A method for segmenting biological pores in soil CT images

Abstract

The present invention discloses a method for segmenting biological pores in soil CT images, including: performing CT scanning on soil samples and performing image preprocessing; using the "Purify" plug-in in the Fiji software to obtain the maximum pore network and isolated pores of each preprocessed image, respectively, and segmenting the coarse biological pores from the maximum pore network and the isolated pores respectively; using Tubeness to detect fine biological pores in isolated pores; separating fine biological pores; superimposing coarse biological pores and fine biological pores together, and naming them as biological pores; calculating biological pore-related characteristic parameters. The present invention combines the Fiji software, the distance transform watershed algorithm and the Tubeness filter, and combines the advantages of multiple image processing algorithms to accurately segment coarse biological pores and fine biological pores, solving the problem in the prior art that biological pores are difficult to accurately segment because they are connected to non-biological pores such as cracks with complex and diverse morphologies.

Description

Technical Field

[0001] The invention relates to the field of soil science and technology, and in particular to a method for segmenting biological pores in a soil CT image. Background Art

[0002] Soil pores are an important part of soil structure and have a significant impact on the physical, chemical and biological properties of soil. Soil pores include biological pores and non-biological pores. Biological pores are mainly produced by roots, biological channels and microbial activities, while non-biological pores are mainly formed by the natural accumulation of soil particles and external forces such as freeze-thaw cycles and dry-wet cycles. Biological pores play a key role in soil ventilation, drainage and root growth. Their distribution and morphological characteristics are crucial to improving soil structure and promoting crop growth.

[0003] With the development of computed tomography (CT) technology, it has become possible to obtain the three-dimensional structure of soil non-destructively. CT images can clearly show the pore structure inside the soil, providing a powerful tool for detailed research on soil pores. For example, the application number is 201911006128.8, and the name of the invention is a method for extracting the pore structure of the entire soil aggregate in a CT digital image. The method optimizes parameters based on the two characteristics of image size and connectivity, and uses the ambient light occlusion calculation method to extract the pore part of the soil aggregate. However, due to the small contrast difference between the biological pores and non-biological pores of soil aggregates in CT images, the segmentation is difficult, and the patented method cannot segment the biological pores and non-biological pores. For example, Li Jian. Study on the characteristics of biological and non-biological pores in typical farmland soils in southern Guangxi [D], a master's thesis of Guilin University of Technology. Section 2.3.2 of this document describes CT scanning and soil pore extraction and separation. However, the method of separating biological pores from non-biological pores recorded in this document requires the assumption that biological pores and macropores are separated. This method can only separate some independent biological pores, but cannot separate biological pores that are connected to non-biological pores. Summary of the invention

[0004] The present invention aims to segment biological pores according to morphological characteristics, thereby providing an important means for in-depth study of soil pore structure.

[0005] The technical solution of the present invention is as follows:

[0006] A method for segmenting biological pores in a soil CT image comprises the following steps:

[0007] Step S1: perform CT scanning on the soil sample and perform image preprocessing, and use the global threshold segmentation method to binarize the image to obtain a pore image;

[0008] Step S2: for each image preprocessed in step S1, the Purify plug-in in the Fiji software is used to obtain the maximum pore network, and the coarse biological pores are segmented from the maximum pore network;

[0009] Step S3: subtract the maximum pore network in step S2 from the pore image in step S1 to obtain isolated pores, and use Tubeness to detect fine biological pores in the isolated pores;

[0010] Step S4: separating the fine biological pores of step S3;

[0011] Step S5: Using the Fiji-Process-Image calculator-Add algorithm, the coarse biological pores of step S2 and the fine biological pores of step S4 are superimposed and named as biological pores;

[0012] Step S6: Calculate the biological pore-related characteristic parameters of step S5.

[0013] Furthermore, in step S1, performing CT scanning on the sample and performing image preprocessing includes the following steps:

[0014] Step S11: Collect soil aggregates or soil column samples, scan them using CT scanning equipment, reconstruct the scanned images using PITRE software, and convert them into Tiff format that can be analyzed. The image resolution is 3.25 μm.

[0015] Step S12: importing the image reconstructed in step S11 into Image J / Fiji software;

[0016] Step S13: Use Adjust-Brightness / Contrast to adjust the brightness and contrast of the image;

[0017] Step S14: using a non-mean filter to perform noise reduction;

[0018] Step S15: using an unsharp mask filter to reduce local volume effects;

[0019] Step S16: In order to remove boundary interference, a region of interest of 700×750×750 voxels in the center of the agglomerate is selected, and the actual size is 2275×2437.5×2437.5 μm 3 .

[0020] Step S17: The image is binarized using a global threshold segmentation method, where black pixels represent pores and white pixels represent soil solid phases.

[0021] Furthermore, step S2 specifically includes the following steps:

[0022] Step S21: Use the Purify plug-in in the Fiji software to obtain the pore network with the largest volume in the image obtained in step S17, and rename it as the largest pore network 1.

[0023] Because the maximum pore network is relatively complex and has a high porosity, the maximum pore network volume of the aggregate accounts for more than 50% of the total pore volume, and the maximum pore network contains non-biological pores with complex shapes. Therefore, the present invention first takes the maximum pore network as the research object for segmenting biological pores.

[0024] Step 22: Use Fiji-MorphoLibJ-Distance Transform Watershed 3D plug-in to segment the largest pore network.

[0025] In the largest pore network, biological pores and non-biological pores are connected together, and biological pores cannot be directly separated. The distance transform watershed algorithm (3D) is needed to segment pores of different forms. The distance transform watershed algorithm combines the advantages of the distance transform and the watershed algorithm to effectively separate and identify objects in three dimensions. It plays an important role in processing three-dimensional images with overlapping objects or complex internal structures. First, it uses the distance transform to calculate the distance from each pixel to the nearest background pixel. In the binary image, the value of each foreground pixel (pore) can be replaced by the distance from it to the nearest background pixel to highlight the boundary and internal features of the pore; then it uses the watershed algorithm to identify the boundaries between different objects to segment pores of different forms, that is, the image is regarded as a topographic map, where the gray value of the pixel represents the height. On such a topographic map, the local minimum is regarded as a puddle, and the water starts to rise from these puddles until the water in different puddles begins to meet and form a watershed. In image segmentation, the watershed represents the boundary between different pores.

[0026] Step S23: Use the Fiji-MorphoLibJ-Connected Components Labeling plug-in to label the image in step S22, label each pore segmented by the watershed algorithm, and then use the Fiji-MorphoLibJ-Labeledition plug-in to manually select the coarse biological pores.

[0027] Coarse biological pores have obvious tubular morphology and can be identified by naked eye during selection.

[0028] Step S24: After manually selecting all the coarse biological pores, in the Labeledition interface, click removeselected to remove all the coarse biological pores in the image, and then click done to obtain the image after removing the coarse biological pores.

[0029] Step S25: The image before the removal of the rough biological pores is subtracted from the image after the removal of the rough biological pores (Fiji-Process-Image calculator-Subtract) to obtain the rough biological pores, and the image is converted into 8 bits, and then the threshold is adjusted to make it binary, and then the image is processed by median filtering (3D) to separate the rough biological pores with larger volume, which are renamed as rough biological pores 1.

[0030] It is worth noting that there may be some finer biological pores in the largest pore network, which can be segmented using the same method as step S3 described later.

[0031] Step S26: Subtract the maximum pore network 1 in step S21 (Fiji-Process-Image calculator-Subtract) from all the pores in step S17 to obtain an isolated pore image, which is renamed as isolated pore 1. Purify the isolated pore 1 to obtain the pore network with the largest volume in the isolated pore image, which is renamed as maximum pore network 2. After inverting the image (Invert), 3D visualization (3D Viewer) is performed to determine whether there are coarse biological pores in the maximum volume pore network 2. If so, the coarse biological pores are separated according to steps S22-S25; if there are no coarse biological pores, the isolated pore 1 is subtracted from the maximum pore network 2 to obtain the isolated pore 2, and then the biological pore segmentation of the maximum pore network is stopped after the maximum pore network 3 is obtained.

[0032] Generally speaking, compared with the maximum pore network one, the maximum pore network two and the maximum pore network three are smaller in volume, simpler in morphology, and the biological pore segmentation steps are simpler and faster; in addition, there are no biological pores in the maximum pore network two and the maximum pore network three of some samples, which greatly speeds up the segmentation efficiency and reduces the interference of non-biological pores on subsequent segmentation.

[0033] Furthermore, step S3 specifically includes the following steps:

[0034] Step S31: Perform Tubeness filtering on the maximum pore network 3 obtained in step S26, estimate the diameter of the fine biological pores in the maximum pore network 3, set the σ value accordingly, obtain the corresponding image results, and name these images respectively; in the embodiment of the present invention, the σ values ​​are set to 5, 10, 15, 20, 25 and 30 respectively, and the corresponding biological pore diameters are 16.25, 32.5, 48.75, 65, 81.25 and 97.5 μm respectively. Obtain the corresponding image results, and name these images respectively: T5, T10, T15, T20, T25, T30.

[0035] Here, the diameter of biological pores is estimated, and then Tubeness filtering is performed by adjusting the σ value. Generally speaking, the smaller the interval of the σ value setting, the more the number, and the more accurate the detection of biological pores, but it takes more time; if the σ value is too large, fine biological pores will be blurred. If the image is large (>1GB), the use of Tubeness filter takes a long time, and the recognition efficiency of tubular structures can be accelerated by reducing the image size (Image-Adjust-Size).

[0036] Tubeness filter is a nonlinear filter used to detect tubular structures (such as blood vessels and nerve fibers) in image processing. Its purpose is to enhance the tubular structures in the image while suppressing non-tubular structures. It can detect tubular structures with specific directions and adapt to tubular structures of different scales. It has a certain robustness to changes in noise and image quality. Tubeness filter detects tubular structures in images based on the Hessian matrix (a second-order partial derivative matrix). The Hessian matrix has three eigenvalues ​​in three-dimensional space (λ1, λ2, λ3, where |λ1|<|λ2|<|λ3|), corresponding to eigenvectors in three orthogonal directions. For spot-like or spherical structures, the Hessian matrix will obtain three dominant eigenvalues, namely |λ1|,

[0037] |λ2|, |λ3| are both large; for the tubular structure, the Hessian matrix will obtain two dominant eigenvalues, namely

[0038] |λ1|≈0<<|λ2|≈|λ3|; For the sheet structure, the Hessian matrix will get a dominant eigenvalue, that is

[0039] |λ1|≈|λ2|≈0, and |λ3| is relatively large. Therefore, by calculating the eigenvalues ​​of the Hessian matrix, image features related to the tubular structure can be extracted.

[0040] Step S32: Perform 3D hysteresis thresholding (Fiji-Plugin-3D-3DHysteresis Thresholding) on ​​the result in step S31, set the upper and lower thresholds to 150 and 90 respectively, and name the obtained images HT5, HT10, HT15, HT20, HT25, and HT30 respectively.

[0041] The upper and lower thresholds are set here to distinguish all biological pores. After the image is processed by Tubeness filtering, biological pores with diameters close to the σ value have higher scores (larger grayscale values), and other pores have smaller grayscale values. It is necessary to continuously adjust the upper and lower thresholds to ensure that all biological pores can be distinguished. If the lower threshold is too large, some small biological pores will be removed.

[0042] 3D Hysteresis Thresholding is a commonly used edge detection method in image processing. It requires setting two thresholds: upper threshold and lower threshold. The gray value of each pixel in the 3D image is compared with the two pre-defined thresholds to segment the research object. The upper threshold is used to identify pixels that may belong to the object. Pixels above the upper threshold will be retained. Pixels with gray values ​​between the upper and lower thresholds will be retained if they are connected to the object pixels, otherwise they will be removed. Pixels with gray values ​​below the lower threshold will be removed. The settings of the upper and lower thresholds need to be adjusted according to the actual image information to achieve the best segmentation effect.

[0043] Step S33: The results in step S32 are superimposed together through mathematical operations (Fiji-Process-Imagecalculator-Max). Since only two images can be calculated at a time, HT5 and HT10 are calculated first, and their results are calculated with HT15, and then the result is calculated with HT20, and so on. Finally, the result of combining six images is obtained and renamed HT.

[0044] Max is used to compare the grayscale values ​​of pixels at the same position in two images and retain the pixels with the larger grayscale value between the two.

[0045] Further, the separation of fine biopores in step S4 comprises the following steps:

[0046] Step S41: Label each pore in the HT image (Connected Components Labeling), use Label size filtering to filter out pores whose volume is larger than the set value voxel, then convert the filtered image to 8 bits, and adjust the threshold to binarize it, the steps are the same as S25.

[0047] Because the biological pores in the aggregates themselves are relatively small in diameter, with a resolution of about 3-5 μm, pores below 300 voxels are mostly non-biological pores in the form of spots, or tubular pores with very small diameters, which can be excluded first by volume size. Therefore, here we can choose to filter out pores with a volume greater than 300 voxels. Of course, we can also choose to filter out pores with a volume greater than 400, 500, etc. voxels according to the specific situation, but the same batch of data needs to be unified when processing.

[0048] Tubeness treatment cannot remove all abiotic pores. After treatment, a large number of abiotic pores are still retained, but biological pores will be retained. Therefore, a large number of tiny abiotic pores are removed first.

[0049] Step S42: Label the binary image in step S41 (Connected Components Labeling), and then use the Analyze Region 3D function in MorphoLibJ to calculate the morphological characteristics of each pore, including volume, surface area, and equivalent ellipsoid radius, and export them as a csv file.

[0050] Step S43: Open the csv file in step S42 in Excel and further calculate the morphological characteristics of the pore structure, including agglomeration B, sphericity S and plate-likeness P. The calculation formula is as follows:

[0051]

[0052] S=36πV 2 / A 3

[0053] P=|R2| / |R1|

[0054] In the formula, each pore is replaced by an equivalent ellipsoid, which is an imaginary ellipsoid with the same volume as the pore. The smaller the agglomeration, the more the object is elongated, that is, the more tubular it is; the larger the sphericity, the more the pore is similar to a sphere; the larger the plate-likeness, the more the pore is plate-like.

[0055] Step S44: In Excel, pores with agglomeration <0.35, sphericity <0.15, and plate-likeness <0.40 are screened out. These pores are defined as fine biological pores, and all such pores are marked as 1, and other non-biological pores are all marked as 0. The title of this column is changed to Classification and saved as a csv file.

[0056] Fiji software cannot open excel files and needs to be saved as csv files.

[0057] Step S45, open the csv file in step S44 in Fiji software, open the Assign Measure to Label function in the MorphoLibJ plug-in, select the opened csv file in the Table Name option, select Classification in Column, set Min Value and Max Value to 0.5 and 1.5 respectively, and then click OK to screen out fine biological pores and name them fine biological pores;

[0058] Min Value and Max Value represent the minimum and maximum values ​​in the Classification column, respectively. In step S44, biological pores and non-biological pores are marked as 1 and 0, respectively. Therefore, the Min Value and Max Value can be set by yourself to separate the biological pores.

[0059] Step S46: Convert the fine biological pores of step S45 into 8 bits and binarize them by adjusting the threshold, using the same method as S25.

[0060] Further, in step S5, the coarse biological pores in steps S25 and S26 and the fine biological pores in step S46 are superimposed together using the Fiji-Process-Image calculator-Add algorithm and named as biological pores.

[0061] Furthermore, in step S6, the calculation of biological pore morphology parameters includes the following steps:

[0062] The biological pores were skeletonized using Skeletonise in Fiji software, and then the length and diameter of the biological pores were calculated; the volume and surface area of ​​the biological pores were calculated using Analyze 3D in the MorphoLibJ plug-in.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] The present invention combines the advantages of Fiji software, distance transform watershed algorithm and Tubeness filter and other image processing algorithms to accurately segment coarse biological pores and fine biological pores, solving the problem in the prior art that biological pores are difficult to accurately segment because they are connected to other non-biological pores such as cracks with complex and diverse shapes.

[0065] The present invention separates the largest pore network and isolated pores, and performs segmentation of coarse and fine biological pores respectively, avoiding the interference of coarse biological pores on the identification of fine biological pores by using Tubeness filter, while ensuring the segmentation efficiency, laying a foundation for in-depth research on soil science, crop growth and pore structure improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 The CT scan image of the soil sample in Example 1 is converted into an 8-bit grayscale image after reconstruction.

[0067] Figure 2 This is the grayscale image of the soil sample in Example 1 after preprocessing.

[0068] Figure 3 The maximum pore network in the binary image of the soil sample of Example 1 is shown.

[0069] Figure 4 Yes Figure 3 The image shown has the effect of being marked.

[0070] Figure 5 This is a flow chart of fine biopore segmentation of the soil sample of Example 1.

[0071] Figure 6 This is a flow chart of using Tubeness filter to detect fine biological pores in the soil sample of Example 1.

[0072] Figure 7 This is the effect diagram of Tubeness filtering processing.

[0073] Figure 8 This is the effect diagram of 3D hysteresis threshold processing. DETAILED DESCRIPTION

[0074] In order to make the technical solutions and advantages of the invention more clear, the invention is further described below in conjunction with specific examples and drawings.

[0075] Step S1: Soil samples with a surface thickness of 0-10 cm were randomly collected from the tillage and fertilization experimental base of Longkang Farm in Huaiyuan County, Anhui Province. After air-drying, 9 soil aggregates with a diameter of 3-5 mm were randomly selected from them. The samples were scanned at the Shanghai Light Source X-ray Imaging and Biomedical Applications Beamline Station (BL13HB) with a scanning energy of 26 KeV and a spatial resolution of 3.25 μm. Each sample was rotated 180° at a uniform speed on the sample stage. The scanning time for each sample was about 10 minutes. A total of 940 images were collected during this process.

[0076] Preprocess the scanned image:

[0077] First, the images were reconstructed using PITRE software and stored in Tiff format; secondly, the 940 images of the first sample were imported into Fiji software, and the brightness and contrast of the images were adjusted using Adjust-Brightness / Contrast, and the images were converted into 8-bit grayscale images; then, the noise reduction was performed using the Non-Local Means Denoising filter in the Biomedgroup plug-in, with the parameters set to noise variance 15 (Sigma=15) and smoothing factor 1 (δ=1). Next, edge enhancement was performed, and an unsharp mask filter with a radius of 1 was applied to reduce the local volume effect caused by image blur. Finally, in order to remove boundary interference, a region of interest of 700×750×750 voxels in the center of the agglomerate was selected, with an actual size of 2275×2437.5×2437.5μm 3 , the cropped grayscale image is as follows Figure 1 As shown. The global threshold segmentation method is used to binarize the image, and the pore structure is segmented out. The black pixels represent the pore structure, as shown in Figure 2 .

[0078] Step S2: Use the Purify plug-in in the Fiji software to obtain the maximum pore network in the above binary image and rename it as the maximum pore network 1, such as Figure 3 Then the Fiji-MorphoLibJ-Distance TransformWatershed (3D) plug-in is used to segment the largest pore network, and the connected components labeling plug-in (ConnectedComponents Labeling) is used to manually label the segmented pores. Each pore segmented by the watershed algorithm is labeled. The software displays each labeled pore in a different color in the image, as shown in the figure below. Figure 4 shown.

[0079] like Figure 5As shown, the Fiji-MorphoLibJ-Labeledition plug-in is then used to manually select the coarse biological pores. The coarse biological pores present a significant tubular structure and are easily identified by the naked eye without the interference of a large number of isolated pores. After manually selecting all the coarse biological pores, in the Labeledition interface, click remove selected to remove all the coarse biological pores in the image, and then click done to obtain the image after the coarse biological pores are removed. The coarse biological pores are obtained by subtracting the image after the coarse biological pores are removed from the image before the coarse biological pores are removed (Fiji-Process-Image calculator-Subtract), and the image is converted to 8 bits, and then the threshold is adjusted to make it binary, and then the image is processed by median filtering (Median filtering (3D)) to separate the coarse biological pores with larger volumes, which are renamed coarse biological pores in this embodiment.

[0080] use Figure 2 Subtract the largest pore network 1 (Fiji-Process-Image calculator-Subtract) from all the pores shown to obtain an isolated pore image, which is renamed as isolated pore 1. Purify the image to obtain the pore network with the largest volume in isolated pore 1, which is renamed as maximum pore network 2. After inverting the image of maximum pore network 2, 3D visualization (3D Viewer) is performed to manually judge whether there are coarse biological pores in the maximum volume pore network 2. If so, the biological pores are separated according to steps S22-S25; if there are no coarse biological pores, the isolated pore 1 is subtracted from the largest pore network 2 to obtain isolated pore 2, and then the maximum pore network 3 is obtained.

[0081] like Figure 5 As shown, in this embodiment, only the largest pore network 1 of the first sample contains biological pores, and there are no biological pores in the largest pore network 2 and the largest pore network 3.

[0082] Step S3: Segmentation process for fine biological pores Figure 5 , 6 As shown in the figure, Tubeness filtering is performed on isolated pores 3 to identify tubular structures in the image. The σ values ​​are set to 5, 10, 15, 20, 25 and 30, respectively, and the corresponding biological pore diameters are 16.25, 32.5, 48.75, 65, 81.25 and 97.5 μm, respectively. The corresponding image results are obtained and these images are named as: T5, T10, T15, T20, T25, T30; as shown in the figure, Figure 7The figures show the effect of Tubeness filtering processing of the samples in this embodiment when the σ values ​​are set to 5 and 10 respectively, where the σ value in a is 5 and the σ value in b is 10.

[0083] These images were then processed with 3D hysteresis thresholding (Fiji-Plugin-3D-3D HysteresisThresholding), with the upper and lower thresholds set to 150 and 90, respectively, and the resulting images were named HT5, HT10, HT15, HT20, HT25, and HT30. Figure 8 The figure shows the effect of 3D hysteresis threshold processing of the sample in this embodiment, where the σ value in a is 5 and the σ value in b is 10.

[0084] These results are superimposed together through mathematical operations (Fiji-Process-Image calculator-Max). Since only two images can be calculated at a time, HT5 and HT10 are calculated first, and their results are calculated with HT15, and then the result is calculated with HT20, and so on. Finally, the result of combining six images is obtained and renamed HT.

[0085] Step S4: Label each pore in the HT image (Connected Components Labeling), use Label size filtering to filter out pores with a volume larger than 300 voxels, then convert the filtered image to 8 bits, and adjust the threshold to binarize it; then label each pore in the binary image (Connected Components Labeling), and then use the Analyze Region 3D function in MorphoLibJ to calculate each morphological feature, including volume, surface area, and equivalent ellipsoid radius, and export it as a csv file.

[0086] The equivalent ellipsoids of different shapes are shown in Table 1:

[0087]

[0088] Open the csv file in Excel and further calculate the morphological characteristics of the pore structure, including agglomeration B, sphericity S, and plate-likeness P. The calculation formula is as follows:

[0089]

[0090] S=36πV 2 / A 3

[0091] P=|R2| / |R1|

[0092] In the formula, each pore is replaced by an equivalent ellipsoid, which is an imaginary ellipsoid with the same volume as the pore. The smaller the agglomeration, the more the object is elongated, that is, the more tubular it is; the larger the sphericity, the more the pore is similar to a sphere; the larger the plate-likeness, the more the pore is plate-like.

[0093] In the formula, R1, R2, and R3 represent the three radii of the equivalent ellipsoid, R1 is the major radius, which is the radius of the ellipsoid along the longest axis, also called the major axis or semi-major axis; R2 is the middle radius, which is the radius of the ellipsoid along the axis perpendicular to the major radius, also called the minor axis or semi-minor axis; R3 is the minor radius, which is the radius of the ellipsoid along the shortest axis, also called the radius, R1>R2>R3. A and V are the actual surface area and actual volume of each pore, respectively.

[0094] In Excel, pores with agglomeration <0.35, sphericity <0.15, and plate-likeness <0.40 were screened out. These pores were defined as fine biological pores, and all such pores were marked as 1, and all other non-biological pores were marked as 0. The title of the column was changed to Classification and saved as a csv file.

[0095] Open the csv file in Fiji software, and open the image after marking the pores with >300 voxels. Use the Assign Measure to Label function in the MorphoLibJ plug-in, select the opened csv file in the Table Name option, select Classification as the Column, set the Min Value and Max Value to 0.5 and 1.5 respectively, and then click OK to screen out the fine biological pores and name them as fine biological pores; convert the fine biological pore image to 8 bits and binarize it by adjusting the threshold.

[0096] Step S5: Using the Fiji-Process-Image calculator-Add algorithm, the coarse biological pores and the fine biological pores are superimposed together and named as biological pores.

[0097] Step S6: The calculation of biological pore morphological parameters includes the following steps:

[0098] In the Fiji software, the biological pores were skeletonized using Skeletonise, and then the length of the biological pores was calculated using Analyzeskeleton. The volume and surface area of ​​the biological pores were calculated using Analyze 3D in the MorphoLibJ plug-in. The results are shown in 2 below:

[0099] Table 2 Calculation results

[0100] quantity Overall length volume Surface Area 198 157mm <![CDATA[0.566mm 3 ]]> <![CDATA[21.4mm 2 ]]>

[0101] The above examples illustrate that the present invention can accurately segment the biological pores in soil aggregates, obtain information such as their binding morphology, spatial distribution and quantity, and segment soil column samples in the same way. This provides an important basis for a deep understanding of the impact of crop root growth on soil structure and its further impact on water and nutrient migration processes. Fiji / ImageJ used in the present invention is open source software, and it can be easily promoted and applied by being slightly familiar with its usage and image processing procedures.

Claims

1. A method for segmenting biological pores in soil CT images, characterized in that: The following steps are involved: Step S1: perform CT scanning on the soil sample and perform image preprocessing, and use the global threshold segmentation method to binarize the image to obtain a pore image; Step S2: for each image preprocessed in step S1, the Purify plug-in in the Fiji software is used to obtain the maximum pore network, and the coarse biological pores are segmented from the maximum pore network; Step S3: subtract the maximum pore network in step S2 from the pore image in step S1 to obtain isolated pores, and use Tubeness to detect fine biological pores in the isolated pores; Step S4: separating the fine biological pores of step S3; Separating fine biopores specifically includes the following steps: Step S41: label each pore in the image obtained in step S33, filter out pores with a volume larger than a set value voxel using label size filtering, then convert the filtered image into 8 bits, and adjust the threshold to binarize it; Step S42: Mark the binary image in step S41, use the Analyze Region 3D function in MorphoLibJ to calculate the morphological characteristics of each pore, including volume, surface area and equivalent ellipsoid radius, and export it as a csv file; Step S43: Open the csv file in step S42 in Excel to further calculate the morphological characteristics of the pore structure, including agglomeration B, sphericity S and plate-likeness P; Step S44: In Excel, pores with agglomeration <0.35, sphericity <0.15, and plate-likeness <0.40 are screened out. These pores are defined as fine biological pores, and all such pores are marked as 1, and all other non-biological pores are marked as 0. The title is changed to Classification and saved as a csv file; Step S45, open the csv file in step S44 in Fiji software, open the AssignMeasure to Label function in the MorphoLibJ plug-in, select the opened csv file in the Table Name option, select Classification in Column, set Min Value and Max Value to 0.5 and 1.5 respectively, and then click OK to screen out fine biological pores and name them as fine biological pores; Step S46: converting the fine biological pores in step S45 into 8 bits and binarizing them by adjusting the threshold; Step S5: Using the Fiji-Process-Image calculator-Add algorithm, the coarse biological pores of step S2 and the fine biological pores of step S4 are superimposed and named as biological pores; Step S6: Calculate the biological pore-related characteristic parameters of step S5.

2. The method for segmenting biological pores in soil CT images according to claim 1, characterized in that: The segmentation of the coarse biological pores in step S2 specifically includes the following steps: Step S21: Use the Purify plug-in in the Fiji software to obtain the pore network with the largest volume in the image obtained in step S1, and rename it as the largest pore network 1; Step 22: Use Fiji-MorphoLibJ-Distance Transform Watershed 3D plug-in to segment the largest pore network one; Step S23: Use the Fiji-MorphoLibJ-Connected Components Labeling plug-in to label the image in step S22, label each pore segmented by the watershed algorithm, and then use the Fiji-MorphoLibJ-Label edition plug-in to manually select the coarse biological pores; Step S24: After manually selecting all the coarse biological pores, in the Label edition interface, click removeselected to remove all the coarse biological pores in the image, and then click done to obtain the image after removing the coarse biological pores; Step S25: subtracting the image after removing the rough biological pores from the image before removing the rough biological pores to obtain the rough biological pores, and converting the image into 8 bits, then adjusting the threshold to make it binary, and then performing median filtering on the image to separate the rough biological pores with larger volume, and renaming them as rough biological pores 1; Step S26: Subtract the maximum pore network one in step S21 from all the pores in step S1 to obtain an isolated pore image, which is renamed as isolated pore one. Purify the isolated pore one to obtain the pore network with the largest volume in the isolated pore image, which is renamed as maximum pore network two. After inverting the image, 3D visualization is performed to determine whether there are coarse biological pores in the maximum volume pore network two. If so, the coarse biological pores are separated according to steps S22-S25. If there are no coarse biological pores, the isolated pore one is subtracted from the maximum pore network two to obtain isolated pore two, and then the maximum pore network three is obtained.

3. The method for segmenting biological pores in soil CT images according to claim 2, characterized in that: Step S3 specifically includes the following steps: Step S31: Perform Tubeness filtering on the maximum pore network 3 obtained in step S26, estimate the diameter of fine biological pores in the maximum pore network 3, set the σ value accordingly, obtain corresponding image results, and name these images respectively; Step S32: performing 3D hysteresis threshold processing on the images obtained in step S31, setting upper and lower thresholds and renaming the images respectively; Step S33: Add the results in step S32 together through mathematical operations.

4. The method for segmenting biological pores in soil CT images according to claim 1, characterized in that: Step S43 calculates the agglomeration B, sphericity S and plate-likeness P of the pore structure, and the calculation formula is as follows: In the formula, R1, R2, and R3 represent the three radii of the equivalent ellipsoid, respectively. R1 is the major radius, which is the radius of the ellipsoid along the longest axis, also called the major axis or semi-major axis; R2 is the median radius, which is the radius of the ellipsoid along the axis perpendicular to the major radius, also called the minor axis or semi-minor axis; R3 is the minor radius, which is the radius of the ellipsoid along the shortest axis, also called the radius. R 1 > R 2 > R 3 ； A and V are the actual surface area and actual volume of each pore, respectively.

5. The method for segmenting biological pores in soil CT images according to claim 1, characterized in that: The calculation of biological pore morphology parameters in step S6 includes the following steps: The biological pores were skeletonized using Skeletonise in the Fiji software, and then the length and diameter of the biological pores were calculated; the volume and surface area of ​​the biological pores were calculated using Analyze 3D in the MorphoLibJ plug-in.

6. The method for segmenting biological pores in soil CT images according to claim 1, characterized in that: In step S1, the sample is CT scanned and image preprocessed, including the following steps: Step S11: collecting soil aggregates or soil column samples, scanning them using CT scanning equipment, reconstructing the scanned images using PITRE software, and converting them into a Tiff format that can be analyzed; Step S12: importing the image reconstructed in step S11 into Image J / Fiji software; Step S13: Use Adjust-Brightness / Contrast to adjust the brightness and contrast of the image; Step S14: using a non-mean filter to perform noise reduction; Step S15: using an unsharp mask filter to reduce local volume effects; Step S16: In order to remove boundary interference, a region of interest with a size of 700×750×750 voxels in the center of the agglomerate is selected, and the actual size is 2275×2437.5×2437.5 μm 3 ; Step S17: The image is binarized using a global threshold segmentation method, where black pixels represent pores and white pixels represent soil solid phases.

Images