A method and system for rapid segmentation of minerals in rock CT scanning images

Abstract

The present invention relates to the field of geological exploration and development evaluation technology, and in particular to a method and system for rapid segmentation of minerals in rock CT scan images, including obtaining CT and QEMSCAN images and preprocessing them; registering the QEMSCAN images after the preprocessing operation with the CT scan images, identifying different mineral component types of the rock CT scan images, and marking different mineral types with different labels; training CT images with different labels and different mineral types, using StyleGAN to obtain a large number of high-quality pseudo CT images, using SegNet to train the pseudo CT images, obtaining a training model, and applying it to new rock CT scan images, which can quickly segment different minerals in the rock CT scan images. A large number of realistic pseudo CT images can be generated by StyleGAN, and combined with SegNet, accurate mineral rapid segmentation can be achieved on actual CT images, effectively improving the mineral segmentation accuracy of CT scan images.

Description

Technical Field

[0001] The invention relates to the technical field of geological exploration and development evaluation, and in particular to a method and system for rapid segmentation of minerals in rock CT scanning images. Background Art

[0002] The characterization of the three-dimensional spatial distribution of different mineral components of rocks is of great significance for accurately describing the internal mineral structure of rocks and fine evaluation of mineral resources. Rock CT scan images can accurately depict the three-dimensional spatial distribution of different mineral components of rocks. However, due to the similar mineral density and blurred boundaries when segmenting minerals in rock CT scan images, traditional image processing methods often find it difficult to balance the accuracy and efficiency of segmentation. Although there is currently a technology that uses the SegNet deep learning method to segment rock minerals on rock CT scan images, the processing effect is relatively poor due to the low accuracy and limited number of manually made mineral labels.

[0003] At present, QEMSCAN can be used to quickly and quantitatively identify and analyze the mineral components in rock samples. However, this technology can only provide two-dimensional information about rock minerals and cannot reflect the three-dimensional spatial distribution of minerals. Therefore, combining CT with QEMSCAN images can theoretically achieve the three-dimensional spatial distribution of different mineral components in rocks. However, since different minerals may have the same CT density distribution range, it is impossible to accurately distinguish minerals with similar densities, resulting in low accuracy in segmenting minerals in CT scan images. Summary of the invention

[0004] The present invention provides a method and system for rapid segmentation of minerals in rock CT scanning images, which is used to solve the above-mentioned problem existing in the prior art, that is, the problem of poor accuracy of mineral segmentation in CT scanning images in the prior art. The present invention provides a method for rapid segmentation of minerals in rock CT scanning images, which comprises:

[0005] Obtain CT scan images and QEMSCAN scan images of rocks and pre-process them;

[0006] The preprocessed QEMSCAN image is registered with the CT scan image by the feature point image registration method, the mineral component type of the CT scan image is identified according to the mapping relationship, and different mineral types are marked;

[0007] The CT scan images marked with different mineral types are used as training data, and the algorithm architecture in StyleGAN is used to train them to obtain the StyleGAN model; after training, a large number of pseudo CT images are generated through the StyleGAN model;

[0008] The pseudo CT images are trained and optimized through SegNet to obtain a training model, which is then applied to new rock CT scan images to identify different minerals and segment the rock CT scan images based on different minerals.

[0009] Optionally, the step of obtaining a CT scan image and a QEMSCAN scan image of the rock and preprocessing the image may include:

[0010] A full-diameter plug-like rock is obtained, and a CT scan is performed on it to obtain a CT scan image. The z-axis of the plug-like rock is smoothly cut, and the cut surface is scanned by QEMSCAN to obtain a QEMSCAN image. The CT scan image and the QEMSCAN image are denoised and contrast adjusted by median filtering, and the CT scan image and the QEMSCAN image are preprocessed.

[0011] Optionally, the CT scan images marked with different mineral types are used as training data, and the algorithm architecture in StyleGAN is used to train the images to obtain the StyleGAN model, which specifically includes:

[0012] CT images marked with different labels and various mineral types are used as training data. The training data set is divided into training set, validation set and test set according to the ratio of 70%, 20% and 10% respectively. The Stylegan2-ada algorithm architecture in StyleGAN is used for training to obtain the StyleGAN model.

[0013] Optionally, the pseudo CT image is trained and optimized by SegNet to obtain a training model, and the training model is applied to a new rock CT scan image to quickly segment different minerals in the rock CT scan image, specifically including:

[0014] A large number of pseudo CT images are used as training data, and the training data set is divided into training set, validation set and test set according to the ratio of 70%, 20% and 10%. Through the SegNet architecture, the cross entropy loss function and Adam optimizer are used to iterate on the training set, and the model performance is determined on the validation set to obtain the training model. The training model is applied to new rock CT scan images to identify different minerals and segment the rock CT scan images based on different minerals.

[0015] The present invention also provides a rock CT scanning image mineral rapid segmentation system, which specifically comprises:

[0016] An acquisition module is used to acquire CT scan images and QEMSCAN scan images of rocks and pre-process them;

[0017] A processing module is used to register the pre-processed QEMSCAN image with the CT scan image by a feature point image registration method, identify the mineral component type of the CT scan image according to the mapping relationship, and mark different mineral types;

[0018] The execution module is used to use CT scan images marked with different mineral types as training data, train them using the algorithm architecture in StyleGAN, and obtain a StyleGAN model; after training, generate a large number of pseudo CT images through the StyleGAN model; train and optimize the pseudo CT images through SegNet to obtain a training model, apply the training model to new rock CT scan images, identify different minerals, and segment the rock CT scan images based on different minerals.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention provides a method for rapid mineral segmentation in rock CT scan images, which obtains and preprocesses CT scan images and QEMSCAN images of rocks, aligns the preprocessed QEMSCAN images with the CT scan images according to a feature point image registration method, identifies different mineral component types in the rock CT scan images according to a mapping relationship of spatial positions, and marks different mineral types. A large number of realistic pseudo CT images can be generated through StyleGAN, and the pseudo CT images are trained in combination with SegNet, and the training model is applied to rock CT scan images, thereby achieving accurate rapid mineral segmentation in actual CT images, effectively improving the mineral segmentation accuracy of CT scan images. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0021] Figure 1 A flowchart of a method for rapid segmentation of minerals in rock CT scan images provided by an embodiment of the present invention;

[0022] Figure 2 A registered sandstone CT image provided by an embodiment of the present invention;

[0023] Figure 3 The QEMSCAN scanned image of sandstone after registration provided by the embodiment of the present invention;

[0024] Figure 4 A fast segmentation image of different minerals in a rock CT scan image provided by an embodiment of the present invention.

[0025] The above drawings have shown clear embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present invention in any way, but to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] First, the nouns involved in the present invention are explained:

[0028] QEMSCAN (Quantitative Evaluation of Minerals by Scanning Electron Microscopy) is a technology that combines scanning electron microscopy (SEM) and energy dispersive spectroscopy (EDS) to quickly and quantitatively identify and analyze mineral components in rock samples.

[0029] StyleGAN: It is a powerful generative adversarial network model that generates realistic images by training a generative model. It is good at generating high-resolution and diverse image data, with good diversity and controllability.

[0030] Rock CT scan images can accurately depict the three-dimensional spatial distribution of different mineral components of rocks. However, when segmenting minerals in rock CT scan images, this technology can only provide two-dimensional information about rock minerals and cannot reflect the three-dimensional spatial distribution of minerals. Therefore, combining CT with QEMSCAN images can theoretically achieve the three-dimensional spatial distribution of different mineral components of rocks. However, since different minerals may have the same CT density distribution range, traditional image processing methods are difficult to balance the accuracy and efficiency of segmentation. Although there is currently a technology that uses the SegNet deep learning method to segment rock minerals on rock CT scan images, the processing effect is relatively poor due to the low accuracy and limited number of manually made mineral labels.

[0031] In view of the above problems, the present invention provides a method for rapid mineral segmentation in rock CT scan images. The method collects CT and QEMSCAN images and preprocesses them, aligns the preprocessed QEMSCAN images with the CT scan images according to a feature point image registration method, identifies different mineral component types in the rock CT scan images according to a mapping relationship of spatial positions, and marks different mineral types. Conditional generation is performed through StyleGAN combined with the label information of the mineral type to obtain a large number of high-quality pseudo CT images. By training the pseudo CT images, SegNet is used to realize rapid segmentation of different minerals in the rock CT scan images, thereby effectively improving the mineral segmentation accuracy of the CT scan images.

[0032] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present invention will be described below in conjunction with the accompanying drawings.

[0033] Figure 1 FIG. 1 is a flow chart of a method for rapid segmentation of minerals in a rock CT scan image provided by an embodiment of the present invention. Figure 1 As shown, taking the sandstone collected from Well A of a certain oil field in China as an example, the method for rapid mineral segmentation of rock CT scanning images shown in this embodiment includes the following steps:

[0034] S1: Obtain CT scan images and QEMSCAN scan images of rocks and preprocess them.

[0035] Optionally, a full-diameter sandstone sample is collected. For example, a plug-like rock with a drill diameter of 2.54 cm and a length of 4.0 cm can be selected, and a CT scan can be performed on it to obtain a CT scan image; wherein the resolution during the CT scan can be selected to be 4.0 μm; then the plug-like rock is smoothly cut at a position of 2.0 cm along the z-axis, and the cut surface is scanned by using QEMSCAN to obtain a QEMSCAN scan image; then the CT scan image and the QEMSCAN scan image are respectively subjected to median filtering for denoising and contrast adjustment for image enhancement, and the CT scan image and the QEMSCAN scan image are preprocessed.

[0036] S2: The preprocessed QEMSCAN image is registered with the CT scan image through the feature point image registration method, the mineral component type of the CT scan image is identified according to the mapping relationship, and different mineral types are marked.

[0037] The QEMSCAN image after preprocessing is registered with the CT scan image using the feature point image registration method, and the different mineral component types of the rock CT scan image are identified according to the mapping relationship of the spatial position. Figure 2 is the sandstone CT image after registration, Figure 3 This is the QEMSCAN scanned image of sandstone after registration. The registered image is as follows Figure 2 and Figure 3 shown.

[0038] After the above-mentioned image registration and recognition are completed, different labels are marked for different mineral types, where the mineral types may be, for example, illite, quartz, feldspar, and heavy minerals.

[0039] S3: StyleGAN is used to train CT scan images labeled with different mineral types and generate pseudo CT images with determined labels and different mineral components.

[0040] Optionally, CT images with different labels and different mineral types marked by S2 are used as training data, and CT scan images marked with different mineral types are trained by StyleGAN to generate a large number of pseudo CT images.

[0041] Specifically, CT scan images labeled with different mineral types are used as training data, and the training data set is divided into three subsets: training set, validation set and test set according to the ratio of 70%, 20% and 10%. By selecting the Stylegan2-ada algorithm architecture in StyleGAN, the hyperparameters are continuously adjusted; wherein, the hyperparameters may be, for example, the learning rate (from 0.0001 to 0.001), the batch size (from 1 to 32), etc., and then the training data set is trained to achieve better training effect. After the training is completed, the trained StyleGAN model is used to generate 10,000 pseudo CT images with determined labels and composed of illite, quartz, feldspar and heavy mineral components.

[0042] Optionally, the learning rate can be adjusted to 0.001 and the batch size can be adjusted to 24.

[0043] S4: Use SegNet to train and optimize the pseudo CT images to obtain a training model, apply the training model to new rock CT scan images, identify different minerals, and segment the rock CT scan images based on different minerals.

[0044] Specifically, the generated 10,000 pseudo CT images are used as training data, and the entire training data set is divided into three subsets: training set, validation set, and test set according to the ratio of 70%, 20%, and 10%. The SegNet architecture is selected, and the cross entropy loss function and Adam optimizer are used. Multiple iterations are performed on the training set. After each round of iteration, the model performance is evaluated on the validation set. The network structure is adjusted according to the evaluation results until the optimal performance is achieved, that is, the training loss and validation loss are both stable, and the validation loss no longer decreases. Finally, the training model is applied to new rock CT scan images. Figure 4 As shown in the figure, SegNet is applied to rock CT scan images to identify different minerals and segment the rock CT scan images based on different minerals.

[0045] The present embodiment provides a method for rapid mineral segmentation in a rock CT scan image. The method obtains a CT image and a QEMSCAN image of the rock and preprocesses them. The QEMSCAN image after the preprocessing operation is registered with the CT scan image according to a feature point image registration method. Different mineral component types of the rock CT scan image are identified according to a mapping relationship of spatial positions, and different mineral types are marked with different labels. A large number of high-quality pseudo CT images with determined labels and different mineral components are obtained through StyleGAN, thereby improving the quality of image data. The pseudo CT images are trained by using SegNet, and the training model is applied to the rock CT scan images, thereby achieving accurate rapid mineral segmentation on the actual CT images, effectively improving the accuracy of mineral segmentation in CT scan images, and further improving the efficiency of mineral segmentation in CT scan images.

[0046] The present invention provides a rock CT scanning image mineral rapid segmentation system, comprising:

[0047] An acquisition module is used to acquire CT scan images and QEMSCAN scan images of rocks and pre-process them;

[0048] A processing module is used to register the pre-processed QEMSCAN image with the CT scan image by a feature point image registration method, identify the mineral component type of the CT scan image according to the mapping relationship, and mark different mineral types;

[0049] The execution module is used to use CT scan images marked with different mineral types as training data, train them using the algorithm architecture in StyleGAN, and obtain a StyleGAN model; after training, generate a large number of pseudo CT images through the StyleGAN model; train and optimize the pseudo CT images through SegNet to obtain a training model, apply the training model to new rock CT scan images, identify different minerals, and segment the rock CT scan images based on different minerals.

[0050] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0051] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of the present invention.

Claims

1. A method for rapid segmentation of minerals in rock CT scan images, characterized in that: The following steps are involved: Obtain CT scan images and QEMSCAN scan images of rocks and pre-process them; The preprocessed QEMSCAN scan image is registered with the CT scan image by the feature point image registration method, the mineral component type of the CT scan image is identified according to the mapping relationship, and different mineral types are marked; CT scan images marked with different mineral types are used as training data, and the algorithm architecture in StyleGAN is used to train them to obtain the StyleGAN model; after training, a large number of pseudo CT images are generated through the StyleGAN model; the pseudo CT images are trained and optimized through SegNet to obtain a training model, which is applied to new rock CT scan images to identify different minerals and segment the rock CT scan images based on different minerals; The step of obtaining the CT scan image and the QEMSCAN scan image of the rock and preprocessing the same specifically includes: Obtain a full-diameter plunger-like rock, perform CT scanning on it, obtain a CT scanning image, perform QEMSCAN scanning on the cut surface by smoothing the z-axis of the plunger-like rock, obtain a QEMSCAN scanning image, perform median filtering denoising and contrast adjustment on the CT scanning image and the QEMSCAN scanning image, and pre-process the CT scanning image and the QEMSCAN scanning image; The CT scan images with different mineral types are used as training data, and the algorithm architecture in StyleGAN is used to train them to obtain the StyleGAN model, which specifically includes: CT scan images marked with different labels and various mineral types are used as training data. The training data set is divided into training set, validation set and test set according to the ratio of 70%, 20% and 10% respectively. The Stylegan2-ada algorithm architecture in StyleGAN is used for training to obtain the StyleGAN model.

2. The method for rapid segmentation of minerals in rock CT scan images according to claim 1, characterized in that: The pseudo CT image is trained and optimized by SegNet to obtain a training model, and the training model is applied to a new rock CT scan image to identify different minerals and segment the rock CT scan image based on different minerals, specifically including: A large number of pseudo CT images are used as training data, and the training data set is divided into training set, validation set and test set according to the ratio of 70%, 20% and 10%. Through the SegNet architecture, the cross entropy loss function and Adam optimizer are used to iterate on the training set, and the model performance is determined on the validation set to obtain the training model. The training model is applied to new rock CT scan images to identify different minerals and segment the rock CT scan images based on different minerals.

3. A rock CT scanning image mineral rapid segmentation system, characterized in that: Specifically include: An acquisition module is used to acquire CT scan images and QEMSCAN scan images of rocks and preprocess them; the acquisition of CT scan images and QEMSCAN scan images of rocks and preprocessing them specifically includes: acquiring a full-diameter plunger-like rock, performing CT scanning on it to obtain a CT scan image, performing QEMSCAN scanning on the cut surface by smoothing the z-axis of the plunger-like rock to obtain a QEMSCAN scan image, and performing median filtering denoising and contrast adjustment on the CT scan image and the QEMSCAN scan image to preprocess the CT scan image and the QEMSCAN scan image; A processing module is used to register the pre-processed QEMSCAN scan image with the CT scan image by a feature point image registration method, identify the mineral component type of the CT scan image according to the mapping relationship, and mark different mineral types; An execution module is used to use CT scan images marked with different mineral types as training data, train them using the algorithm architecture in StyleGAN, and obtain a StyleGAN model; after training, generate a large number of pseudo CT images through the StyleGAN model; train and optimize the pseudo CT images through SegNet to obtain a training model, apply the training model to new rock CT scan images, identify different minerals, and segment the rock CT scan images based on different minerals; the CT scan images marked with different mineral types are used as training data, and the algorithm architecture in StyleGAN is used to train them to obtain the StyleGAN model, specifically including: using CT scan images marked with different labels and having multiple mineral types as training data, dividing the training data set into a training set, a validation set, and a test set according to the ratio of 70%, 20%, and 10%, and training them using the Stylegan2-ada algorithm architecture in StyleGAN to obtain the StyleGAN model.

Images