Deep learning driven method and system for identifying particles in insulating oil

Abstract

The invention relates to a method and a system for identifying particles in insulating oil driven by deep learning, which can realize automatic identification and classification of particles in insulating oil by combining a micro-flow digital imaging technology and a semantic segmentation method of deep learning, constructing a unified training data set and combining a multi-task segmentation model. Images of the particles in the transformer oil are obtained through the micro-flow digital imaging module, and pixel-level classification and particle recognition are carried out on the images through the multi-task segmentation model. The segmentation branches in the model are responsible for extraction of particle boundaries, and the classification branches are used for analyzing particle morphology and outputting particle categories. The model adopts multi-scale feature fusion, an attention mechanism and a deep residual network structure, and the problems of particle morphology diversity, size change, image boundary blur and the like are solved. Through the high-precision particle recognition algorithm and system, the accuracy and efficiency of detection of the particles in the insulating oil are effectively improved, manual intervention is reduced, and the safety and long-term stability of the transformer are improved.

Description

Technical Field

[0001] The present invention belongs to the field of transformers, and specifically relates to a deep learning-driven method and system for identifying particles in insulating oil. Background Art

[0002] With the continuous development of power systems, transformers, as key equipment in power systems, are particularly important for safety and stability. Transformer oil not only serves as the insulating medium of power transformers, but also bears the functions of heat dissipation and protection. Its performance plays a decisive role in transformers. However, during the operation of transformers, they are inevitably contaminated by impurities. When the concentration of particulate matter reaches a certain level, it may affect the dielectric strength of the oil, reduce the insulation performance of the transformer, and even cause partial discharge or equipment failure. Therefore, timely detection and identification of particulate matter in transformer oil has become a research hotspot in the power industry.

[0003] At present, the detection methods of transformer oil particles mainly include optical image analysis and chemical analysis, etc. Although these traditional detection methods can provide certain detection effects in certain specific application scenarios, they usually have problems such as poor detection accuracy, inaccurate recognition effect, and great human influence.

[0004] With the rise of deep learning technology, especially the rapid development of convolutional neural networks (CNN) and semantic segmentation technology, new opportunities have been brought to the field of particle detection. Semantic segmentation technology can classify images at the pixel level, greatly improving the accuracy of particle recognition. Deep learning methods can automatically extract the morphological features of particles through a large amount of training data and powerful feature learning capabilities, thereby achieving more accurate particle recognition.

[0005] However, despite the significant progress made by deep learning technology in the field of image processing, existing methods still face several challenges. For example, the morphology of particles in transformer oil is complex and changeable, the particle size varies greatly, and the boundaries of particle images are blurred. These problems make it difficult for existing deep learning models to achieve efficient and accurate particle recognition in practical applications. When processing particle images, existing models often rely on relatively simple feature extraction methods and lack the ability to capture subtle differences between particles. Therefore, further innovation and optimization are still needed.

[0006] In addition, although some existing studies have attempted to combine micro-flow digital imaging technology with deep learning for particle detection, due to the diversity and complexity of transformer oil particles, existing technologies still cannot meet the needs of real-time detection, accurate classification, and efficient processing. Therefore, how to propose a high-precision and high-efficiency transformer oil particle recognition technology based on micro-flow digital imaging technology combined with a deep learning model is still a hot topic and difficulty in current research.

[0007] Therefore, the present invention proposes a new particle recognition scheme by combining a multi-task segmentation model of deep learning, targeting the diversity and size changes of particle images, aiming to solve the problems of poor particle detection accuracy, inaccurate recognition effect, and great human influence in the prior art, and provide a more intelligent and efficient technical solution for the safe operation of transformers. Summary of the invention

[0008] In order to solve the deficiencies in the prior art, the present invention provides a deep learning-driven method for identifying particles in insulating oil, which solves the problems of the lack of a unified and effective training data set for particles in insulating oil, blurred image segmentation boundaries, diverse impurity particle shapes, variable sizes, and poor recognition accuracy. The present invention combines microflow digital imaging technology with semantic segmentation technology based on deep learning, and proposes a method suitable for transformer oil particle detection. By constructing a unified and effective training data set for particles in oil, and using semantic segmentation technology to perform pixel-level classification on microflow imaging images, automatic recognition and classification of particulate matter is achieved. This method not only overcomes the limitations of traditional technologies in dealing with the diversity of particle shapes, size changes, and blurred segmentation edges, but also improves the accuracy and efficiency of detection, and reduces reliance on manual operations. This technical solution provides a more intelligent and efficient solution for online monitoring and fault warning of power equipment, which helps to improve the long-term stability and safety of transformers.

[0009] The present invention adopts the following technical solution.

[0010] The present invention provides a method for identifying particles in insulating oil driven by deep learning, comprising:

[0011] Acquire an image to be detected and input it into a pre-trained multi-task segmentation model; the multi-task segmentation model outputs an oil particle image detection result based on the image to be detected; the multi-task segmentation model is trained through the following steps: acquire an oil particle image and construct a multi-task segmentation data set; pre-process the multi-task segmentation data set; use a dual-branch framework of a segmentation branch and a classification branch as a basic structure to build a multi-task segmentation model; the segmentation branch is used to segment particles in the oil and obtain boundary information of the particles; the classification branch is used for identification, and morphological analysis is performed based on the information obtained in the segmentation branch, and then identification is performed in combination with the image extracted from the classification branch to obtain the type of particles in the oil; the constructed multi-task segmentation model is trained using the pre-processed multi-task segmentation data set; set network training parameters, and use the loss function to train the constructed multi-task segmentation model.

[0012] Preferably, the step of acquiring the particle images in oil and constructing a multi-task segmentation dataset comprises:

[0013] Images of particles in transformer oil are acquired through a microfluidic digital imaging module; synthetic particle data and difficult samples are added to form the final multi-task segmentation dataset.

[0014] Preferably, the preprocessing of the multi-task segmentation dataset includes data enhancement of images in the multi-task segmentation dataset, and the data enhancement includes horizontal flipping, vertical flipping, rotation, translation, scaling, cropping, erasing and color dithering operations.

[0015] Preferably, the multi-task segmentation model built by training the preprocessed multi-task segmentation dataset includes:

[0016] In the segmentation branch, the image features output by the skeleton network are used to guide segmentation feature learning, and the image features include high-order features and low-order features; in the classification branch, the high-order features input by the segmentation branch are used as auxiliary discriminant information to classify the confusion items.

[0017] Preferably, the skeleton network is an optimized Xception deep residual network, and the Xception includes stage one, stage two, stage three, stage four and stage five.

[0018] Preferably, the extracting the output features of the skeleton network in the segmentation branch includes:

[0019] After the high-order features output by the stage 4 are sent to the multi-scale context comparison local module, the local features output by the stage 4 and the low-order features output by the stage 1, stage 2, and stage 3 are respectively sent to the double-layer attention convolution GRU.

[0020] Preferably, the multi-scale contextual comparison local module is processed by connecting the local features fed into the stage four with four 2×2 convolutions with dilated convolution rates of 2, 3, 6 and 8 respectively, and then the processed local features are subtracted and fused and connected to the input end of the channel attention module.

[0021] Preferably, the channel attention module uses global average pooling and two fully connected layers to calculate channel weights.

[0022] The present invention also provides a deep learning driven system for identifying particles in insulating oil, the system being a system used in the aforementioned deep learning driven method for identifying particles in insulating oil, comprising:

[0023] The microfluidic digital imaging module is used to obtain the image to be detected and input it into the pre-trained multi-task segmentation model;

[0024] An image recognition module, used for the multi-task segmentation model to output a particle image recognition result based on the image to be detected;

[0025] Multi-scale contextual comparison local module, used to enhance the capture and optimization of image boundaries;

[0026] Channel attention module, used to mine the intrinsic semantic relationship between multi-scale feature maps;

[0027] The pyramid pooling module is used to perform pooling operations of four different scales on the input feature map to generate a multi-scale feature map.

[0028] The present invention also provides a terminal, including a processor and a storage medium;

[0029] The storage medium is used to store instructions;

[0030] The processor is configured to operate according to the instructions to execute the steps according to the aforementioned method.

[0031] The present invention also provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of the above method are implemented.

[0032] The beneficial effect of the present invention is that, compared with the prior art,

[0033] 1. Improve particle recognition accuracy: The present invention introduces a multi-task segmentation model, combines the segmentation branch and the classification branch, and utilizes the effective fusion of high-order features and low-order features to accurately segment and classify particles according to their morphology. Compared with the prior art, the present invention significantly improves the accuracy of particle recognition, especially when the particles have complex morphology and variable sizes. Through the comprehensive application of the global attention upsampling module, the mixed pool module, and the feature pyramid attention module, the present invention can better capture local details and global information, ensuring high-precision recognition and classification of particles.

[0034] 2. Solve the problem of blurred image boundaries: The present invention further enhances the capture and optimization of image boundaries by adopting a multi-scale contextual contrast local module and a double-layer attention convolution GRU (Gated Recurrent Unit). In traditional methods, it is often difficult to accurately segment particles due to their complex morphology and blurred boundaries. By combining the fusion of global features, local features, and multi-scale features, the present invention can effectively improve the clarity and accuracy of image boundaries, especially when the particle boundaries are blurred, it can significantly improve the segmentation effect and generate more accurate particle segmentation results.

[0035] 3. Enhanced processing of inter-class similarities and intra-class differences: The present invention effectively solves the confusion problem caused by the similarity of particle morphology by combining global context information with multi-scale fusion of local features. This method enhances the ability to distinguish different types of particles, especially when the particles are similar in morphology and vary greatly in size, accurate classification and segmentation can still be achieved. In addition, the introduction of the pyramid pooling module further optimizes the model's ability to handle particles with variable morphology and large size differences, ensuring high-precision classification and segmentation of particles.

[0036] 4. Construct a unified and effective training dataset for particles in oil: The present invention significantly improves the accuracy and robustness of particle recognition by constructing a unified and effective training dataset for particles in oil. The dataset is composed of real particle images obtained by microfluidic digital imaging technology, and the diversity of the dataset is expanded by synthetic data and difficult samples. Combined with data enhancement technology, the model can adapt to changes in various particle morphologies and sizes. Through standardized and unified dataset design, the problems of data scarcity and low recognition accuracy in the prior art are solved, and the model's ability to distinguish between similarities between classes and differences within classes is significantly improved, thereby ensuring more accurate particle recognition and segmentation effects.

[0037] 5. Optimize training and loss calculation: This invention introduces the Balanced Segmentation Loss (BSL), which not only focuses on the overall segmentation accuracy, but also finely optimizes the segmentation boundaries. The loss function combines the cross entropy loss and the intersection over union loss (IoU loss), further improving the model's sensitivity to particle boundaries and effectively reducing mis-segmentation and misclassification. This optimization of the loss function enables the model to more accurately identify particle shapes and boundaries during training, improving the accuracy and stability of segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is an overall flow chart of a method and system for identifying particles in insulating oil driven by deep learning provided by the present invention;

[0039] Figure 2 It is a network framework diagram of a method and system for identifying particles in insulating oil driven by deep learning provided by the present invention;

[0040] Figure 3 It is a schematic diagram of the attention convolution GRU in a deep learning driven method for identifying particles in insulating oil and a system provided by the present invention;

[0041] Figure 4 It is a schematic diagram of a local module of multi-scale context comparison in a method for identifying particles in insulating oil driven by deep learning and a system provided by the present invention;

[0042] Figure 5 It is a schematic diagram of a channel attention module in a deep learning driven method and system for identifying particles in insulating oil provided by the present invention;

[0043] Figure 6 It is a schematic diagram of a pyramid pooling module in a deep learning-driven method for identifying particles in insulating oil and a system provided by the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the scope of protection of the present invention.

[0045] like Figure 1 As shown, in a first aspect, the present invention provides a method for identifying particles in insulating oil driven by deep learning, comprising:

[0046] Obtain the image to be detected and input it into the pre-trained multi-task segmentation model;

[0047] The multi-task segmentation model outputs an oil particle image detection result based on the image to be detected;

[0048] The multi-task segmentation model is trained by the following steps:

[0049] Obtain images of particles in oil and build a multi-task segmentation dataset;

[0050] Preprocess the multi-task segmentation dataset;

[0051] A multi-task segmentation model is built using a dual-branch framework of segmentation branch and classification branch as the basic structure;

[0052] The segmentation branch is used to segment particles in the oil and obtain the boundary information of the particles; the classification branch is used for identification. Morphological analysis is performed based on the information obtained from the segmentation branch, and then combined with the image extracted from the classification branch for identification to obtain the type of particles in the oil.

[0053] Use the preprocessed multi-task segmentation dataset to train the built multi-task segmentation model;

[0054] Set network training parameters and use loss function to train the built multi-task segmentation model.

[0055] In combination with the first aspect, further, the step of acquiring the oil particle image and constructing the multi-task segmentation dataset includes:

[0056] Acquire images of particles in transformer oil using a microfluidic digital imaging module;

[0057] Add synthetic granular data and difficult samples to form the final multi-task segmentation dataset.

[0058] In combination with the first aspect, further, the preprocessing of the multi-task segmentation dataset includes data enhancement of images in the multi-task segmentation dataset, and the data enhancement of images in the multi-task segmentation dataset includes horizontal flipping, vertical flipping, rotation, translation, scaling, cropping, erasing and color dithering operations on the images in the multi-task segmentation dataset.

[0059] In combination with the first aspect, further, the multi-task segmentation model built by training using the preprocessed multi-task segmentation dataset includes:

[0060] In the segmentation branch, the image features output by the skeleton network are used to guide the segmentation feature learning, and the image features include high-order features and low-order features; in the classification branch, the input high-order features are used as auxiliary discrimination information to classify and solve the problem of confusion items. The classification task is completed through the network framework of the segmentation and classification dual branches, and finally the identification of particles in insulating oil is realized.

[0061] In combination with the first aspect, further, the skeleton network is an optimized Xception deep residual network, and the Xception includes stage one, stage two, stage three, stage four and stage five; the improvement of the Xception deep residual network includes: replacing the ordinary convolutions of the five stages of the network with void convolutions, thereby enhancing the network's processing capabilities for multi-scale features and improving the network's performance when processing particles of different sizes.

[0062] In combination with the first aspect, further, extracting output features of the skeleton network in the segmentation branch includes:

[0063] like Figure 4 As shown in the figure, the high-order features output from the fourth stage are sent to the multi-scale contextual contrast local module for processing. The local module extracts features through four 2×2 convolutions with dilated convolution rates of 2, 3, 6 and 8 respectively. After feature processing, the local features and the low-order features output from the first, second and third stages are sent to the double-layer attention convolution GRU respectively. This operation effectively combines feature information of different scales, enhances the ability to capture the morphological details of particles, and especially significantly improves the segmentation accuracy of tiny particles.

[0064] like Figure 3As shown in the figure, the two-layer attention GRU network replaces the traditional fully connected layer in this task, and processes the generated tensors through the convolution layer, thereby reducing the number of model parameters and improving the computational efficiency of the model. At the same time, the introduction of GRU enables the model to effectively process the temporal relationship between features and further optimize the decision-making process of the model using the attention mechanism. By adding the attention mechanism to GRU, the model can dynamically adjust the degree of attention to features at each scale and improve segmentation accuracy.

[0065] like Figure 5 As shown in the figure, the attention convolution GRU can help the model to mine the intrinsic semantic relationship between multi-scale feature maps by introducing a channel attention module. Specifically, the attention mechanism limits the influence of the model on specific feature maps by introducing a gating structure, and can quickly adjust the model parameters according to the changes in the current target scene, thereby improving the accuracy and robustness of segmentation.

[0066] Combined with the first aspect, further, after receiving the output of the double-layer attention convolution GRU, the channel attention module adds it to the features output by the GRU, and then upsamples it to the original image size and feeds it into the 1×1 convolution to integrate the features. The channel attention module uses global average pooling and two fully connected layers to calculate the channel weights, and then selectively enhances effective features and suppresses invalid features to achieve the purpose of feature recalibration. The global average pooling operation effectively extracts the spatial information of the features and ensures the spatial consistency of the features. At the same time, the global maximum pooling retains the most significant information in the features. These operations help to enhance the representativeness and discriminability of the features.

[0067] like Figure 6 As shown, in combination with the first aspect, further, the classification branch processes the output features of the Xception network stages one, two, three, four, and five through a pyramid pooling module, extracts the global information of the image, and fuses it with the results of the segmentation branch, and finally outputs the classification result. The pyramid pooling module generates a multi-scale feature map by performing pooling operations on the input feature map at four different scales, and then adjusts the number of channels through 1×1 convolution, and upsamples it to the original size to complete feature fusion. Through this multi-scale pooling and feature fusion method, the network can accurately classify particles from multiple scales of perspective and enhance the global consistency of the segmentation results.

[0068] In combination with the first aspect, further, setting network training parameters and using the loss function to train the constructed semantic segmentation model includes:

[0069] Use the images in the preprocessed semantic segmentation image dataset as input images and unify the size of the input images;

[0070] The semantic segmentation network model is trained using a stochastic gradient descent algorithm with momentum, wherein the weight initialization method in the semantic segmentation network model is Kaiming and the activation function is ReLu;

[0071] The loss function is used to predict the difference between the actual data, and the loss function is a balanced segmentation loss function (BSL).

[0072] The formula of the loss function is:

[0073] BSL=αL CE +βL IoU

[0074] Among them, L CE is the cross entropy loss, L IoU is the IoU loss, α and β are hyperparameters that control the weights of the two parts, with a value range of 0-1, and usually need to be tuned through cross-validation.

[0075] The cross entropy loss function formula is:

[0076]

[0077] Where N is the pixel value, C is the number of categories, and y i,c is the true label (0 or 1) of the i-th pixel belonging to category c, P i,c is the predicted probability that the ith pixel belongs to category c.

[0078] The formula for the intersection-over-union loss function is:

[0079]

[0080] Among them, A∪B is the intersection of the predicted area and the true area, and A∪B is the union of the predicted area and the true area.

[0081] like Figure 2 As shown, in the second aspect, further, the present invention provides a deep learning driven particle identification system in insulating oil, comprising:

[0082] The microfluidic digital imaging module is used to obtain the image to be detected and input it into the pre-trained multi-task segmentation model;

[0083] An image recognition module, used for the multi-task segmentation model to output a particle image recognition result based on the image to be detected;

[0084] Multi-scale contextual comparison local module, used to enhance the capture and optimization of image boundaries;

[0085] Channel attention module, used to mine the intrinsic semantic relationship between multi-scale feature maps;

[0086] The pyramid pooling module is used to perform pooling operations of four different scales on the input feature map to generate a multi-scale feature map.

[0087] The multi-task segmentation model is trained by the following steps:

[0088] Obtain particle images and build a multi-task segmentation dataset;

[0089] Preprocess the multi-task segmentation dataset;

[0090] A multi-task segmentation model is built using a dual-branch framework of segmentation branch and classification branch as the basic structure;

[0091] Use the preprocessed multi-task segmentation dataset to train the built multi-task segmentation model;

[0092] Set network training parameters and use loss function to train the built multi-task segmentation model.

[0093] In a third aspect, the present invention provides a deep learning driven particle identification system in insulating oil, comprising a processor and a storage medium;

[0094] The storage medium is used to store instructions;

[0095] The processor is used to operate according to the instructions to execute the steps of any method described in the first aspect.

[0096] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the methods described in the first aspect.

[0097] The beneficial effect of the present invention is that, compared with the prior art,

[0098] 1. Improve particle recognition accuracy: The present invention introduces a multi-task segmentation model, combines the segmentation branch and the classification branch, and utilizes the effective fusion of high-order features and low-order features to accurately segment and classify particles according to their morphology. Compared with the prior art, the present invention significantly improves the accuracy of particle recognition, especially when the particles have complex morphology and variable sizes. Through the comprehensive application of the global attention upsampling module, the mixed pool module, and the feature pyramid attention module, the present invention can better capture local details and global information, ensuring high-precision recognition and classification of particles.

[0099] 2. Solve the problem of blurred image boundaries: The present invention further enhances the capture and optimization of image boundaries by adopting a multi-scale contextual contrast local module and a double-layer attention convolution GRU (Gated Recurrent Unit). In traditional methods, it is often difficult to accurately segment particles due to their complex morphology and blurred boundaries. By combining the fusion of global features, local features, and multi-scale features, the present invention can effectively improve the clarity and accuracy of image boundaries, especially when the particle boundaries are blurred, it can significantly improve the segmentation effect and generate more accurate particle segmentation results.

[0100] 3. Enhanced processing of inter-class similarities and intra-class differences: The present invention effectively solves the confusion problem caused by the similarity of particle morphology by combining global context information with multi-scale fusion of local features. This method enhances the ability to distinguish different types of particles, especially when the particles are similar in morphology and vary greatly in size, accurate classification and segmentation can still be achieved. In addition, the introduction of the pyramid pooling module further optimizes the model's ability to handle particles with variable morphology and large size differences, ensuring high-precision classification and segmentation of particles.

[0101] 4. Construct a unified and effective training dataset for particles in oil: The present invention significantly improves the accuracy and robustness of particle recognition by constructing a unified and effective training dataset for particles in oil. The dataset is composed of real particle images obtained by microfluidic digital imaging technology, and the diversity of the dataset is expanded by synthetic data and difficult samples. Combined with data enhancement technology, the model can adapt to changes in various particle morphologies and sizes. Through standardized and unified dataset design, the problems of data scarcity and low recognition accuracy in the prior art are solved, and the model's ability to distinguish between similarities between classes and differences within classes is significantly improved, thereby ensuring more accurate particle recognition and segmentation effects.

[0102] 5. Optimize training and loss calculation: This invention introduces the Balanced Segmentation Loss (BSL), which not only focuses on the overall segmentation accuracy, but also finely optimizes the segmentation boundaries. The loss function combines the cross entropy loss and the intersection over union loss (IoU loss), further improving the model's sensitivity to particle boundaries and effectively reducing mis-segmentation and misclassification. This optimization of the loss function enables the model to more accurately identify particle shapes and boundaries during training, improving the accuracy and stability of segmentation.

[0103] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0104] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but 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 suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0105] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0106] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A deep learning driven method for identifying particles in insulating oil, comprising: The image to be detected is obtained and input into a pre-trained multi-task segmentation model; the multi-task segmentation model outputs the detection result of the oil particle image based on the image to be detected; the multi-task segmentation model is trained by the following steps: obtaining the oil particle image and constructing a multi-task segmentation data set; preprocessing the multi-task segmentation data set; using the dual-branch framework of the segmentation branch and the classification branch as the basic structure to build a multi-task segmentation model; the segmentation branch is used to segment the particles in the oil and obtain the boundary information of the particles; the classification branch is used for identification, and morphological analysis is performed according to the information obtained in the segmentation branch, and then combined with the image extracted from the classification branch for identification to obtain the type of particles in the oil; Use the preprocessed multi-task segmentation dataset to train the built multi-task segmentation model; set the network training parameters and use the loss function to train the built multi-task segmentation model.

2. The method for identifying particles in insulating oil driven by deep learning according to claim 1, characterized in that: The step of acquiring the particle images in oil and constructing a multi-task segmentation dataset includes: Images of particles in transformer oil are acquired through a microfluidic digital imaging module; synthetic particle data and difficult samples are added to form the final multi-task segmentation dataset.

3. The method for identifying particles in insulating oil driven by deep learning according to claim 2, characterized in that: The preprocessing of the multi-task segmentation dataset includes data enhancement of images in the multi-task segmentation dataset, and the data enhancement includes horizontal flipping, vertical flipping, rotation, translation, scaling, cropping, erasing and color jittering operations.

4. The method for identifying particles in insulating oil driven by deep learning according to claim 3 is characterized in that: The multi-task segmentation model built by training the preprocessed multi-task segmentation dataset includes: In the segmentation branch, the image features output by the skeleton network are used to guide segmentation feature learning, and the image features include high-order features and low-order features; in the classification branch, the high-order features input by the segmentation branch are used as auxiliary discriminant information to classify the confusion items.

5. The method for identifying particles in insulating oil driven by deep learning according to claim 4, characterized in that: The skeleton network is an optimized Xception deep residual network, and the Xception includes stage one, stage two, stage three, stage four and stage five.

6. The method for identifying particles in insulating oil driven by deep learning according to claim 5, characterized in that: The step of extracting the output features of the skeleton network in the segmentation branch includes: After the high-order features output by the stage 4 are sent to the multi-scale context comparison local module, the local features output by the stage 4 and the low-order features output by the stage 1, stage 2, and stage 3 are respectively sent to the double-layer attention convolution GRU.

7. The method for identifying particles in insulating oil driven by deep learning according to claim 6, characterized in that: The multi-scale contextual comparison local module processes the local features fed into the fourth stage by connecting them with four 2×2 convolutions with dilated convolution rates of 2, 3, 6 and 8 respectively, and then subtracts and fuses the processed local features and connects them to the input end of the channel attention module.

8. The method for identifying particles in insulating oil driven by deep learning according to claim 7, characterized in that: The channel attention module uses global average pooling and two fully connected layers to calculate channel weights.

9. A deep learning driven system for identifying particles in insulating oil, the system being a system used in a deep learning driven method for identifying particles in insulating oil as claimed in any one of claims 1 to 8, comprising: The microfluidic digital imaging module is used to obtain the image to be detected and input it into the pre-trained multi-task segmentation model; An image recognition module, used for the multi-task segmentation model to output a particle image recognition result based on the image to be detected; Multi-scale contextual comparison local module, used to enhance the capture and optimization of image boundaries; Channel attention module, used to mine the intrinsic semantic relationship between multi-scale feature maps; The pyramid pooling module is used to perform pooling operations of four different scales on the input feature map to generate a multi-scale feature map.

10. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images