Method for identifying quartz sand fluid inclusion
By segmenting the image of quartz sand particles, the pixel ratio between the fluid inclusion and the quartz sand particles is determined, and the problem of low accuracy of quartz sand fluid inclusion recognition in the prior art is solved, and an efficient and fine recognition process is achieved, saving labor costs.
Patent Information
- Application Number
- CN202410506168.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-25
- Publication Date
- 2025-07-01
AI Technical Summary
The prior art is difficult to efficiently identify fluid inclusions in quartz sand, resulting in low recognition accuracy, cumbersome steps and high labor costs.
By acquiring multiple images to be detected in quartz sand particles, performing image segmentation processing, determining the pixel ratio of the fluid inclusions and the quartz sand particles, and realizing the identification of the quartz sand fluid inclusions.
It improves the recognition accuracy of quartz sand fluid inclusions, saves labor costs, and realizes large-scale fine detection of quartz sand particles.
Smart Images
Figure CN120236279A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image processing technology. Specifically, it relates to a method for identifying fluid inclusions in quartz sand. Background Art
[0002] There are a large number of fluid inclusions in quartz minerals. The trace impurity elements contained in the fluid inclusions are congenital conditions for evaluating whether quartz minerals can be used to prepare high-purity quartz. Generally, only based on the actual use results of quartz sand can the content of fluid inclusions in quartz sand be roughly obtained. Therefore, it is necessary to detect and screen quartz sand particles before the quartz sand is officially put into use. Since the main components of the fluid inclusions are substances such as liquids and gases generated or left during the purification of quartz sand, it is difficult to identify the fluid inclusions in quartz sand by conventional methods.
[0003] In related technologies, for the detection and identification of fluid inclusions in quartz sand, refractive oil is usually used to treat quartz sand particles, and then the treated particles are photographed using a polarized light microscope. However, this method is cumbersome and costly in terms of labor on the one hand. On the other hand, using a polarized light microscope can only view some areas of quartz sand particles, and it is impossible to perform large-scale and refined detection and identification of quartz sand particles, resulting in a low accuracy rate for identifying fluid inclusions in quartz sand. Summary of the Invention
[0004] The invention object of this application is to provide a method for identifying fluid inclusions in quartz sand, which can perform image processing on the images of a large number of quartz sand particles to determine the pixel ratio of fluid inclusions to quartz sand particles in the images, so as to complete the identification of fluid inclusions in quartz sand and improve the accuracy rate of identifying fluid inclusions in quartz sand.
[0005] The technical solution of this application is implemented as follows:
[0006] This application provides a method for identifying fluid inclusions in quartz sand, including:
[0007] Obtain a plurality of images to be detected corresponding to quartz sand particles;
[0008] For each of the images to be detected, perform image segmentation processing on the image to be detected to obtain a segmentation result corresponding to the image to be detected;
[0009] Determine the pixel ratio of fluid inclusions to quartz sand particles from the segmentation result;
[0010] Use the pixel ratio of each image to be detected as the identification result of fluid inclusions in quartz sand.
[0011] In some embodiments, the obtaining of the images to be detected corresponding to quartz sand particles includes:
[0012] Multiple original images of at least one quartz sand particle are obtained by a photographing device.
[0013] Each original image of the quartz sand particle is segmented respectively to obtain a quartz sand particle image as the image to be detected.
[0014] In some embodiments, the image segmentation process is performed on the image to be detected to obtain a segmentation result corresponding to the image to be detected, including:
[0015] Feature extraction is performed on the image to be detected to obtain a feature map of the image to be detected.
[0016] Multi-scale feature information is sampled from the feature map, and the classification probability of each pixel point in the image to be detected is determined according to the multi-scale feature information.
[0017] Each pixel point is classified according to the classification probability, and the obtained classification result is used as the segmentation result corresponding to the image to be detected.
[0018] In some embodiments, the categories of pixel points include fluid inclusion pixel points and quartz sand pixel points. The determining the classification probability of each pixel point in the image to be detected according to the multi-scale feature information includes:
[0019] Mapping processing is performed on the multi-scale feature information to obtain the probability that each pixel point belongs to a fluid inclusion pixel point and the probability that it belongs to a quartz sand particle pixel point respectively.
[0020] In some embodiments, the classifying each pixel point according to the classification probability includes:
[0021] For each pixel point in the image to be detected, when the probability that the pixel point belongs to a fluid inclusion pixel point is greater than the probability that it belongs to a quartz sand particle pixel point, it is determined that the pixel point is a fluid inclusion pixel point.
[0022] When the probability that the pixel point belongs to a fluid inclusion pixel point is less than the
[0023] probability
[0024] that it belongs to a quartz sand particle pixel point, it is determined that the pixel point is a quartz sand particle pixel point.
[0025] In some embodiments, the determining the pixel ratio between the fluid inclusion and the quartz sand particle from the segmentation result includes:
[0026] Determine the first total number of pixels of fluid inclusion pixel points and the second total number of pixels of quartz sand particle pixel points from the segmentation result;
[0027] Determine the ratio of the first total number of pixels to the second total number of pixels as the pixel ratio between the fluid inclusion and the quartz sand particle.
[0028] In some embodiments, it is characterized in that the image segmentation processing of the image to be detected is implemented by calling an image segmentation model, and the training method of the image segmentation model includes:
[0029] Obtain a sample of the image to be detected of quartz sand particles, wherein the sample of the image to be detected is marked with a true label;
[0030] Input the sample of the image to be detected into the image segmentation model for forward propagation to obtain a sample prediction result;
[0031] Take the difference between the sample prediction result and the true label as the loss value of the image segmentation model, and perform backpropagation in the image segmentation model through the loss value to update the parameters of the image segmentation model.
[0032] The embodiments of the present application have the following beneficial effects:
[0033] By obtaining a plurality of images to be detected corresponding to quartz sand particles, and performing image segmentation processing on each image to be detected to obtain the segmentation result corresponding to the image to be detected, and then determining the pixel ratio between the fluid inclusion and the quartz sand particle from the segmentation result. Thus, the identification process of the quartz sand fluid inclusion is transformed into a classification process based on image processing, and the fluid inclusion in the quartz sand is distinguished by using image segmentation and pixel calculation, which can more precisely identify the quartz sand fluid inclusion, and without any manual steps, it not only saves labor costs but also improves the identification accuracy. Description of the Drawings
[0034] Figure 1 is a flowchart of the method for identifying quartz sand fluid inclusions provided by the embodiments of the present application;
[0035] Figure 2 is the original image of the quartz sand particle provided by the embodiments of the present application;
[0036] Figure 3 is a schematic diagram of the area of the identified quartz sand particle provided by the embodiments of the present application;
[0037] Figure 4 is a schematic diagram of the area of the identified image background provided by the embodiments of the present application. Detailed Embodiments
[0038] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.
[0039] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.
[0040] In the following description, the terms "first", "second", and "third" are only used to distinguish similar objects, and do not represent a specific order for the objects. It can be understood that "first", "second", and "third" can be interchanged with a specific order or sequence when allowed, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.
[0041] Unless otherwise defined, all technical and scientific terms used in the present application have the same meaning as commonly understood by those skilled in the technical field to which the present application belongs. The terms used in the present application are only for the purpose of describing the embodiments of the present application, and are not intended to limit the present application.
[0042] Before further describing the embodiments of the present application, the nouns and terms involved in the embodiments of the present application are described. The nouns and terms involved in the embodiments of the present application are applicable to the following explanations:
[0043] 1) Fluid inclusion: It refers to the inclusion substances existing inside the stone minerals, generally gas and liquid. It is generated during the purification of stone minerals and may have an impact when the products of stone minerals are put into use. The fluid inclusions in the embodiments of the present application mainly refer to the fluid inclusions in quartz minerals (quartz sand).
[0044] 2) Image segmentation model: Generally refers to the model algorithm used for image segmentation, such as some neural network models. Image segmentation generally refers to semantic segmentation of images, which is essentially pixel-level classification, that is, feature extraction of images to classify the pixels in the images, so as to divide the images into regions.
[0045] The implementation of the method for identifying fluid inclusions in quartz sand provided by the present application will be further described in detail below in conjunction with the accompanying drawings of the specification.
[0046] See Figure 1 , Figure 1It is a schematic flowchart of a method for identifying fluid inclusions in quartz sand provided by an embodiment of the present application. The execution subject can be a server, an image processing system on a terminal, or other electronic devices, and will be described in combination with Figure 1 the steps shown below.
[0047] In step 101, a plurality of images to be detected corresponding to quartz sand particles are obtained.
[0048] First, for a large number of quartz sand particles (the particle size and quantity are not limited), the original images of the quartz sand particles are obtained by directly taking pictures or capturing frames through photography, and then the corresponding images to be detected are obtained by preprocessing or screening the original images.
[0049] In some embodiments, Figure 1 it is shown in [reference] that step 101 can be implemented through the following steps 1011 to 1012, and the following is a specific description.
[0050] In step 1011, the quartz sand particles are photographed multiple times by a photographing device to obtain multiple different original images of the quartz sand particles.
[0051] Here, the sample of the quartz sand particles can be placed under a paper sheet or a cover glass background, and then the quartz sand particles are photographed multiple times by a photographing device to obtain multiple different original images of the quartz sand particles. There can be multiple quartz sand particles here, and the quantity is not limited. The photographing device can be an industrial camera, a video camera, or a polarized light microscope with a photographing lens, and the quartz sand particles can be photographed multiple times from multiple angles or by controlling the movement of the device containing the quartz sand particles to obtain multiple original images of the quartz sand particles. As shown in Figure 2 it shows a micrograph of the quartz sand particles photographed by a polarized light microscope with a photographing lens as the original image.
[0052] In step 1012, each original image of the quartz sand particles is respectively segmented to obtain a quartz sand particle image as the image to be detected.
[0053] After the original image of the quartz sand particles is photographed, in order to facilitate the observation of the quartz sand particles, for each original image of the quartz sand, each original image of the quartz sand particles can be respectively segmented to obtain a quartz sand particle image as the image to be detected. Here, because it is considered that in the photographed original image, there is a background in addition to the quartz sand, for each original image, the SAM (Segment Anything Model) segmentation model can be used to segment the original image. Here, the segmentation process can be foreground segmentation, separating the quartz sand belonging to the foreground area from the background area, so as to remove the background area in the original image. For specific reference, see Figure 3 andFigure 4 , Figure 3 The schematic diagram of the area of quartz sand particles is shown in Figure 3 The darker-colored fragment area in it is the area of the segmented quartz sand particles. And Figure 4 The schematic diagram of the area of the recognized image background shown in Figure 4 The darker-colored background area in it is the area of the image background segmented by the image segmentation model. Thus, the original image of each quartz sand particle is segmented into multiple quartz sand particle images, which are used as multiple images to be detected.
[0054] In step 102, for each image to be detected, image segmentation processing is performed on the image to be detected to obtain the segmentation result corresponding to the image to be detected.
[0055] After obtaining multiple images to be detected of quartz sand particles through step 101, for each image to be detected, an image processing algorithm can be called to perform image segmentation processing on the image to be detected to obtain the segmentation result corresponding to the image to be detected. Here, the image processing algorithm can adopt a semantic-based image segmentation model, such as the HRNet (High-Resolution Network) model. Since there are multiple images to be detected, here, the semantic-based image segmentation model is used to perform image segmentation processing on each image to be detected respectively to obtain a segmented image with semantic annotations, and each pixel point in the segmented image is marked with a category. Thus, a corresponding segmentation result can be obtained for each image to be detected. The specific image segmentation processing process will be introduced below.
[0056] In some embodiments, Figure 1 It is shown in
[0057] In step 1021, feature extraction processing is performed on the image to be detected to obtain the feature map of the image to be detected.
[0058] In some embodiments, before inputting the image to be detected into the image segmentation model, generally, the size of the image to be detected is adjusted first. Since the images to be detected are all small pictures of quartz sand particles and the image sizes may vary, in order to facilitate the processing by the image segmentation model, all the images to be detected can be set to a fixed image scale, such as 256*256. Then, the image to be detected is input into the image segmentation model, and the convolutional layer in the image segmentation model will perform feature extraction processing on the image to be detected to obtain the feature map of the image to be detected, that is, the feature information of the image to be detected is extracted through the convolutional layer.
[0059] In step 1022, multi-scale feature information is sampled from the feature map, and the classification probability of each pixel point in the image to be detected is determined according to the multi-scale feature information.
[0060] Continuing from the above embodiments, after extracting the feature map of the image to be detected, the sampling layer in the image segmentation model is used to sample the feature map. The sampling process generally adjusts the scale of the feature map, and a feature map with a fixed scale can be obtained for each sampling. Multiple samplings can obtain multiple feature maps with different scales. The sampling process is divided into upsampling and downsampling. Through multiple upsamplings and downsamplings, multi-scale feature information of the image can be sampled from the feature map, and then the classification probability of each pixel point in the image to be detected is determined based on this multi-scale feature information. The following is a specific description.
[0061] In some embodiments, after sampling multi-scale feature information from the feature map, next, the multi-
[0062] scale feature information is subjected to mapping processing to obtain the probability that each pixel point belongs to a fluid inclusion pixel point and the probability that it belongs to a quartz sand particle pixel point, respectively.
[0063] Here, by integrating these multi-scale feature information and inputting this feature information into the classifier of the image segmentation model for mapping processing, these classifiers can be a fully connected layer neural network or a classification algorithm of machine learning. The mapping processing process is directed at each pixel point in the image, and the feature information is mapped and calculated to obtain the classification probability of the pixel point belonging to each category. In the embodiments of the present application, since there are fluid inclusions in the quartz sand particles, the categories of the pixel points are two categories: fluid inclusion pixel points and quartz sand particle pixel points. Therefore, in the mapping processing process, the obtained classification probability of the pixel point is the probability of belonging to a fluid inclusion pixel point and the probability of belonging to a quartz sand particle pixel point, that is, each pixel point in the image is mapped and calculated through the feature information, and a probability of belonging to a fluid inclusion pixel point and a probability of belonging to a quartz sand particle pixel point will be obtained.
[0064] In step 1023, each pixel point is classified according to the classification probability, and the obtained classification result is used as the segmentation result corresponding to the image to be detected.
[0065] Through step 1022, after obtaining the classification probability of each pixel point (that is, the probability of belonging to a fluid inclusion pixel point and the probability of belonging to a quartz sand particle pixel point) through mapping processing, each pixel point is classified according to the classification probability, and the obtained classification result is used as the segmentation result of the image to be detected.
[0066] In some embodiments, for each pixel point in the image to be detected, when the probability that the pixel point belongs to a fluid inclusion pixel point is greater than the probability that it belongs to a quartz sand particle pixel point, the pixel point is determined to be a fluid inclusion pixel point. Here, by comparing the classification probabilities of the pixel points, the classification of the pixel points can be determined. Because when the probability that the pixel point belongs to an image background pixel point is greater than the probability that it belongs to a quartz sand particle pixel point, it indicates that after image segmentation of the image to be detected, the possibility that this pixel point is judged to belong to the pixel points of the image background is very high, which means that the image segmentation model recognizes and classifies this pixel point as a pixel point of a fluid inclusion. Therefore, this pixel point can be directly determined to be a fluid inclusion pixel point and marked with a category.
[0067] In some embodiments, for each pixel point in the image to be detected, when the probability that the pixel point belongs to a fluid inclusion pixel point is less than the probability that it belongs to a quartz sand particle pixel point, the pixel point is determined to be a quartz sand particle pixel point. Because when the probability that the pixel point belongs to a fluid inclusion pixel point is less than the probability that it belongs to a quartz sand particle pixel point, it indicates that after image segmentation of the image to be detected, the possibility that this pixel point is judged to belong to the pixel points of quartz sand particles is very high, which means that the image segmentation model recognizes and classifies this pixel point as a quartz sand particle pixel point. Therefore, this pixel point can be determined to be a quartz sand particle pixel point and marked with a category.
[0068] Thus, for each pixel point in the image to be detected, by classifying each pixel point according to the classification probability, it can be judged that each pixel point belongs to a fluid inclusion pixel point and a quartz sand image pixel point, and thus the final classification result is obtained. Next, the obtained classification result is used as the segmentation result corresponding to the image to be detected. Because from the perspective of the image to be detected, the pixel points of the entire image to be detected will be divided into a region range belonging to fluid inclusions and a region range belonging to quartz sand particles, that is, the segmentation result corresponding to the image to be detected.
[0069] Continue to refer to Figure 1 , in step 103, the pixel ratio of fluid inclusions to quartz sand particles is determined from the segmentation result.
[0070] For the segmentation result of the image to be detected, the quartz sand particles and fluid inclusions have been divided. Next, relevant calculations need to be performed on the pixels of the quartz sand particles and fluid inclusions according to actual requirements, that is, the pixel ratio of fluid inclusions to quartz sand particles is determined from the segmentation result.
[0071] In some embodiments, Figure 1 The step 103 shown in
[0072] In step 1031, determine the first total number of pixels of fluid inclusion pixel points and the second total number of pixels of quartz sand particle pixel points from the segmentation result.
[0073] In some embodiments, it is first necessary to determine the first total number of pixels of fluid inclusion pixel points and the second total number of pixels of quartz sand particle pixel points from the segmentation result. Since in the segmentation result of the image to be detected, each pixel point in the image has been identified and classified, and corresponding category labels have also been made. Here, directly perform statistics according to the corresponding category labels, that is, count the total number of fluid inclusion pixel points according to the fluid inclusion label as the first total number of pixel points. Similarly, the total number of quartz sand particle pixel points can be directly counted according to the quartz sand label as the second total number of pixel points.
[0074] In step 1042, determine the pixel ratio of the quartz sand fluid inclusion according to the first total number of pixels and the second total number of pixels.
[0075] Next, perform relevant calculations on the first total number of pixel points and the second total number of pixel points as the pixel ratio of the fluid inclusion in the quartz sand particles.
[0076] In some embodiments, the actual requirement may be to classify the level of quartz sand particles to determine the quality of quartz sand. For example, calculate the IQR (Inclusions Quartz Ratio) index of quartz sand particles, that is, the pixel ratio of fluid inclusions in quartz sand particles, and the calculation method can be to directly calculate the ratio of the sum of the pixel points of fluid inclusions to the sum of the pixel points of quartz sand particles, that is, calculate the ratio of the first total number of pixel points to the second total number of pixel points, and then multiply the ratio by 100% to obtain the pixel ratio, that is, the IQR (Inclusions Quartz Ratio) index, as the pixel ratio of the quartz sand fluid inclusion.
[0077] In some other embodiments, the actual requirement may only be to determine the content of fluid inclusions in quartz sand. The calculation method is to first calculate the sum of the pixel points of the entire quartz sand particle, that is, calculate the sum of the total number of pixel points of fluid inclusions and the total number of pixel points of quartz sand particles as the sum of the pixel points of the quartz sand particle, then calculate the ratio of the first total number of pixel points of the fluid inclusion to the sum of the pixel points of the quartz sand particle, and then multiply the ratio by 100% as the pixel ratio of the quartz sand fluid inclusion to measure the content of fluid inclusions in quartz sand.
[0078] In step 104, use the pixel ratio corresponding to each image to be detected as the recognition result of the quartz sand fluid inclusion.
[0079] Through step 103, a pixel ratio can be finally calculated for each image to be detected, and this pixel ratio is the recognition result of fluid inclusions in quartz sand particles in the image to be detected. Since a large amount of recognition data of fluid inclusions is required for the detection and evaluation of quartz sand particles, here, the pixel ratio corresponding to each image to be detected can be recorded as the recognition result of fluid inclusions in quartz sand, and thus the whole process of recognizing fluid inclusions in quartz sand in the embodiments of the present application is introduced.
[0080] In the subsequent process, this recognition result can be used as a reference for the evaluation or classification of quartz sand. For example, the quartz sand particles can be classified according to the corresponding pixel ratio, or the standard deviation of these pixel ratios can be calculated to determine the dispersion of these data for the evaluation of quartz sand particles.
[0081] Exemplarily, after calculating each pixel ratio, according to a selected specific requirement index (IQR), the permeability of quartz sand particles can be determined using the pixel ratio (IQR), and they can be classified into permeable particles (the pixel ratio of fluid inclusions is less than 0.3%), semi-permeable particles (the pixel ratio of fluid inclusions is 0.3% to 5%), and impermeable particles (the pixel ratio of fluid inclusions is higher than 5%). The quality of a large number of quartz sands can also be classified according to the number of permeable particles to determine the criteria for putting them into use.
[0082] In some embodiments, the image segmentation processing of the image to be detected is implemented by invoking an image segmentation model, such as the HRNet (High-Resolution Network) model. The image segmentation model is trained by making samples with a large number of original images of quartz sand particles before recognizing fluid inclusions in quartz sand. The training method of the image segmentation model is specifically introduced below.
[0083] First, obtain the image samples to be detected of quartz sand particles. Generally, 20 to 30 groups of original images are preferably used as samples, which is not limited here. Of course, the more sample images, the better the model training effect and the higher the final recognition accuracy. Then each group of original images is respectively segmented to obtain multiple quartz sand particle images as the image samples to be detected. Here, considering that the image segmentation process of the image segmentation model is essentially a classification process, true labels are marked in the image samples to be detected to mark the pixel points belonging to fluid inclusions and the pixel points belonging to quartz sand.
[0084] After constructing the image sample of quartz sand particles to be detected, the image sample to be detected is input into the image segmentation model for forward propagation to obtain the sample prediction result. The forward propagation process here is to use the convolutional layer of the image segmentation model to perform feature extraction on the image sample to be detected to obtain the feature map of the image sample to be detected, and then use the sampling layer of the image segmentation model to sample multi-scale feature information from the feature map. Next, the classifier is used to determine the classification probability of each pixel point in the image to be detected according to the multi-scale feature information. Finally, each pixel point is classified according to the classification probability, and the obtained classification result is used as the sample prediction result (that is, each pixel point in the sample is predicted to belong to a fluid inclusion or quartz sand).
[0085] In each round of training, the image sample to be detected is subjected to forward propagation once. After each round of training, the sample prediction result obtained by forward propagation is compared with the true label, and the error (difference) between the sample prediction result and the true label is calculated as the loss value of the image segmentation model. Finally, the loss value is used for backpropagation in the image segmentation model, and the gradient descent optimization algorithm is used to update the parameters of the image segmentation model. The training is stopped until the loss value converges or reaches the specified number of training rounds.
[0086] The trained image segmentation model can be directly used for prediction. The image to be detected of quartz sand particles is directly input into the trained image segmentation model for prediction. After obtaining the prediction result, it is compared with the true label, and then the confusion matrix is calculated. Then, IoU (Intersection over Union) and PA (Pixel Accuracy) are introduced as evaluation parameters according to the confusion matrix. Finally, the final precision Precision is calculated through IoU and PA as the evaluation index of the model to measure the quality of the image segmentation model.
[0087] In summary, through the embodiments of the present application, the identification process of fluid inclusions is transformed into a process of processing image pixels. Based on multiple quartz sand particle images, the processing result of image segmentation is used to complete the division of the area range of quartz sand particles and the area range of fluid inclusions, and then relevant pixel calculations are performed in units of pixel points to complete the distinction of fluid inclusions in quartz sand. Even for a large number of quartz sand particles, the fluid inclusions in quartz sand can be accurately identified. Moreover, in the whole process, except for manually operating to take the original image of quartz sand particles, the whole identification process does not require any manual steps, which not only saves labor costs but also improves the identification accuracy.
[0088] The above is only the embodiments of the present application and is not intended to limit the protection scope of the present application. Any modifications, equivalent replacements, and improvements made within the spirit and scope of the present application are included in the protection scope of the present application.
Claims
1. A method for identifying quartz sand fluid inclusions, characterized in that: The method comprises: Acquire multiple images to be detected corresponding to quartz sand particles; For each of the images to be detected, performing image segmentation processing on the image to be detected to obtain a segmentation result corresponding to the image to be detected; Determining a pixel ratio of the fluid inclusion to the quartz sand grain from the segmentation result; The pixel ratio corresponding to each of the images to be detected is used as the recognition result of the quartz sand fluid inclusion.
2. The method according to claim 1, characterized in that The step of obtaining a plurality of images to be detected corresponding to the quartz sand particles includes: The quartz sand particles are photographed multiple times by a photographing device to obtain multiple different original images of the quartz sand particles; Each original image of quartz sand particles is segmented respectively to obtain multiple quartz sand particle images as images to be detected.
3. The method according to claim 1, characterized in that The performing image segmentation processing on the image to be detected to obtain a segmentation result corresponding to the image to be detected includes: Performing feature extraction processing on the image to be detected to obtain a feature map of the image to be detected; Sampling multi-scale feature information from the feature map, and determining the classification probability of each pixel in the image to be detected according to the multi-scale feature information; Each pixel point is classified according to the classification probability, and the obtained classification result is used as the segmentation result corresponding to the image to be detected.
4. The method according to claim 3, characterized in that The categories of pixel points include fluid inclusion pixel points and quartz sand particle pixel points, and the determining the classification probability of each pixel point in the image to be detected according to the multi-scale feature information includes: The multi-scale feature information is mapped to obtain the probability that each pixel belongs to a fluid inclusion pixel and the probability that each pixel belongs to a quartz sand particle pixel.
5. The method according to claim 3, characterized in that: The classifying each pixel point according to the classification probability includes: For each pixel point in the image to be detected, when the probability that the pixel point belongs to a fluid inclusion pixel point is greater than the probability that the pixel point belongs to a quartz sand particle pixel point, the pixel point is determined to be a fluid inclusion pixel point; When the probability that the pixel point belongs to the fluid inclusion pixel point is less than the probability that the pixel point belongs to the quartz sand particle pixel point, the pixel point is determined to be a quartz sand particle pixel point.
6. The method according to claim 1, characterized in that Determining the pixel ratio of the fluid inclusion to the quartz sand particles from the segmentation result includes: Determine a first total number of pixels of the fluid inclusion pixels and a second total number of pixels of the quartz sand particles from the segmentation results; The pixel ratio of the quartz sand fluid inclusion is determined according to the first total number of pixels and the second total number of pixels.
7. The method according to claim 1, characterized in that The image segmentation processing of the image to be detected is implemented by calling an image segmentation model, and the training method of the image segmentation model includes: Obtaining an image sample to be detected of quartz sand particles, wherein the image sample to be detected is marked with a real label; Inputting the image sample to be detected into the image segmentation model for forward propagation to obtain a sample prediction result; The difference between the sample prediction result and the true label is used as the loss value of the image segmentation model, and the loss value is used to back-propagate in the image segmentation model to update the parameters of the image segmentation model.