Mobile phone visible light and infrared transformer bushing image detection defect method and system

Abstract

The application provides a mobile phone visible light and infrared transformer bushing image detection defect method and system, comprising the following steps: S1: obtaining infrared and visible light images of a transformer bushing based on a mobile phone multi-camera and / or a data transmission interface; S2: performing fusion processing on the infrared and visible light images on the mobile phone side; and S3: performing fault diagnosis based on a trained defect detection neural network by using the fusion-processed images on the mobile phone side to obtain a diagnosis result. The application retains a large amount of important information in the source images, can compress the information, reduces the operation amount, is beneficial to real-time processing of the algorithm, and the extracted feature information is directly related to decision analysis, the final image fusion result can maximize the feature information required for decision analysis, thereby improving the target detection capability of the system and being more beneficial to decision of the system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power equipment detection, in particular to a transformer bushing image detection defect method and system based on mobile phone visible light and infrared, and more particularly to an infrared and visible light image fusion and defect diagnosis method for detecting transformer bushing based on mobile phone multi-camera. BACKGROUND

[0002] As the main equipment for voltage conversion and power transmission in the power system, the stable and reliable operation of the power transformer directly affects the safety and stability of the power grid. As an important part of the transformer, the bushing bears the function of fixing the lead wire and ensuring its insulation from the outside world, and is also a frequent fault position. In actual operation, the bushing often causes transformer shutdown and even explosion and other serious accidents due to defects such as damaged insulation or local overheating. Therefore, fault diagnosis and evaluation of the transformer bushing are particularly important. In order to ensure its healthy and stable working state, preventive tests are often needed under the condition of transformer shutdown, which greatly reduces the reliability and timeliness of the equipment.

[0003] Common problems of transformer bushing include lack of oil in the bushing, poor contact of the conductive loop connecting piece, etc., which causes abnormal temperature rise at the joint. The heating and oxidation of the bushing connecting end are the basis for judging whether the transformer is abnormal. The thermal imager can timely find out whether the bushing is in good working condition, timely find out the problem, improve the work efficiency, and avoid accidents.

[0004] At present, the application of live detection technology is relatively mature and widespread. The infrared temperature measurement technology is often used to detect the oil shortage fault and local overheating fault of the transformer bushing due to its non-contact measurement, high precision and immunity to electromagnetic interference. Due to the poor adaptability of human extraction of descriptive features, the judgment of the equipment operating state is too subjective, and the given conclusion is fuzzy and incomplete. The detection personnel often misjudge the fault of the bushing. At the same time, the storage capacity of unstructured image data is huge, and manual inspection and identification not only affects the accuracy of the judgment result, but also leads to low efficiency, thereby affecting the degree of power equipment monitoring automation.

[0005] Patent document CN115761428A (application number: 202211525616.1) discloses a method for fusing infrared and visible light images of a UHV converter transformer bushing, first obtaining infrared light images and visible light images of the same bushing; preprocessing the infrared light images and the visible light images; image segmentation is performed on the preprocessed infrared light images and visible light images; the segmented infrared light images and visible light images are subjected to a refinement process to form a corresponding weight map; the preprocessed infrared light images and visible light images are subjected to brightness extraction; the preprocessed infrared light images and visible light images are subjected to single-scale weighting with the corresponding weight map to form a pre-fusion image; the pre-fusion image and the brightness layer after brightness extraction are added to obtain a final fusion image.

[0006] Patent document CN113344475B (application number: 202110894419.6) discloses a transformer bushing defect identification method based on sequence modal decomposition, relating to the technical field of transformer bushings, which uses an image fusion algorithm to fuse the obtained infrared images of the transformer bushing and visible light images of the transformer bushing; a Mask-RCNN algorithm is used to segment the transformer bushing fusion image; a CEEMDAN method is used to adaptively decompose nonlinear signals; the eliminated intrinsic mode functions are reconstructed; an LSTM training model is used to model the training set data and optimize the network parameters; the operating power parameter is input into the LSTM training model to obtain the temperature value for discrimination; the hot spot temperature and the relative temperature difference are used to make a transformer bushing defect determination.

[0007] Patent document CN116342952A (application number: 202310318777.1) discloses a transformer bushing anomaly identification method and system, first obtaining the original infrared images of each phase bushing of the transformer; preprocessing the original infrared images to obtain infrared pictures of each phase transformer bushing; simulating the fault types of the transformer bushing to obtain simulation images of the transformer bushing; converting the simulation images into simulation infrared images under the corresponding fault types; inputting the simulation infrared images and the bushing infrared images as training samples into a pre-set convolutional neural network to generate an online anomaly identification model of the transformer bushing; using the online anomaly identification model to identify the fault type of the current transformer bushing.

[0008] Patent document CN113344475B (application number: 202110894419.6) discloses a transformer bushing defect identification method based on sequence modal decomposition, relating to the technical field of transformer bushing, which adopts an image fusion algorithm to fuse the obtained transformer bushing infrared image and transformer bushing visible light image; adopts a Mask-RCNN algorithm to segment the transformer bushing fusion image; adopts a CEEMDAN method to adaptively decompose nonlinear signals; reconstructs the eliminated intrinsic modal function; models the training set data using an LSTM training model, optimizes the network parameters; inputs the operating power parameters into the LSTM training model to obtain the temperature value for discrimination; and makes a transformer bushing defect determination based on the hotspot temperature and relative temperature difference.

[0009] Patent document CN113313013B (application number: 202110580695.5) discloses a transformer bushing target detection method based on infrared image processing technology, including the following steps: collecting the infrared image of the transformer bushing in the substation; identifying and extracting the transformer bushing in the infrared image using a multi-template multi-matching method; extracting the ROI region to segment the image of the area where the bushing is located for detection; filtering and denoising; performing semantic segmentation on the transformer bushing after image preprocessing, removing the background using the OTSU algorithm, extracting the edge using the Canny operator, removing the non-connected region using the morphological operation of the image, and segmenting the bushing base and umbrella skirt part using histogram binarization; performing fine segmentation on each part of the transformer bushing to generate a temperature measurement reference baseline.

[0010] Patent document CN111488868B (application number: 202010227133.8) discloses a high-temperature area identification method and system based on transformer infrared images, including obtaining the temperature interval value and colorimetric value in the image; obtaining the temperature colorimetric value at high temperature using the obtained colorimetric value of the temperature interval; obtaining all colorimetric values of the image and establishing a colorimetric distribution chart; using image colorimetric and LOG operators to divide the image edges to obtain the identified high-temperature area, which combines computer batch processing, saves time and cost compared to manual identification, and has high accuracy, allowing the high-temperature temperature interval to be set according to demand, and providing great help for transformer external temperature monitoring.

[0011] Patent document CN110378424A (application number: 201910664061.0) discloses a transformer bushing fault infrared image recognition method based on a convolutional neural network, comprising the following steps: A, constructing a convolutional neural network model; the convolutional neural network model includes an input layer and an output layer, a plurality of convolutional layers and sampling layers are arranged between the input layer and the output layer, the convolutional layers and the sampling layers are alternately arranged, and a fully connected layer is arranged between the output layer and the last sampling layer; B, training the convolutional neural network model; the excitation propagation and weight update are performed in a loop until the target function converges to the preset range; C, using the convolutional neural network model trained in step B to recognize the transformer bushing fault infrared image.

[0012] Patent document CN111798405A (application number: 201911145289.5) discloses a method for diagnosing faults of transformer bushing infrared images using deep learning, comprising a bushing extraction module, a fault area extraction module, and a fault diagnosis module, which realizes the function of diagnosing faults of input bushing infrared images. The bushing extraction module includes an offline training part and an online extraction part, the fault area extraction module includes an SLIC algorithm preprocessing part and a PCNN extraction part, and the fault diagnosis module includes a fault feature vector extraction and fault classification part. The detection algorithm is based on deep learning, which trains the labeled bushing image to extract the bushing image in a complex background, then extracts the fault and its feature vector, and uses the diagnostic algorithm to achieve the purpose of fault diagnosis, thereby realizing the infrared image fault detection of the bushing and effectively ensuring the safe and stable operation of the bushing.

[0013] In the prior art, the transformer bushing image acquisition uses an infrared and visible light camera in the form of a ball machine or a gun machine structure, which has a large device volume, and the image processing still needs additional computer processing equipment, making it inconvenient for users to use. The current smart phones are basically equipped with a visible light image camera, and have a data output interface. The data output interface and the external infrared camera can be used to form a double-image acquisition device for transformer bushing detection. Although there are many infrared and visible light image fusion algorithms, due to the limited processing speed of the mobile phone, it is difficult to perform image processing based on the mobile phone system, and there is usually no infrared camera with temperature reading on the mobile phone, so it is difficult to match the external camera data with the self-camera image, and therefore the transformer bushing image fusion and diagnosis method capable of running on the mobile phone still needs to be developed.

[0014] In view of the above problems, the present application provides an infrared and visible light image fusion and defect diagnosis method for detecting transformer bushing based on a mobile phone multi-camera. SUMMARY

[0015] Aiming at the defects in the prior art, the present application aims to provide a mobile phone visible light and infrared transformer bushing image defect detection method and system.

[0016] According to the present application, a mobile phone visible light and infrared transformer bushing image defect detection method is provided, comprising:

[0017] Step S1: based on the mobile phone multi-camera and / or data transmission interface, the infrared and visible light images of the transformer bushing are obtained;

[0018] Step S2: the mobile phone fuses the infrared and visible light images;

[0019] Step S3: the mobile phone uses the fused image to perform fault diagnosis based on the trained defect detection neural network to obtain a diagnosis result.

[0020] Preferably, the step S1 uses: when the mobile phone has a built-in temperature detection high-definition infrared camera and a visible light camera, the infrared and visible light images of the transformer bushing are obtained based on the mobile phone camera; when the mobile phone only has a visible light camera, the infrared image of an external high-definition infrared camera with temperature detection is read through the data transmission interface of the mobile phone.

[0021] Preferably, the step S2 uses:

[0022] Step S2.1: the visible light and infrared images are respectively feature-extracted to obtain features including edges, shapes, contours and local features;

[0023] Step S2.2: the features including edges, shapes, contours and local features are registered to obtain a first feature registration;

[0024] Step S2.3: according to the infrared feature size and the registration area size, the size of the visible light image is adjusted, the visible light image features are extracted again, and the visible light image features are registered with the infrared image features again to generate a second feature registration, to obtain an image with consistent resolution of the visible light and infrared images in the feature area, and then perform pixel fusion.

[0025] Preferably, the step S2.1 uses: the visible light and infrared images are feature-extracted through the SIFT, SURF, BRIEF, FAST, BRISEK and FREAK detection algorithms in the OpenCV library function.

[0026] Preferably, the defect detection neural network comprises a convolutional neural network, a support vector machine and an LSTM neural network.

[0027] Preferably, the infrared image and the visible light image collected by the mobile phone are sent to a preset image processor based on a mobile phone wireless network, the preset image processor performs image fusion analysis on the infrared image and the visible light image, judges the state of the transformer bushing comprehensively, and feeds back the state of the transformer bushing to the mobile phone end through the wireless network after diagnosis; when the state of the transformer bushing obtained by using the preset image processor is the same as the diagnosis result of the mobile phone end, the preset image processor and the mobile phone end are mutually verified; when the state of the transformer bushing obtained by using the preset image processor is different from the diagnosis result of the mobile phone end, further judgment is performed.

[0028] According to the mobile phone visible light and infrared transformer bushing image defect detection system provided by the application, the infrared image and the visible light image of the transformer bushing are obtained based on the mobile phone multi-camera and / or data transmission interface.

[0029] Module M1: obtaining the infrared and visible light images of the transformer bushing based on the mobile phone multi-camera and / or data transmission interface;

[0030] Module M2: the mobile phone end performs fusion processing on the infrared and visible light images;

[0031] Module M3: the mobile phone end performs fault diagnosis based on the trained defect detection neural network to obtain a diagnosis result based on the fusion-processed images.

[0032] Preferably, when the mobile phone is equipped with a high-definition infrared camera and a visible light camera with temperature detection, the infrared and visible light images of the transformer bushing are obtained based on the mobile phone camera; when the mobile phone is only equipped with a visible light camera, the infrared image of the external high-definition infrared camera with temperature detection is read through the data transmission interface of the mobile phone.

[0033] Preferably, the module M2 adopts:

[0034] Module M2.1: respectively extracting features of the visible light and infrared images, and obtaining edge, shape, contour and local features;

[0035] Module M2.2: registering the edge, shape, contour and local features to obtain a first feature registration;

[0036] Module M2.3: adjusting the size of the visible light image according to the infrared feature size and the registration area size, extracting the features of the visible light image again, registering the features of the visible light image and the infrared image again to generate a second feature registration, obtaining an image with consistent resolution of the visible light and infrared images in the feature area, and then performing pixel fusion;

[0037] The module M2.1 adopts: the SIFT, SURF, BRIEF, FAST, BRISEK and FREAK detection algorithms in the OpenCV library function are used to extract features of the visible light and infrared images.

[0038] Preferably, the infrared image and the visible light image collected by the mobile phone are sent to a preset image processor based on a mobile phone wireless network, the preset image processor performs image fusion analysis on the infrared image and the visible light image, comprehensively judges the state of the transformer bushing, and feeds back the state of the transformer bushing to the mobile phone end through the wireless network after diagnosis; when the state of the transformer bushing obtained by using the preset image processor is the same as the diagnosis result of the mobile phone end, the preset image processor and the mobile phone end are mutually verified; when the state of the transformer bushing obtained by using the preset image processor is different from the diagnosis result of the mobile phone end, further judgment is performed.

[0039] Compared with the prior art, the present application has the following beneficial effects:

[0040] 1. Although computers can use complex and advanced image processing algorithms, computers are not as convenient as mobile phones for actual on-site detection, and since mobile phones have limited performance, the image processing algorithms used by mobile phones are still in the exploratory stage, and the image processing algorithm of the present application can meet the on-site detection requirements;

[0041] 2. The present application integrates a high-definition infrared camera and a visible light camera in a mobile phone, develops APP software based on the mobile phone system to collect infrared images and visible light images, and facilitates users to test on site;

[0042] 3. The present application can fuse infrared images and visible light images through a mobile phone to find defects in transformer bushings, facilitate on-site testing, give real-time diagnosis references, and can timely find problems;

[0043] 4. The present application provides two detection modes, which can be performed simultaneously through local and remote diagnosis, to improve diagnosis accuracy and timeliness. BRIEF DESCRIPTION OF DRAWINGS

[0044] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments with reference to the attached drawings:

[0045] Figure 1 A flow chart of an infrared and visible light image fusion and defect diagnosis method for detecting transformer bushings based on a mobile phone multi-camera. DETAILED DESCRIPTION

[0046] The present application will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any form. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present application. These all belong to the protection scope of the present application.

[0047] Example 1

[0048] In order to overcome the deficiencies of the prior art, an infrared and visible light image fusion and defect diagnosis method and system based on a mobile phone multi-camera detection transformer bushing are provided, infrared and visible light images based on a mobile phone multi-camera detection transformer bushing are fused and processed by the mobile phone, a defect feature database is established through sample learning, the classifier is trained, the trained neural network is used for fault diagnosis, and the fault type is generated; based on wireless network transmission, the image is transmitted to a special image processor for fault diagnosis, and then feedback is given to the mobile phone for display.

[0049] Although there are many infrared and visible light image fusion algorithms, due to the limited operation speed of the mobile phone, the transformer bushing image fusion and diagnosis method capable of running on the mobile phone still needs to be developed. Since the visible light image size of the transformer bushing collected by the mobile phone is large and the infrared image size is small, two kinds of image feature extraction, registration and fusion are required, the infrared image is embedded into the reduced visible light image, and various features are displayed in a total image. The present application retains a large amount of important information in the source image, and can also compress the information, reduce the operation amount, be beneficial to real-time processing of the algorithm, and the extracted feature information is directly related to decision analysis, and the final image fusion result can maximize the feature information required for decision analysis, thereby improving the target detection capability of the system and being more beneficial to the decision of the system. The transformer bushing diagnosis can directly use the convolutional neural network classification algorithm on the mobile phone, a defect feature database is established through sample learning, the convolutional neural network classifier is trained, the trained convolutional neural network is used for transformer bushing fault diagnosis, and the fault type is generated; in addition, the transformer bushing diagnosis can also use remote defect diagnosis, the collected infrared image and visible light image are sent to a special image processor through a wireless network, the special image processor is processed, and then the fault type is fed back to the mobile phone through the wireless network.

[0050] According to the infrared and visible light image fusion and defect diagnosis method based on a mobile phone multi-camera detection transformer bushing provided by the present application, as shown in Figure 1 The infrared and visible light image fusion and defect diagnosis method based on a mobile phone multi-camera detection transformer bushing provided by the present application, as shown in

[0051] Specifically, the mobile phone reads the infrared image and the visible light image. When the mobile phone is equipped with a high-definition infrared camera and a visible light camera with temperature detection, the infrared image and the visible light image of the self camera are directly read. If the mobile phone is only equipped with a visible light camera, the infrared image of the external high-definition infrared camera with temperature detection is read through the data transmission interface of the mobile phone.

[0052] Specifically, the mobile phone processes the infrared image and the visible light image. The visible light image of the transformer bushing collected by the mobile phone has a large size, and the infrared image has a small size. First, the features of the visible light image and the infrared image are extracted respectively to obtain multiple information such as edges, shapes, contours, and local features. Then, the features are matched to obtain feature matching 1. According to the infrared feature size and the matching area size, the size of the visible light image is adjusted. Then, the features of the visible light image are extracted again, and are matched with the infrared image features to generate feature matching 2. An image with consistent resolution of the visible light and the infrared image in the feature area is obtained, and then pixel fusion is performed.

[0053] The feature extraction algorithm of the image can be a detection algorithm such as SIFT, SURF, BRIEF, FAST, BRISEK, and FREAK in the OpenCV library function.

[0054] After feature extraction, the image matching can be performed based on the brute force matcher or the FLANN matcher (fast library for approximate nearest neighbors) in the OpenCV library function.

[0055] The image fusion is performed according to the feature positioning to determine the position of the infrared image in the visible light image, and then a linear fusion method is used to fuse the infrared image into the visible light image.

[0056] Specifically, the transformer bushing diagnosis includes: the mobile phone image processing and the defect diagnosis and the remote defect diagnosis can be freely switched by selecting the button.

[0057] In the transformer bushing mobile phone diagnosis, a trained convolutional neural network is used for transformer bushing fault diagnosis. In addition, support vector machines, LSTM neural networks, etc. can also be used. Based on the convolutional neural network classification algorithm, a defect feature database is established by sample learning, and the convolutional neural network classifier is trained. The training can be performed by the mobile phone or by the computer network training. Then, the corresponding parameters are assigned to the convolutional neural network on the mobile phone. The convolutional neural network classifier should include the possible fault types.

[0058] The transformer bushing adopts remote defect diagnosis, the collected infrared image and visible light image are sent to a special image processor based on mobile phone wireless network, the special image processor can analyze the infrared image and visible light image respectively, then image fusion is carried out for analysis, the state of the transformer bushing is judged in all directions, and the fault type is fed back to the mobile phone through the wireless network after diagnosis.

[0059] Further requirements, the transformer bushing diagnosis, characterized in that: the mobile phone itself image processing defect diagnosis and remote defect diagnosis method are consistent, the generated fault type is consistent, when the mobile phone itself processing diagnosis and remote diagnosis method are inconsistent, the results of the two can be used as a reference for the tester to make a judgment.

[0060] According to the infrared and visible light image fusion and defect diagnosis system for detecting transformer bushing based on mobile phone multi-camera provided by the application, the infrared image and visible light image are directly processed by the mobile phone, such as image feature extraction, image registration and image fusion; the transformer bushing diagnosis can directly use convolutional neural network classification algorithm on the mobile phone, a defect feature database is established through sample learning, the convolutional neural network classifier is trained, the trained convolutional neural network is used for transformer bushing fault diagnosis, and a fault type is generated; in addition, the transformer bushing diagnosis can also use remote defect diagnosis, the collected infrared image and visible light image are sent to a special image processor through wireless network, and the fault type is fed back to the mobile phone through wireless network after processing of the special image processor.

[0061] Specifically, the mobile phone reads the infrared image and visible light image, when the mobile phone has a temperature detection high-definition infrared camera and a visible light camera, the infrared image and visible light image of the self camera are directly read; if the mobile phone only has a visible light camera, the infrared image of an external high-definition infrared camera with temperature detection is read through the data transmission interface of the mobile phone.

[0062] Specifically, the mobile phone processes the infrared image and visible light image, the visible light image size of the transformer bushing collected by the mobile phone is large, and the infrared image size is small, first, the visible light image and the infrared image are respectively subjected to feature extraction, a plurality of information such as edge, shape, contour and local feature is obtained, then the features are registered to obtain feature registration 1, according to the infrared feature size and the registration area size, then the size of the visible light image is adjusted, the visible light image features are extracted again, the visible light image features are registered with the infrared image features again to generate feature registration 2, the image with consistent resolution of visible light and infrared image in the feature area is obtained, and then pixel fusion is carried out.

[0063] Wherein, the feature extraction algorithm of the image can be through the SIFT, SURF, BRIEF, FAST, BRISEK and FREAK detection algorithms in the OpenCV library function.

[0064] Wherein, the image registration can be through the feature extraction, and the feature matching is performed based on the brute force matcher in the OpenCV library function or based on the FLANN matcher-fast approximate nearest neighbor library.

[0065] Wherein, the image fusion is to determine the position of the infrared image in the visible light image according to the positioning of the features, and then the linear fusion method is used to fuse the infrared image into the visible light image.

[0066] Specifically, the transformer bushing diagnosis includes: the mobile phone image processing and defect diagnosis and the remote defect diagnosis can be freely switched by selecting the button.

[0067] Wherein, the transformer bushing mobile phone diagnosis adopts the trained convolutional neural network for transformer bushing fault diagnosis, and in addition, support vector machines, LSTM neural networks, etc. can be used. Among them, based on the convolutional neural network classification algorithm, the defect feature database is established by sample learning, and the convolutional neural network classifier is trained. The training can be performed through the mobile phone, or the network training can be performed through the computer. Then the corresponding parameters are assigned to the convolutional neural network on the mobile phone. The convolutional neural network classifier should include the possible fault types.

[0068] Wherein, the transformer bushing adopts the remote defect diagnosis, based on the mobile phone wireless network, the collected infrared image and visible light image are sent to the special image processor. The special image processor can analyze the infrared image and visible light image respectively, and then perform image fusion for analysis, and judge the state of the transformer bushing in all directions. After diagnosis, the fault type is fed back to the mobile phone through the wireless network.

[0069] Further requirements, the transformer bushing diagnosis, characterized in that: the mobile phone itself image processing defect diagnosis and the remote defect diagnosis method are consistent, the generated fault type is consistent, when the mobile phone itself processing diagnosis and the remote diagnosis method are inconsistent, the results of the two can be used as a reference for the tester to make a judgment.

[0070] Example 2

[0071] Embodiment 2 is a preferred example of embodiment 1

[0072] According to the application, a kind of infrared and visible light image fusion and defect diagnosis method for detecting transformer bushing based on mobile phone multi camera is provided.Install the APP software described in the application on Android smart phone, data transmission interface is connected with the camera of Aiwei mobile phone thermal imager series, then run software to read the image of smart phone self camera and external infrared camera, then carry out image processing;Based on the trained convolutional neural network classifier, obtain fault diagnosis type.

[0073] Those skilled in the art know that, in addition to implementing the system provided by the application and each device, module, unit thereof in a pure computer readable program code manner, the same function can be realized by logically programming method steps in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers and embedded microcontrollers, etc.The system provided by the application and each device, module, unit thereof can be considered as a hardware component, and the devices, modules and units included therein for realizing various functions can also be considered as structures within the hardware component;The devices, modules and units for realizing various functions can also be considered as both software modules for realizing methods and structures within the hardware component.

[0074] The specific embodiments of the application are described above.It needs to be understood that the application is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essential content of the application.In the case of no conflict, the embodiments of the application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for detecting defects in a transformer bushing image of a mobile phone visible light and infrared, characterized by, The method comprises the steps of: Step S1: obtaining infrared and visible light images of the transformer bushing based on the multiple cameras and / or data transmission interface of the mobile phone; Step S2: performing fusion processing on the infrared and visible light images on the mobile phone side; Step S3: performing fault diagnosis based on the trained defect detection neural network using the fused images on the mobile phone side to obtain a diagnosis result; In the step S1, when the mobile phone is equipped with a high-definition infrared camera and a visible light camera for temperature detection, the infrared and visible light images of the transformer bushing are obtained based on the cameras of the mobile phone; when the mobile phone is only equipped with a visible light camera, the infrared image of an external high-definition infrared camera with temperature detection is read through the data transmission interface of the mobile phone; In the step S2, the following steps are performed: Step S2.1: feature extraction is performed on the visible light and infrared images respectively to obtain edges, shapes, contours and local features; Step S2.2: registration is performed on the edges, shapes, contours and local features to obtain a first feature registration; Step S2.3: the size of the visible light image is adjusted according to the infrared feature size and the registration area size, the feature of the visible light image is extracted again, the visible light image feature is registered with the infrared image feature again to generate a second feature registration, an image with consistent resolution of the visible light and infrared images in the feature area is obtained, and then pixel fusion is performed; The infrared image and the visible light image collected by the mobile phone are sent to a preset image processor based on the wireless network of the mobile phone, the image processor performs image fusion analysis on the infrared image and the visible light image, judges the state of the transformer bushing in all directions, and feeds back the state of the transformer bushing to the mobile phone side through the wireless network after diagnosis; when the state of the transformer bushing obtained by the preset image processor is the same as the diagnosis result of the mobile phone side, the preset image processor and the mobile phone side are mutually verified; when the state of the transformer bushing obtained by the preset image processor is different from the diagnosis result of the mobile phone side, further judgment is performed.

2. The method of claim 1, wherein the method is performed by a mobile phone visible light and infrared transformer bushing image defect detection method. In the step S2.1, the SIFT, SURF, BRIEF, FAST, BRISEK and FREAK detection algorithms in the OpenCV library function are used for feature extraction on the visible light and infrared images.

3. The method of claim 1, wherein the method is performed by a mobile phone visible light and infrared transformer bushing image defect detection method. The defect detection neural network comprises a convolutional neural network, a support vector machine and an LSTM neural network.

4. A mobile phone visible light and infrared transformer bushing image defect detection system, characterized in that, The method comprises the steps of: Module M1: obtaining infrared and visible light images of the transformer bushing based on the multiple cameras and / or data transmission interface of the mobile phone; Module M2: performing fusion processing on the infrared and visible light images on the mobile phone side; Module M3: performing fault diagnosis based on the trained defect detection neural network using the fused images on the mobile phone side to obtain a diagnosis result; In the module M1, when the mobile phone is equipped with a high-definition infrared camera and a visible light camera for temperature detection, the infrared and visible light images of the transformer bushing are obtained based on the cameras of the mobile phone; when the mobile phone is only equipped with a visible light camera, the infrared image of an external high-definition infrared camera with temperature detection is read through the data transmission interface of the mobile phone; In the module M2, the following steps are performed: Module M2.1: feature extraction is performed on the visible light and infrared images respectively to obtain edges, shapes, contours and local features; Module M2.2: the first feature registration is obtained by registering the edge, shape, contour and local feature; Module M2.3: according to the infrared feature size and the registration area size, the size of the visible light image is adjusted, the visible light image feature is extracted again, the infrared image feature is registered again to generate the second feature registration, the image with consistent resolution of the visible light and infrared image in the feature area is obtained, and then pixel fusion is performed; The module M2.1 adopts: the visible light and infrared image is subjected to feature extraction by using the detection algorithms including SIFT, SURF, BRIEF, FAST, BRISEK and FREAK in the OpenCV library function; The infrared image and the visible light image collected by the mobile phone are sent to a preset image processor based on the mobile phone wireless network, the image fusion analysis is performed on the infrared image and the visible light image by the preset image processor, the state of the variable voltage sleeve is comprehensively judged, the variable voltage sleeve state is fed back to the mobile phone end after diagnosis through the wireless network; when the variable voltage sleeve state obtained by using the preset image processor is the same as the diagnosis result of the mobile phone end, the preset image processor and the mobile phone end are mutually verified; when the variable voltage sleeve state obtained by using the preset image processor is different from the diagnosis result of the mobile phone end, further judgment is performed.

Images