore area, pore shape, pore distribution and connectivity.

9. The method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on the software ImageJ according to claim 1, characterized in that The calculation method of the fractal dimension in step 8 is: first set the minimum and maximum box sizes and the step size of the box size, start the calculation, and FracLac will generate a report and chart containing the fractal dimension.

10. The method for identifying pore parameters in magnesium phosphate cementitious material images by batch processing based on the software ImageJ according to claim 1, characterized in that The Box-Counting described in step 8 uses square grids to cover the image at different scales, counts the number of grids containing pores, and calculates the fractal dimension as the grid size changes.

Images