Rock debris white light picture classification method, system, equipment, medium and program
Through the white light image classification method of rock debris based on semi-supervised learning, the key features of rock debris are extracted using deep learning network technology, and the problem of inefficient classification of rock debris type is solved, efficient and accurate judgment of rock debris type is achieved, and detection is improved Rate and work efficiency.
Patent Information
- Application Number
- CN202410999797.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-15
AI Technical Summary
The inefficient classification of rock cutting types leads to slow detection speed, and relying on manual judgment is prone to misjudgment. Especially for new or inexperienced technicians, it is difficult to quickly and accurately complete rock cutting classification work.
Using the method of classification of rock cutting white light picture based on semi-supervised learning, the data set is constructed and the classification network model is trained, and the KL divergence is used as a loss function for iterative training to generate the final classification. Network model for judging the cutting type.
The classification work efficiency and detection rate of rock cutting types are significantly improved, manual intervention is reduced, the model's adaptability to unknown or new types of rock cuttings is enhanced, and the classification accuracy and working efficiency is improved.
Smart Images

Figure CN118968140A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of oil / gas detection, and in particular relates to a method, system, device, medium and program for classifying white light images of rock debris based on semi-supervised learning. Background Art
[0002] The diversity of rock chip samples is obvious. From hard quartz sandstone to soft mudstone, from fossil-rich limestone to mineral-rich igneous rock, the rock chips are of various types and shapes. In addition, the rock chips may be worn, contaminated or mixed during the ascent process, causing the original clear features to become blurred or difficult to identify. These factors greatly increase the difficulty of accurately identifying the rock chip type.
[0003] The professionalism and proficiency of technicians have a crucial impact on the analysis results. Even experienced technicians may make misjudgments due to fatigue, distraction or personal judgment when faced with complex and changeable rock cuttings samples. For new or inexperienced technicians, it is easy to fall into confusion and find it difficult to complete the rock cuttings classification work quickly and accurately.
[0004] In the field of oil and gas exploration, when a large amount of rock cuttings are obtained from oil wells, it is often necessary to rely on visual observation to determine the type of the rock cuttings. Since there are many impurities in the rock cuttings, this work is very challenging for the technical skills of the technicians. Because the professionalism and proficiency of different technicians vary, this work is often time-consuming and labor-intensive and cannot guarantee a high accuracy rate. Summary of the invention
[0005] Aiming at the problem that the classification efficiency of rock cuttings types in the prior art is low, resulting in slow detection rate, the present invention provides a rock cuttings white light image classification method based on semi-supervised learning, which can effectively help detection technicians to judge the rock cuttings types, improve the classification efficiency of rock cuttings types, and increase the detection rate.
[0006] In order to achieve the above object, the present invention provides the following technical solutions.
[0007] In a first aspect, the present invention provides a method for classifying white light images of rock debris based on semi-supervised learning, comprising:
[0008] Process the obtained rock cuttings white light image data and construct a data set;
[0009] Use the dataset to train a rock debris white light image classification network based on semi-supervised learning to obtain a classification network model;
[0010] Design a loss function to iteratively train the classification network model to obtain the final classification network model;
[0011] The collected Baiyan pictures are analyzed based on the final classification network model, and classification results are generated according to the analysis results.
[0012] As a further improvement of the present invention, the step of processing the obtained rock cuttings white light image data to construct a data set includes:
[0013] The rock cuttings white light images are preprocessed to obtain rock cuttings white light image data and image type labels, and then the data set is obtained.
[0014] As a further improvement of the present invention, a dataset is used to train a rock debris white light image classification network based on semi-supervised learning to obtain a classification network model, including:
[0015] The data set is input into the rock debris white light image classification network based on semi-supervised learning to extract features and form different rock debris types;
[0016] Classification training is performed for different types of rock cuttings to obtain a classification network model.
[0017] As a further improvement of the present invention, the designed loss function performs iterative training on the classification network model to obtain the final classification network model, including:
[0018] Input the data set into the backbone network of the classification network model for feature extraction, and design the loss function based on the classification results;
[0019] The KL divergence is used as the loss function to train the network, and then the parameters of the classification network are updated in the direction to complete the iterative training of the classification network model to obtain the final classification network model.
[0020] As a further improvement of the present invention, the step of inputting the data set into the backbone network in the classification network model for feature extraction and designing a loss function according to the classification result includes:
[0021] After the features are extracted, the second feature extraction is performed through ResNet and MoblieNet in the backbone network, and then the difference between ResNet and MoblieNet is calculated through KL divergence.
[0022] As a further improvement of the present invention, the difference between ResNet and MoblieNet is calculated by KL divergence, which is specifically expressed as follows:
[0023]
[0024] Where p(x) is the probability distribution of the large network, q(x) is the probability distribution of the small network, i = [0, N], and N is the total number of samples.
[0025] In a second aspect, the present invention provides a rock debris white light image classification system based on semi-supervised learning, comprising:
[0026] Model building module: used to process the obtained rock cuttings white light image data and build a data set;
[0027] Classification model module: used to use the data set to train the rock debris white light image classification network based on semi-supervised learning to obtain the classification network model;
[0028] Final network model module: used to design the loss function to iteratively train the classification network model to obtain the final classification network model;
[0029] Classification result module: used to analyze the collected Baiyan pictures based on the final classification network model and generate classification results according to the analysis results.
[0030] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of a method for classifying white light images of rock debris based on semi-supervised learning when executing the computer program.
[0031] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of a method for classifying white-light images of rock debris based on semi-supervised learning.
[0032] In a fifth aspect, the present invention provides a computer program product, characterized in that it includes computer instructions, which, when executed by a processor, implement the steps of the rock debris white light image classification method based on semi-supervised learning.
[0033] Compared with the prior art, the present invention has the following beneficial effects:
[0034] The present invention can automatically extract key features from a large number of white light images of rock debris through deep learning network technology, and these features are essential for distinguishing different types of rock debris. Compared with the traditional manual classification method, it not only greatly improves the classification efficiency, but also significantly improves the accuracy of classification through the optimization of machine learning algorithms. At the same time, while ensuring the classification accuracy, the complexity and calculation amount of the model are reduced, and the classification speed is further improved. Secondly, the traditional supervised learning method requires a large amount of labeled data, and in practical applications, it is often costly and time-consuming to obtain high-quality, labeled rock debris image data. The present invention adopts a semi-supervised learning method, which can make full use of unlabeled or mislabeled image data, and automatically learn the inherent structure and laws in the data through the algorithm, thereby improving the generalization ability and classification performance of the model without significantly increasing the data annotation cost. Semi-supervised learning combines the advantages of small sample learning and unsupervised learning, so that the model can learn more diverse feature representations during the training process, and enhances the model's adaptability to unknown or new types of rock debris. This feature is particularly important for complex and changeable detection environments, and helps to improve the robustness of the overall detection process. The present invention greatly reduces manual intervention, reduces labor costs, and improves work efficiency through automated and intelligent classification methods. The efficient generation of rock cuttings classification results can quickly provide accurate geological information to geological prospectors, helping them make decisions faster and adjust exploration strategies, thereby improving the efficiency and success rate of the entire exploration process. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] The drawings described herein are for illustrative purposes only and are not intended to limit the scope of the present disclosure in any way. In the drawings:
[0036] Figure 1 A schematic diagram of a process of a rock debris white light image classification method based on semi-supervised learning according to the present invention;
[0037] Figure 2 A schematic diagram of a specific process of a rock debris white light image classification method based on semi-supervised learning according to the present invention;
[0038] Figure 3 A schematic diagram of a network model of a rock debris white light image classification method based on semi-supervised learning according to the present invention;
[0039] Figure 4 A schematic diagram of the analysis process of a rock debris white light image classification method based on semi-supervised learning according to the present invention;
[0040] Figure 5 It is a structural schematic diagram of a rock debris white light image classification system based on semi-supervised learning of the present invention;
[0041] Figure 6Schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION
[0042] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. The described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.
[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.
[0044] Aiming at the problem that the classification efficiency of rock cuttings types in the prior art is low, resulting in a slow detection rate, the present invention provides a rock cuttings white light image classification method based on semi-supervised learning, such as Figure 1 As shown, the method includes:
[0045] Process the obtained rock cuttings white light image data and construct a data set;
[0046] Use the dataset to train a rock debris white light image classification network based on semi-supervised learning to obtain a classification network model;
[0047] Design a loss function to iteratively train the classification network model to obtain the final classification network model;
[0048] The collected Baiyan pictures are analyzed based on the final classification network model, and classification results are generated according to the analysis results.
[0049] The present invention can effectively help detection technicians to judge the type of rock cuttings, improve the efficiency of rock cutting type classification, and increase the detection rate.
[0050] The present invention is further explained below in conjunction with specific drawings.
[0051] like Figure 2 As shown, the present invention provides a rock debris white light image classification technology based on semi-supervised learning, including:
[0052] Collect and organize rock debris white light image datasets;
[0053] The white light images of rock cuttings are preprocessed to meet the requirements of subsequent semi-supervised training.
[0054] Use semi-supervised learning to train the classification network to improve data utilization;
[0055] Using semi-supervised learning to train classification networks can make more efficient use of data. For example, unlabeled or mislabeled data can also be used to train the network, which can greatly improve data utilization.
[0056] Input the image data into the backbone network for feature extraction, and obtain the classification results to calculate the loss;
[0057] The backbone network uses ResNet and MoblieNet to train both large and small networks at the same time. The small network is trained at the same time and the large network is distilled to make the network model more accurate. This can reduce the number of parameters in the classification of the network model and improve efficiency.
[0058] When calculating the loss, in order to efficiently utilize unlabeled data, the idea of mask matrix is adopted when designing the loss function, which can effectively solve the impact of unlabeled data on the network during cross entropy calculation.
[0059] The training method used when training large and small networks simultaneously is to use KL divergence as the loss function to calculate the difference between ResNet and MoblieNet. The formula is:
[0060]
[0061] Where p(x) is the probability distribution of the large network, q(x) is the probability distribution of the small network, i = [0, N], and N is the total number of samples.
[0062] Using KL divergence as the loss function to train the network can more effectively utilize unlabeled or mislabeled data for the small network to distill the features of the large network. This can ensure the efficiency of the small network while learning the characteristics of the large network's feature extraction, thereby improving the efficiency of the network.
[0063] The features are extracted through two networks for the second time and then the loss of each network is calculated by KL divergence.
[0064] Back propagate the network to update parameters and complete the network training.
[0065] Adopt image hierarchical training and perform hierarchical training on different standards to improve the accuracy of network classification;
[0066] For different rock cuttings type classification standards, a hierarchical training method is used to train the network more specifically for each different classification, thereby improving the accuracy of the network;
[0067] Use the trained network model to classify rock cuttings.
[0068] Input the white light image of rock cuttings of corresponding specifications to obtain its classification results, and analyze and generate the classification results of the network through the result analysis program.
[0069] The present invention is described in detail below with reference to specific embodiments:
[0070] The present invention provides a rock debris white light image classification technology based on semi-supervised learning, comprising the following steps:
[0071] S1: Collect and organize rock debris white light image data.
[0072] Specifically, the present invention collects white light images of rock cuttings and their type labels, organizes them, and uses them in the subsequent semi-supervised training process.
[0073] S2: Process the collected image data to build a dataset.
[0074] Specifically, the previously collected rock cuttings white light data are preprocessed, and corresponding vectors are created for their labels. Because the present invention is based on semi-supervised learning technology, unlabeled and wrong labels need to be specially marked for subsequent network training.
[0075] S3: Training a network for classifying white light images of rock debris using semi-supervised learning.
[0076] Specifically, the network model diagram is as follows Figure 3 As shown, the backbone network of the present invention uses ResNet and MobileNet to train two networks at the same time, inputs the processed image data into the backbone network to obtain the extracted feature values of the two networks, passes the feature values of the two networks through the two networks respectively, obtains the predicted value, and finally uses KL divergence as the loss of the large network and the small network to calculate the loss. The loss calculated by the KL divergence loss function is used to distill the small network MobileNet to obtain the characteristics of the large network ResNet, thereby ensuring that the network model has the operating efficiency of the small model while improving its accuracy. At the same time, the present invention also designs a hierarchical training strategy for the network according to the characteristics of specific rock cuttings. Specifically, it is to layer them according to different types of labels, and then each layer of the network is trained for a corresponding type of label, thereby improving the prediction accuracy of the network.
[0077] S4: Design the loss function based on the idea of mask matrix.
[0078] Specifically, in order to efficiently utilize image data, the present invention adopts semi-supervised learning, and can simultaneously use unlabeled data to train the network. When designing the loss function, the prediction result of the network for unlabeled images will affect the loss calculation process. Therefore, the present invention adopts the idea of mask matrix when designing the loss function. When reading data, the label will be converted to the corresponding class number, such as N classes, which are numbered from 0 to N-1. Therefore, when merging files, the label of unlabeled data can be recorded as None, and then when converting to class numbers, first check whether the label of the data is in the class label. If it is, it is normally converted to the class number, otherwise it is encoded as the number of classes N. Then, when calculating the loss, you only need to generate a mask matrix and process the incoming class number modulo N. The method for generating the mask matrix is as follows: you only need to determine whether the class number is less than N, if it is less than N, record it as 1, otherwise record it as 0. Modulo N processing can make the label of unlabeled data become 0, so cross entropy calculation can also be performed. After calculating the loss of each sample, it is multiplied by the mask matrix. The contribution of unlabeled data to the cross entropy loss is 0, thus ensuring that the final classification loss only calculates the unlabeled data.
[0079] S5: Perform training based on the above dataset and model improvements, and perform verification and inference.
[0080] Specifically, if Figure 4 As shown, based on the above collected and collated data sets, a specially designed loss function is used to iterate the training of the improved model, and finally a classification network model is obtained. When in use, the white light image of the cuttings can be inferred based on the model to obtain the classification result of the cuttings.
[0081] S6: According to the classification results of the above network, the classification results are input into the analysis program to analyze the types of rock cuttings.
[0082] The description of the above embodiments is only used to help understand the core idea of the present invention; at the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the idea of the present invention. In summary, the content of this specification should not be understood as limiting the present invention.
[0083] In summary, the invention provides a rock debris white light image classification technology based on semi-supervised learning, and combines deep learning network technology to extract and classify the white light of rock debris. The invention uses small network distillation technology to improve the network training and classification efficiency, and uses semi-supervised learning methods to improve data utilization, without wasting unlabeled or wrongly labeled image data. The invention can greatly improve the efficiency of rock debris type classification, improve detection rate, and have a very positive impact on the entire detection process.
[0084] The second object of the present invention is to propose a rock debris white light image classification system based on semi-supervised learning, such as Figure 5 As shown, including:
[0085] Model building module: used to process the obtained rock cuttings white light image data and build a data set;
[0086] Classification model module: used to use the data set to train the rock debris white light image classification network based on semi-supervised learning to obtain the classification network model;
[0087] Final network model module: used to design the loss function to iteratively train the classification network model to obtain the final classification network model;
[0088] Classification result module: used to analyze the collected Baiyan pictures based on the final classification network model and generate classification results according to the analysis results.
[0089] like Figure 6 As shown, the third object of the present invention is to provide an electronic device, the electronic device comprising: a processor, a memory and a display screen. The memory and the display screen are both connected to the processor, such as through a bus. Optionally, the electronic device may also include a transceiver. It should be noted that in actual applications, the transceiver is not limited to one, and the structure of the electronic device does not constitute a limitation on the embodiments of the present application.
[0090] The processor may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.
[0091] The bus may include a path to transmit information between the above components. The bus may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc.
[0092] The memory can be ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or it can be EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to this.
[0093] The memory is used to store the application code for executing the solution of the present application, and the execution is controlled by the processor. The processor is used to execute the application code stored in the memory to implement the content shown in the above method embodiment.
[0094] Figure 6 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.
[0095] The fourth object of the present invention is to provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program is stored thereon, and when the program is executed by a processor, the aforementioned Figure 1 The various processes of the method embodiment shown in the figure include, for example, a memory including instructions, and the instructions can be executed by a processor of an electronic device to complete the method.
[0096] A computer-readable storage medium may be a tangible device that holds and stores instructions used by an instruction execution device. A computer-readable storage medium may be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, a computer-readable storage medium may be a portable computer disk, a hard disk, a USB flash drive, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a podium random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, an optical disk, a magnetic disk, a mechanical encoding device, and any combination thereof.
[0097] A fifth object of the present invention is to provide a computer program product, comprising computer instructions, which, when executed by a processor, implement the above Figure 1 The various processes of the method embodiment shown can achieve the same technical effect, and will not be described again here to avoid repetition.
[0098] Many embodiments and many applications beyond the examples provided will be apparent to those skilled in the art upon reading the above description. Therefore, the scope of the present teachings should not be determined with reference to the above description, but rather with reference to the preceding claims and the full scope of equivalents to which such claims are entitled. For the purpose of comprehensiveness, all articles and references, including disclosures of patent applications and publications, are incorporated herein by reference. The omission of any aspect of the subject matter disclosed herein in the preceding claims is not intended to be a disclaimer of such subject matter, nor should it be considered that the applicants did not consider such subject matter to be part of the disclosed inventive subject matter.
[0099] The above content is a further detailed description of the present invention. It cannot be determined that the specific implementation methods of the present invention are limited to these. For ordinary technicians in the technical field to which the present invention belongs, without departing from the concept of the present invention, they can also make several simple deductions or substitutions, which should be regarded as belonging to the protection scope of the present invention determined by the submitted claims.
Claims
1. A method for classifying white light images of rock debris based on semi-supervised learning, characterized in that: include: Process the obtained rock cuttings white light image data and construct a data set; Use the dataset to train a rock debris white light image classification network based on semi-supervised learning to obtain a classification network model; Design a loss function to iteratively train the classification network model to obtain the final classification network model; The collected Baiyan pictures are analyzed based on the final classification network model, and classification results are generated according to the analysis results.
2. The method for classifying rock debris white light images based on semi-supervised learning according to claim 1, characterized in that: The step of processing the obtained rock cuttings white light image data and constructing a data set includes: The rock cuttings white light images are preprocessed to obtain rock cuttings white light image data and image type labels, and then the data set is obtained.
3. The method for classifying rock debris white light images based on semi-supervised learning according to claim 1, characterized in that: The dataset is used to train a rock debris white light image classification network based on semi-supervised learning, and a classification network model is obtained, including: The data set is input into the rock debris white light image classification network based on semi-supervised learning to extract features and form different rock debris types; Classification training is performed for different types of rock cuttings to obtain a classification network model.
4. The method for classifying rock debris white light images based on semi-supervised learning according to claim 1, characterized in that: The designed loss function iteratively trains the classification network model to obtain the final classification network model, including: Input the data set into the backbone network of the classification network model for feature extraction, and design the loss function based on the classification results; The KL divergence is used as the loss function to train the network, and then the parameters of the classification network are updated in the direction to complete the iterative training of the classification network model to obtain the final classification network model.
5. The method for classifying white light images of rock debris based on semi-supervised learning according to claim 4, characterized in that: The data set is input into the backbone network in the classification network model for feature extraction, and a loss function is designed according to the classification result, including: After the features are extracted, the second feature extraction is performed through ResNet and MoblieNet in the backbone network, and then the difference between ResNet and MoblieNet is calculated through KL divergence.
6. The method for classifying rock debris white light images based on semi-supervised learning according to claim 5, characterized in that: The difference between ResNet and MoblieNet is calculated by KL divergence, which is specifically expressed as follows: Where p(x) is the probability distribution of the large network, q(x) is the probability distribution of the small network, i = [0, N], and N is the total number of samples.
7. A rock debris white light image classification system based on semi-supervised learning, characterized in that: include: Model building module: used to process the obtained rock cuttings white light image data and build a data set; Classification model module: used to use the data set to train the rock debris white light image classification network based on semi-supervised learning to obtain the classification network model; Final network model module: used to design the loss function to iteratively train the classification network model to obtain the final classification network model; Classification result module: used to analyze the collected Baiyan pictures based on the final classification network model and generate classification results according to the analysis results.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of a method for classifying white light images of rock debris based on semi-supervised learning as described in any one of claims 1 to 6 when executing the computer program.
9. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of a method for classifying white-light images of rock debris based on semi-supervised learning as claimed in any one of claims 1 to 6.
10. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the steps of a method for classifying white light images of rock debris based on semi-supervised learning as described in any one of claims 1 to 6.
Citation Information
Patent Citations
Semi-supervised three-dimensional point cloud target detection method
CN113536920A
Radio frequency fine feature information extraction method and system based on semi-supervised meta learning
CN113705646A
Rock sample intelligent identification method and system based on image deep learning
CN114581666A
Rock debris type identification method and device, electronic equipment and storage medium
CN114863187A
Model training method, category prediction method and device
CN116824288A
