CN113378818A

CN113378818A - Electrical equipment defect determining method and device, electronic equipment and storage medium

Info

Publication number: CN113378818A
Application number: CN202110688346.5A
Authority: CN
Inventors: 全晓方; 郑奇凯; 孙上元; 陈何成; 李顺; 姚日斌; 黄繁朝; 刘彬; 杨海亮; 杨武志
Original assignee: Liuzhou Bureau of Extra High Voltage Power Transmission Co
Current assignee: Liuzhou Bureau of Extra High Voltage Power Transmission Co
Priority date: 2021-06-21
Filing date: 2021-06-21
Publication date: 2021-09-10

Abstract

The embodiment of the application relates to the technical field of equipment detection, and discloses a method and a device for determining defects of electrical equipment, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring a first infrared image of the electrical equipment to be identified; acquiring temperature information of the electrical equipment to be identified according to the first infrared image; determining a temperature abnormal area in the first infrared image according to the temperature information; extracting the characteristics of the temperature abnormal region through a trained defect determination model, generating a plurality of first masks corresponding to the temperature abnormal region according to the characteristics, and determining a first mask coefficient of a defective component in the temperature abnormal region; generating a second mask of the temperature abnormal region according to the plurality of first masks and the first mask coefficients through a defect determining model; and determining the image position of the defective part of the electrical equipment to be identified in the first infrared image according to the second mask through the defect determination model. The method can accurately position the defective part and improve the identification accuracy of the defective part of the electrical equipment.

Description

Electrical equipment defect determining method and device, electronic equipment and storage medium

Technical Field

The present disclosure relates to the field of device detection technologies, and in particular, to a method and an apparatus for determining defects of an electrical device, an electronic device, and a storage medium.

Background

The electric equipment in the power system runs in a high-voltage environment for a long time, and the electric equipment inevitably has defects of abnormal temperature, structural damage and the like under the action of long-term voltage, heat generation, mechanical stress and environmental factors, so that the defects of the electric equipment are timely discovered and eliminated, and the safety of the electric equipment and the power system is favorably improved.

At present, an infrared image is usually applied in an electric power system to judge whether the temperature of the electrical equipment is abnormal, but the infrared image can only reflect the area of the electrical equipment with abnormal temperature, and the identification accuracy of defective components causing the temperature abnormality in the electrical equipment is low.

Disclosure of Invention

The embodiment of the application discloses a method and a device for determining the defect of electrical equipment, electronic equipment and a storage medium, which can accurately identify the defective component of the electrical equipment and improve the identification accuracy of the defective component in the electrical equipment.

The first aspect of the embodiment of the present application discloses a method for determining defects of electrical equipment, including:

acquiring a first infrared image of the electrical equipment to be identified;

acquiring temperature information of the electrical equipment to be identified according to the first infrared image;

determining a temperature abnormal area in the first infrared image according to the temperature information;

extracting the characteristics of the temperature abnormal region through a trained defect determination model, generating a plurality of first masks corresponding to the temperature abnormal region according to the characteristics, and determining first mask coefficients of defective components in the temperature abnormal region;

generating a second mask of the temperature abnormal region according to the plurality of first masks and the first mask coefficients through the defect determination model;

and determining the image position of the defective component of the electrical equipment to be identified in the first infrared image according to the second mask through the defect determination model.

As an optional implementation manner, in the first aspect of the embodiments of the present application, the defect determining model is obtained by training through a training image set, where the training image set includes infrared images corresponding to components in the electrical device respectively serving as the defective component, and the infrared images are marked with information of the components in the electrical device respectively serving as the defective component.

As an optional implementation manner, in the first aspect of the embodiments of the present application, after the determining, by the defect determination model, the image position of the defective component of the electrical device to be identified in the first infrared image according to the second mask, the method further includes:

intercepting a first infrared sub-image corresponding to the defective component from the first infrared image according to the image position;

acquiring temperature information of the defective part according to the first infrared subimage;

respectively comparing the first infrared subimage and the temperature information of the defective component with a second infrared image of each component of the to-be-identified electrical equipment and the temperature information of each component in a database, wherein the second infrared image is an infrared image of each component of the to-be-identified electrical equipment in a normal state;

and determining the part information of the defective part according to the comparison result.

As an alternative implementation, in the first aspect of the embodiments of the present application,

the defect determination model comprises a depth residual error network with a plurality of convolution modules, and the characteristic of the temperature abnormal region is extracted through the trained defect determination model, and the method comprises the following steps:

processing the temperature abnormal area through a plurality of convolution modules of the depth residual error network to obtain a plurality of first characteristic graphs with different output sizes;

processing the first characteristic diagram with the minimum output size through a convolution layer to obtain a second characteristic diagram;

performing convolution and down-sampling operation on the second feature map to obtain a deep network feature map;

performing convolution processing on the first characteristic diagram with the second smallest output size to obtain a third characteristic diagram;

and amplifying the second characteristic diagram, and summing the amplified second characteristic diagram and the third characteristic diagram to obtain a shallow network characteristic diagram.

the defect determining model further comprises a mask generating network and a coefficient generating network, wherein the mask generating network and the coefficient generating network are parallel networks;

the generating a plurality of first masks corresponding to the temperature abnormal region according to the characteristics and determining a first mask coefficient of a defective component in the temperature abnormal region includes:

inputting the shallow network characteristic diagram into the mask generation network, and generating a plurality of first masks corresponding to the temperature abnormal areas through the mask generation network based on the shallow network characteristic diagram;

and inputting the second feature map, the deep network feature map and the shallow network feature map into the coefficient generation network together, and determining a first mask coefficient of the defective component in the temperature abnormal region through the coefficient generation network based on the second feature map, the deep network feature map and the shallow network feature map.

As an optional implementation manner, in the first aspect of the embodiments of the present application, the determining, by the defect determination model, an image position of the defective component of the electrical device to be identified in the first infrared image according to the second mask includes:

dividing the second mask of the temperature abnormal region through the defect determining model to obtain a third mask corresponding to each defective component in the temperature abnormal region;

and performing image binarization processing on the third mask through the defect determining model, and determining the image position of the defective component in the first infrared image according to the binarized third mask.

the determining a first mask coefficient of a defective component in the temperature anomaly region includes:

in the defect determining model, generating a plurality of prediction frames of the defective component in the temperature abnormal region according to the characteristics, and determining classification confidence coefficients corresponding to the prediction frames one by one and second mask coefficients corresponding to the prediction frames one by one;

and screening the plurality of prediction frames according to the classification confidence degrees corresponding to the prediction frames, determining a target prediction frame uniquely corresponding to the defective part, and determining a second mask coefficient corresponding to the target prediction frame as a first mask coefficient.

In the embodiment, an image corresponding to a temperature abnormal region in an electrical device to be identified is determined through temperature information acquired from a first infrared image of the electrical device to be identified, after features of the corresponding image are extracted through a trained defect determination model, a plurality of first masks with the same size as the corresponding image are generated according to the features, first mask coefficients of a defective component in the temperature abnormal region are determined, a second mask capable of indicating the defective component in the temperature abnormal region is generated according to the plurality of first masks and the first mask coefficients, and the image position of the defective component in the first infrared image can be accurately determined according to the second mask, so that the defective component can be accurately positioned, and the identification accuracy of the defective component in the electrical device can be effectively improved.

A second aspect of the embodiments of the present application discloses an electrical device defect determining apparatus, including:

the image acquisition module is used for acquiring a first infrared image of the electrical equipment to be identified, which is acquired by the infrared camera;

the temperature acquisition module is used for acquiring the temperature information of the electrical equipment to be identified according to the first infrared image;

the region intercepting module is used for determining a temperature abnormal region in the infrared image according to the temperature information;

the mask acquisition module is used for extracting the characteristics of the temperature abnormal region through a trained defect determination model, generating a plurality of first masks corresponding to the temperature abnormal region according to the characteristics, and determining first mask coefficients of defective components in the temperature abnormal region;

a mask determination module, configured to generate a second mask of the temperature abnormal region according to the plurality of first masks and the first mask coefficients through the defect determination model;

and the defect determining module is used for determining the image position of the defective component of the electrical equipment to be identified in the first infrared image according to the second mask through the defect determining model.

A third aspect of the embodiments of the present application discloses an electronic device, including: a memory and a processor, the memory having stored thereon a computer program executable by the processor to cause the processor to implement the electrical device defect determining method according to the first aspect of the embodiments of the present application.

A fourth aspect of embodiments of the present application discloses a computer-readable storage medium, in which a computer program is stored, the computer program being adapted to be loaded and executed by a processor, so as to enable the processor to implement a method for determining defects of an electrical device as disclosed in the first aspect of embodiments of the present application.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.

Fig. 1 is a schematic flow chart of a method for determining defects of an electrical device according to an embodiment of the present disclosure;

fig. 2 is a schematic structural diagram of a defect determination model disclosed in an embodiment of the present application, which is a Yolact model;

FIG. 3 is a schematic diagram of temperature anomaly regions obtained by different region determination methods disclosed in the embodiments of the present application;

FIG. 4 is a schematic flowchart of a process for generating a plurality of first masks corresponding to a temperature anomaly region and determining first mask coefficients of a defective component in the temperature anomaly region, according to an embodiment of the present disclosure;

FIG. 5 is a schematic diagram of different size feature map acquisition manners disclosed in embodiments of the present application;

FIG. 6 is a schematic flow chart illustrating a further method for determining defects of an electrical device according to an embodiment of the present disclosure;

fig. 7 is a block diagram of an electrical device defect determining apparatus according to an embodiment of the present application;

fig. 8 is a block diagram of another electrical device defect determining apparatus disclosed in an embodiment of the present application;

fig. 9 is a schematic structural diagram of an electronic device disclosed in an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

It should be noted that the terms "first", "second", "third" and "fourth", etc. in the description and claims of the present application are used for distinguishing different objects, and are not used for describing a specific order. The terms "comprises," "comprising," and "having," and any variations thereof, of the embodiments of the present application, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

At present, an infrared image is usually applied in an electric power system to judge whether the temperature of the electrical equipment is abnormal, but the infrared image can only reflect the area with the abnormal temperature of the electrical equipment, and due to the low resolution of the infrared image, the infrared image is combined with the existing image recognition or image comparison mode and then only adopts feature extraction or feature comparison operation, so that the position of a defective component in the image cannot be determined, and the recognition accuracy of the defective component causing the abnormal temperature in the electrical equipment is low.

In order to solve the technical problem, the embodiment of the application discloses a method and a device for determining the defect of electrical equipment, electronic equipment and a storage medium, which can accurately identify the position of the defect part of the electrical equipment in an infrared image and improve the accuracy of the defect identification of the electrical equipment. The following detailed description is made with reference to the accompanying drawings.

Referring to fig. 1, fig. 1 is a schematic flow chart illustrating a method for determining defects of an electrical device according to an embodiment of the present disclosure. As shown in fig. 1, the electrical device defect determining method may include the following steps.

101. A first infrared image of the electrical device to be identified is acquired.

In the embodiment of the present application, the electrical device is a generic name of a device used for ensuring normal operation and transmission of electric power in an electric power system, and the electrical device may include a generator, a transformer, a power line, a circuit breaker, and other devices. The infrared image is an image formed by acquiring the intensity of infrared light of an object, is obtained by measuring the heat radiated outside by the object, and is an image formed by acquiring the radiation of a target in an infrared band through an infrared imaging device.

In this embodiment of the application, the first infrared image is an infrared image of the electrical device to be identified, which is acquired through an infrared imaging device, where the electrical device to be identified is an electrical device to be identified, which has one or more defective components, and the infrared imaging device may include an infrared camera, and the like.

In the embodiment of the present application, the method for determining the defect of the electrical device is applicable to electronic devices such as a terminal device or a server, where an operating system of the electronic device may include, but is not limited to, an Android operating system, an IOS operating system, a Symbian operating system, a Black Berry operating system, a Windows Phone8 operating system, and the like.

In this embodiment, the electronic device may be provided with a User Interface (UI), an Interface module, and a Central Processing Unit (CPU).

In the embodiment of the application, after the infrared camera shooting imaging device collects the first infrared image of the electrical equipment to be identified, the first infrared image is transmitted to the electronic equipment. The transmission mode may be wired transmission or wireless network transmission. The electronic device may support network technologies including, but not limited to: global System for Mobile Communications (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), wideband Code Division Multiple Access (W-CDMA), CDMA2000, IMT Single Carrier (IMT Single Carrier), Enhanced Data rate GSM Evolution (Enhanced Data Rates for GSM Evolution, EDGE), Long-Term Evolution (Long-Term Evolution, LTE), advanced Long-Term Evolution (LTE), Time-Division Long-Term Evolution (Time-Division LTE, TD-LTE), High-Performance Radio Local Area Network (High-Performance Radio Local Area Network, High-Performance lan), High-Performance wide Area Network (High wan), Local multi-point dispatch Service (Local multi-point dispatch, LMDS), worldwide interoperability for microwave (OFDM), bluetooth (orthogonal frequency Division multiplexing), and bluetooth (orthogonal frequency Division multiplexing) technologies, High capacity spatial division multiple access (HC-SDMA), Universal Mobile Telecommunications System (UMTS), universal mobile telecommunications system time division duplex (UMTS-TDD), evolved high speed packet access (HSPA +), time division synchronous code division multiple access (TD-SCDMA), evolution data optimized (EV-DO), Digital Enhanced Cordless Telecommunications (DECT), WIFI and others.

102. And acquiring the temperature information of the electrical equipment to be identified according to the first infrared image.

In some embodiments, the electronic device obtains temperature information of the electrical device to be identified according to the first infrared image, and specifically, the first infrared image may be converted into a grayscale image, and the grayscale value of each pixel point may be converted into a corresponding temperature value according to a linear relationship between the grayscale value and the temperature value. The electronic device can also convert each pixel point in the first infrared image into a corresponding temperature value and the like according to the relationship between the RGB value and the temperature value. The present application is not particularly limited to the manner in which the temperature information is acquired from the infrared image.

103. And determining a temperature abnormal area in the first infrared image according to the temperature information.

In this embodiment, the electronic device may compare the temperature value corresponding to each pixel point in the first infrared image with a preset normal temperature range, and determine one or more image areas formed by pixel points whose temperature values exceed the normal temperature range as temperature abnormal areas.

104. The method comprises the steps of extracting features of a temperature abnormal region through a trained defect determination model, generating a plurality of first masks corresponding to the temperature abnormal region according to the features, and determining first mask coefficients of defective components in the temperature abnormal region.

In the embodiment of the present application, the trained defect determination model may be a trained example segmentation model, and the example segmentation model may include, but is not limited to, an R-CNN (region with CNN feature) model, a Fast R-CNN model, a Mask R-CNN model, or a Yolact model as shown in fig. 2.

In some embodiments, the defect-determining model is obtained by training a training image set, where the training image set includes infrared images of the electrical device when each component is a defective component, and the infrared images of the electrical device when each component is a defective component are labeled with information of the electrical device when each component is a defective component. The electronic device may adopt infrared images of each component of a plurality of different electrical devices as a defective component, for example, infrared images of the electrical devices when valves of the electrical devices are defective components, as a training image set to train the defect determination model, and the infrared images of each component as a defective component are labeled with information of each component as a defective component, the information may include defect types of the defective component, for example, when a valve of the electrical device is a defective component, the information is labeled that the defective type is damaged or the defective type is not closed. The objects which can be identified by the defect training model are only the defect parts corresponding to all the component parts in the electrical equipment, and the defect types of the defect parts can be determined according to the labels after the defect parts are identified. The method has the advantages that the defective parts in the electrical equipment to be identified can be better identified and divided, the problems of the defective parts can be effectively displayed, the accuracy of defect determination is improved, and meanwhile, the efficiency of the repair process of the defective parts is improved.

In some embodiments, the electronic device may use the infrared image region corresponding to the determined temperature anomaly region in the first infrared image as the second infrared sub-image. For the feature extraction of the temperature abnormal region, the second infrared sub-image may be specifically input into a trained defect determination model, and the feature of the second infrared sub-image, that is, the feature of the temperature abnormal region, is extracted through a convolution module included in the defect determination model. The features extracted from the temperature abnormal region may be in the form of a multi-scale feature map. The network used for feature extraction in the defect determination model is provided with a plurality of convolution modules, each convolution module comprises a plurality of convolution layers, the sizes of convolution kernels of the convolution layers and feature graphs output by the convolution modules are different, and detection of objects with different sizes can be achieved.

For example, when the trained defect determination model is a trained Yolact instance segmentation model, the Yolact instance segmentation model adopts a depth residual error network to extract a multi-scale feature map of the second infrared subimage, so as to realize feature extraction of the temperature abnormal region. For example, the deep residual network may include 5 convolution modules, each convolution module includes 3 convolution layers, and each convolution module is sorted according to the size of the feature map output by the convolution module from large to small, and may be named as conv1, conv2, conv3, conv4, and conv5, respectively. The feature map sizes of the sub-images of the temperature abnormal region output by the 5 convolution modules are different, and can be 112 × 112, 56 × 56, 28 × 28, 14 × 14 and 7 × 7, respectively. In the present application, the number of convolution layers in each convolution module and the size of convolution kernels in each convolution layer are not particularly limited.

In some embodiments, the trained defect determination model includes a mask generation network and a coefficient generation network. After the temperature abnormal region is subjected to feature extraction through the defect determining model, the extracted features are respectively input into a mask generating network and a coefficient generating network of the defect determining model, and the mask generating network outputs a plurality of first masks with the same size as the second infrared subimages according to the input features of the temperature abnormal region; the coefficient generation network outputs a series of first mask coefficients predicted for the defective part in the temperature abnormality region based on the input characteristics of the temperature abnormality region.

In an embodiment of the present application, the first mask is a mask including components in the temperature abnormal region, and the first mask may be used to share the components in the temperature abnormal region included in the mask, wherein the combination manner of the components in each first mask is different from each other. For the plurality of first masks, specifically, by inputting the characteristics of the temperature abnormal region into the mask generation network composed of the plurality of convolution layers, one first mask set, that is, a plurality of first masks, is output, the set includes a plurality of first masks having the same size, for example, the output first mask set is represented as 138 × 138 × 32, then 138 × 138 is represented as the size of each first mask, and 32 is represented as the number of first masks. Wherein, the mask generation network can be selected as the FCN network.

The first mask coefficient refers to a coefficient of a region where the characteristic of the temperature abnormality region is present, and is used to determine a first mask required for determining a region where each characteristic constituting the temperature abnormality region is present. In the embodiment of the present application, for the first mask coefficient, specifically, the characteristic of the temperature abnormal region is input into a coefficient generation network, and the coefficient generation network includes a prediction layer. In general, the characteristic of the abnormal temperature region is in the form of a characteristic diagram, and in this case, the coefficient generation network has the same number of prediction layers as the characteristic diagram, and parameters are shared among the prediction layers.

In some embodiments, in the coefficient generation network, a prediction frame (anchor) is predicted for the features of the input temperature abnormal region, and each pixel generates a plurality of prediction frames according to different proportions, that is, a range of different shapes is generated for each feature in the sub-image of the temperature abnormal region, for example, 3 prediction frames are generated for each feature. The base side length of the prediction box for each feature may be set, for example, 5 features, and then the base side lengths of the prediction boxes may be set to 24, 48, 96, 192, and 384, respectively. After the prediction boxes are generated, the coefficient generation network respectively predicts a first mask coefficient set for each generated prediction box, wherein the first mask coefficient set may include the confidence of each prediction box corresponding to a different defective part, the position information of each prediction box, and the first mask coefficient of each prediction box. The number of first mask coefficients of each prediction box is the same as the number of first masks. The first mask coefficients refer to coefficients of the respective prediction blocks for determining the first mask required to construct the respective prediction blocks.

105. And generating a second mask of the temperature abnormal region according to the plurality of first masks and the first mask coefficient through the defect determining model.

In this embodiment of the application, the electronic device generates, according to the obtained multiple first masks and the first mask coefficients, the second mask corresponding to the second infrared sub-image of the temperature abnormal region through the trained defect determination model, and specifically, the second mask corresponding to the temperature abnormal region may be obtained by linearly combining the first mask coefficients, corresponding to the first masks, of the features with the corresponding first masks. Wherein the second mask is used to identify defective parts in the temperature anomaly region. That is, the first mask is a mask of a different combination of the respective constituent members of the temperature abnormal region. For example, the mask generation network generates a first mask set with a dimension of 138 × 32, that is, generates 32 first masks of temperature abnormal regions with a size of 138 × 138, and each first mask includes different components of the temperature abnormal region. And the first mask coefficients generated by the coefficient generation network are the mask coefficients of the prediction frames of the defective parts in the temperature abnormal region, such as 3 x 32, namely the prediction frames of 3 defective parts and the mask coefficients of each prediction frame relative to 32 first masks. In this case, the linear combination of the plurality of first masks and the first mask coefficients may adopt matrix multiplication, specifically, M ═ PC^TWhere P is the first mask set, i.e., the plurality of first masks generated by the mask generation network, C is the first mask coefficient of each prediction box, P is 138 is 32, and C is 3 is 32.

106. And determining the image position of the defective part of the electrical equipment to be identified in the first infrared image according to the second mask through the defect determination model.

In some embodiments, the electronic device may perform pixel filtering on the first infrared image through a trained defect determination model according to an obtained second mask including a prediction frame of a defective component in the temperature abnormal region, display one or more defective components in the temperature abnormal region corresponding to the prediction frame in the second mask according to a position of the prediction frame in the first infrared image, and further determine a position of the one or more defective components in the electrical device to be identified.

In the embodiment of the application, an image corresponding to a temperature abnormal region in an electrical device to be identified is determined through temperature information obtained from a first infrared image of the electrical device to be identified, after features of the corresponding image are extracted through a trained defect determination model, a plurality of first masks with the same size as the corresponding image are generated according to the features, first mask coefficients of a defective component in the temperature abnormal region are determined, a second mask capable of indicating the defective component in the temperature abnormal region is generated according to the plurality of first masks and the first mask coefficients, and the image position of the defective component in the first infrared image can be accurately determined according to the second mask, so that the defective component can be accurately positioned, and the identification accuracy of the defective component in the electrical device can be effectively improved.

In some embodiments, because the first infrared image has poor resolution and a blurred visual effect, the first infrared image is denoised before the temperature information of the electrical device to be identified is acquired according to the first infrared image. The denoising process specifically includes: decomposing the first infrared image into three corresponding component images according to R, G, B, respectively performing image filtering on the three component images obtained by decomposition by adopting a filtering function, and merging the component images after filtering into a new first infrared image, namely synthesizing the filtered first infrared image. The filter function may be a two-dimensional median filter function, a block filter function, a mean filter function, a gaussian filter function, or a bilateral filter function.

In the embodiment of the application, the resolution ratio of the first infrared image can be improved, and the accuracy of the acquired temperature information is further improved.

In some embodiments, for an image region composed of pixels with temperature values exceeding the normal temperature range, the pixels with temperature values exceeding the normal temperature range are defined as abnormal pixels. According to the temperature information, determining a temperature abnormal area in the first infrared image, comprising:

dividing the abnormal pixel points and the pixel points adjacent to the abnormal pixel points or the pixel points in a certain pixel interval into the same abnormal area; the abnormal region and the abnormal region adjacent to the abnormal region or the abnormal region within a certain pixel interval are determined as temperature abnormal regions.

For example, as shown in fig. 3, in the first infrared image with a size of 10 × 10, each cell 1 represents one pixel point, and the pixel point with the star mark is an abnormal pixel point. For the temperature abnormal region determined in the first infrared image, the temperature abnormal region may be an image region 2 composed of 6 pixel points with star marks, or an image region 3 composed of 6 pixel points with star marks and pixel points adjacent to the 6 pixel points with star marks.

In some embodiments, after dividing the abnormal pixel point and the pixel point adjacent to the abnormal pixel point or the pixel point in a certain pixel interval into the same abnormal region, the method further includes:

and counting the number of abnormal pixel points contained in each abnormal area, comparing the number with a set abnormal pixel point threshold value, and if the number of the abnormal pixel points contained in the abnormal area is less than the set abnormal pixel point threshold value, discharging the abnormal area. The method can avoid the problem that the area is divided into abnormal areas due to the abnormal temperature values of a few pixel points, so that the range of the abnormal temperature areas is increased, and the calculation amount is increased, thereby reducing the calculation amount and improving the image processing efficiency. In some embodiments

As an alternative implementation, in the step 106 of determining the image position of the defective component of the electrical device to be identified in the first infrared image according to the second mask by the defect determination model, the following steps may be performed:

dividing the second mask of the temperature abnormal region through a defect determining model to obtain a third mask corresponding to each defective component in the temperature abnormal region;

and performing image binarization processing on the third mask through a defect determining model, and determining the image position of the defective part in the first infrared image according to the binarized third mask.

In the embodiment of the application, after the electronic device obtains the second mask of the temperature abnormal region, the prediction frames of the defective parts in the second mask are divided by the defect determination model to obtain the third mask, for example, the prediction frames of the three defective parts in the second mask may be included, and the three prediction frames in one second mask may be divided into three third masks each including one prediction frame by Crop operation in the Yolact model. The electronic device performs image binarization processing on each third mask, that is, the gray value of a pixel point in the prediction frame of each third mask is changed to 0 or 255 by Threshold operation in the Yolact model according to a set Threshold (for example, 0.5, 0.4, and the like, which is not limited herein), so that the display effect of a defective part in the prediction frame can be improved. And accurately marking the image position of each defective component in the temperature abnormal area of the prediction frame according to each third mask after binarization processing, and further converting the image position into a position in the electrical equipment to be identified according to the image position in the first infrared image of each defective component. Defective components in the electrical device to be identified can be seen more clearly.

As an alternative embodiment, after the process of determining the image position of the defective component of the electrical device to be identified in the first infrared image according to the second mask by means of the defect determination model is performed in step 106, the following steps may be performed:

respectively comparing the first infrared subimage and the temperature information of the defective component with a second infrared image of each component of the electrical equipment to be identified and the temperature information of each component in the database, wherein the second infrared image is an infrared image of each component of the electrical equipment to be identified in a normal state;

In the embodiment of the application, the electronic device intercepts a first infrared sub-image corresponding to a defective component, such as a sub-image of the defective component of a valve, from a first infrared image according to the determined image position of the defective component of the electrical device to be identified in the first infrared image, and determines the infrared image of the area corresponding to the defective component as the first infrared sub-image. For the image region corresponding to the defective component, the image region can be determined by using the prediction frame of the defective component in the second mask.

In the embodiment of the application, the electronic device further obtains the temperature information of the first infrared subimage, that is, the temperature information of the defective component, according to the intercepted first infrared subimage where the defective component is located. The manner of acquiring the temperature information of the defective component according to the first infrared sub-image is the same as the manner of acquiring the temperature information of the first infrared image, and is not described herein again. In the embodiment of the application, the electronic device performs image comparison on the first infrared subimage and second infrared images of all components in the electrical device to be identified in the database, and compares temperature information of a defective component with temperature information of all components in a normal state, wherein the second infrared images are infrared images of all components of the electrical device to be identified in the normal state.

For example, the database stores the infrared images of the components of the electrical device to be identified in the normal state, that is, the second infrared image, including the valves, the pipelines, the transformers, and the like. And the electronic equipment performs image comparison on the first infrared subimages of the defective part and the second infrared images one by one in a traversing mode so as to find out the second infrared images corresponding to the first infrared subimages.

As a specific embodiment, the image comparison method between the first infrared sub-image and each second infrared image may include: comparing the information of each pixel point in the first infrared subimage with the information of the corresponding pixel point in each second infrared image; and judging whether the two pixel points belong to the matched pixel points or not according to the set similarity threshold. And if the similarity between the information of each pixel point in the first infrared sub-image and the information of the corresponding pixel point in the second infrared image exceeds a similarity threshold, considering the two pixel points as matched pixel points. And judging whether the first infrared subimage corresponds to the second infrared image according to the set pixel point threshold. And if the number of matched pixel points in the first infrared sub-image and the second infrared image exceeds a pixel point threshold value, the first infrared sub-image is considered to correspond to the second infrared image.

In the embodiment of the present application, for comparison between the temperature information of the defective component and the temperature information of each component in the normal state, specifically, by calculating a difference between the temperature information of the defective component and the temperature information of each component in the normal state, it is determined whether the difference is greater than a set difference threshold, and if so, the component is further determined to be the defective component.

In this embodiment, each second infrared image is bound with information of a corresponding component of the electrical device, for example, when one second infrared image is an infrared image in a normal state of the valve, the second infrared image is bound with information of a valve component, and the information may include information such as a name of the component, a size of the component, and a position of the component in the electrical device. The information binding manner between each second infrared image and the corresponding component of the electrical device is not particularly limited herein.

In the embodiment of the application, the electronic device obtains information of the component of the electrical device bound by the second infrared image according to a comparison result of the first infrared sub-image and the second infrared image, that is, according to the second infrared image corresponding to the first infrared sub-image after image comparison, and further determines component information of the defective component, that is, obtains information of the name, the size, the position in the electrical device, and the like of the defective component. And according to the comparison result of the temperature information of the defective component and the temperature information of each component in the normal state, the accuracy of determining the defective component is further verified, so that the defective component can be further determined and the related information of the defective component can be effectively determined, the replacement or maintenance of the subsequent defective component and the improvement of the identification accuracy of the defective component are facilitated, and the safety of the electrical equipment is effectively improved.

In an embodiment, as shown in fig. 4, the step of extracting features of the temperature abnormal region through a trained defect determination model, generating a plurality of first masks corresponding to the temperature abnormal region according to the features, and determining first mask coefficients of a defective component in the temperature abnormal region may include the following steps:

401. and processing the temperature abnormal area through a plurality of convolution modules of the depth residual error network to obtain a plurality of first characteristic graphs with different output sizes.

In the embodiment of the application, the electronic device inputs the second infrared subimage into the network for feature extraction to obtain a plurality of first feature maps with different sizes. And taking the infrared image part of the determined temperature abnormal area corresponding to the first infrared image as a second infrared sub-image. The network used for feature extraction in the defect determination model is a deep residual network with a plurality of convolution modules. And the depth residual error network performs feature extraction on the second infrared sub-images and outputs a plurality of first feature maps.

402. And processing the first characteristic diagram with the minimum output size through a convolution layer to obtain a second characteristic diagram.

403. And performing convolution and downsampling operation on the second feature map to obtain the features of the deep network feature map.

404. And performing convolution processing on the first feature map with the second smallest output size to obtain a third feature map.

405. And amplifying the second characteristic diagram, and summing the amplified second characteristic diagram and the third characteristic diagram to obtain a shallow network characteristic diagram.

In the embodiment of the application, in order to enrich the features contained in the feature map of the temperature abnormal region, generate multi-scale feature representation and be more favorable for determining the defect component in the electrical equipment, the trained defect determination model may be provided with a network structure for further feature extraction processing of the first feature map. Wherein, the network structure can be selected as FPN network.

In the embodiment of the application, after the feature extraction processing is performed on the second infrared subimage, the depth residual error network in the trained defect determination model obtains a plurality of first feature maps, and the first feature maps are sorted from large to small according to size. After the first characteristic diagrams are sorted, the defect determining model respectively performs convolution operation on the characteristic diagram with the minimum size and the characteristic diagram with the second minimum size in the first characteristic diagrams to obtain a second characteristic diagram and a third characteristic diagram. And expanding the second characteristic diagram by adopting bilinear interpolation on the obtained second characteristic diagram, and adding the expanded second characteristic diagram and the third characteristic diagram to obtain a deep network characteristic diagram. And in addition, the second feature map is subjected to convolution and downsampling operation to obtain a shallow network feature map.

For example, after feature extraction processing is performed on the second infrared sub-image by the depth residual error network in the trained defect determination model, three first feature maps are obtained, and the first feature maps are respectively C3-C5 in descending order of size. As shown in fig. 5, the trained defect determining model performs convolution operation on the feature map C5 with the smallest size in the first feature map by 1 convolutional layer to obtain a second feature map, that is, a feature map P5, then performs bilinear interpolation on the second feature map to expand the second feature map by one time, and adds the expanded second feature map and the third feature map C4 with the second smallest output size after the convolution operation to obtain a deep-layer network feature map P4. The second feature map P5 is subjected to convolution and downsampling operations to obtain a shallow network feature map P6. The deep network characteristic diagram is obtained through down sampling, so that the deep network characteristic diagram has higher characteristics that the receptive field is more abstract, and the information of the first infrared subimage can be better expressed; the shallow network characteristic graph has larger size and more detailed information due to the expansion processing.

By implementing the method, the characteristic graphs containing different receptive fields and different sizes can be obtained, the first mask generated according to the characteristic graphs and the determined first mask coefficient can contain the information of the defective part as much as possible, and the accuracy of defect identification is improved.

In some embodiments, to acquire more different receptive fields and different size profiles, may include: performing convolution processing on the first feature map with the third smallest output size to obtain a fourth feature map;

amplifying the shallow network characteristic diagram, and summing the amplified shallow network characteristic diagram and the fourth characteristic diagram to obtain a second shallow network characteristic diagram;

and carrying out convolution and downsampling operation on the deep network feature map to obtain a second deep network feature map.

For example, when the trained defect determining model performs feature extraction processing on the second infrared subimage to obtain 5 first feature maps, the first feature maps are sorted from large to small in size to be C1-C5, respectively. After the defect determining model obtains the second feature map, the third feature map and the fourth feature map through the feature extraction process, the defect determining model can further perform feature extraction processing on the second feature map, the shallow network feature map and the deep network feature map through the trained defect determining model on the basis of the third feature map and the fourth feature map to obtain more feature maps with different scales. The concrete mode is as follows: and similarly, bilinear interpolation is adopted for the shallow layer network feature map P4 to enable the feature map to be enlarged by one time, and then the feature map is added with a fourth feature map C3 to obtain a second shallow layer network feature map P3, wherein the fourth feature map C3 is obtained by performing convolution operation on the first feature map with the third smallest output size. The deep network feature map P6 is subjected to the same convolution and lower adoption operation as the first feature map P5 of the second smallest output size, and the second deep network feature map P7 is obtained.

According to the defect identification method and device, more feature maps with different sizes can be obtained according to the obtained first feature maps, and the accuracy of defect identification is improved.

406. And generating a plurality of first masks corresponding to the temperature abnormal area according to the second characteristic diagram, the deep layer network characteristic diagram and the shallow layer network characteristic diagram, and determining a first mask coefficient of the defective component in the temperature abnormal area.

In some embodiments, as an optional implementation manner, the defect determination model further includes a mask generation network and a coefficient generation network, and the mask generation network and the coefficient generation network are parallel networks.

Step 406 may include: inputting the shallow network characteristic diagram into a mask generation network, and generating a plurality of first masks corresponding to the temperature abnormal areas through the mask generation network based on the shallow network characteristic diagram; and inputting the second characteristic diagram, the shallow layer network characteristic diagram and the deep layer network characteristic diagram into a coefficient generation network together, and determining a first mask coefficient of the defective component in the temperature abnormal area through the coefficient generation network based on the second characteristic diagram, the shallow layer network characteristic diagram and the deep layer network characteristic diagram.

In the embodiment of the application, if 3 first feature maps are further subjected to feature processing in a trained defect determination model to obtain a second feature map, a shallow network feature map and a deep network feature map, the shallow network feature map P4 is input into a mask generation network to generate a plurality of first masks, and the mask generation network may be a full convolution network. The second feature map P5, the shallow network feature map P4, and the deep network feature map P6 are input into the coefficient generation network to generate a first mask coefficient. If 5 first feature maps are further subjected to feature processing in the trained defect determination model to obtain a second feature map, a shallow network feature map, a deep network feature map, a second shallow network feature map P3 and a second deep network feature map P7, the second shallow network feature map P3 is input into the mask generation network to generate a plurality of first masks.

According to the second feature map after the convolution operation, the shallow layer network feature map after the image expansion operation, the deep layer network feature map after the down sampling operation and the acquisition processes of the feature maps P3 and P7, the feature map with larger size and more detail information of the temperature abnormal region is input into the mask generation network. The mask generation network may be selected as a full convolution network. The second feature map P5, the shallow network feature map P4, the deep network feature map P6, the second shallow network feature map P3 and the second deep network feature map P7 are input into a coefficient generation network to generate first mask coefficients. The coefficient generation network may be selected as a Retina Net employing a shared convolutional network. In the trained defect determining model, the process of generating a plurality of first masks after the network input feature diagram is generated by the masks and the process of generating the first mask coefficients after the network input feature diagram is generated by the coefficients serve as two parallel tasks, so that the speed of the process of determining the defective parts in the electrical equipment by the trained defect determining model can be increased, and the recognition efficiency is improved.

By adopting the method for determining the defect of the electrical equipment described in the embodiment, the characteristic diagrams with different sizes can be obtained, and the obtained characteristic diagrams with different sizes can better express the global information of the temperature abnormal area and the information of the defective component in the temperature abnormal area, so that the accuracy of defect determination is improved.

Referring to fig. 6, fig. 6 is a schematic flow chart illustrating another electrical device defect determining method according to an embodiment of the present disclosure. As shown in fig. 6, the electrical device defect determining method may include the steps of:

601. a first infrared image of the electrical device to be identified is acquired.

602. And acquiring the temperature information of the electrical equipment to be identified according to the first infrared image.

603. And determining a temperature abnormal area in the first infrared image according to the temperature information.

604. And extracting a shallow network characteristic diagram of the temperature abnormal region through the trained defect determination model, and generating a plurality of first masks corresponding to the temperature abnormal region according to the shallow network characteristic diagram.

605. In the defect determining model, a plurality of prediction frames of the defective parts in the temperature abnormal region are generated according to the second feature map, the shallow layer network feature map and the deep layer network feature map, and classification confidence coefficients corresponding to the prediction frames one by one and second mask coefficients corresponding to the prediction frames one by one are determined.

606. And screening the plurality of prediction frames according to the classification confidence degrees corresponding to the prediction frames through a defect determination model, determining a target prediction frame uniquely corresponding to the defective part, and determining a second mask coefficient corresponding to the target prediction frame as a first mask coefficient.

607. And generating a second mask of the temperature abnormal region according to the plurality of first masks and the first mask coefficient through the defect determining model.

608. And determining the image position of the defective part of the electrical equipment to be identified in the first infrared image according to the second mask through the defect determination model.

In the embodiment of the present application, the classification confidence of the prediction box, that is, the prediction box described above, is the corresponding confidence when the prediction box is a different defective component, for example, the confidence when the prediction box is a defective valve is 0.1, and the confidence when the prediction box is a defective pipe is 0.9.

In the embodiment of the application, an NMS algorithm can be adopted to screen the prediction boxes according to the classification confidence, specifically, 5 prediction boxes are provided for one defective component in the temperature abnormal region, and the confidence degrees of the 5 prediction boxes corresponding to the defective component are ranked from high to low to obtain B1-B5; calculating the interaction ratio (IOU) among the 5 prediction frames through matrix operation to obtain a symmetric matrix; deleting diagonal lines and lower triangular elements of the symmetric matrix, and taking the maximum value of each column in the rest matrixes to obtain the IOU value of each prediction box; according to the set threshold value, the prediction frame that the IOU value is larger than the threshold value is omitted. This is because each element in the matrix is row number smaller than column number, and the sequence numbers are arranged in descending order according to the confidence corresponding to the defective component, so that any element is larger than the threshold, which means that the prediction frame corresponding to this column is too overlapped with the prediction frame with higher confidence than the prediction frame, and therefore needs to be discarded. The redundant prediction frames can be effectively removed, and the prediction frame with the highest degree of overlapping with the standard prediction frame is reserved.

In the embodiment of the application, each prediction frame is screened through a defect determination model, a first prediction frame uniquely corresponding to each defective component is determined, that is, the prediction frame uniquely corresponding to each defective component, and a mask coefficient corresponding to the prediction frame is determined as a first mask coefficient, that is, a mask coefficient of the prediction frame relative to each first mask is determined as a first mask coefficient of the defective component corresponding to the confidence.

Referring to fig. 7, fig. 7 is a block diagram of an electrical device defect determining apparatus according to an embodiment of the present disclosure. As shown in fig. 7, the electrical device defect determining apparatus may include:

the image acquisition module 701 is used for acquiring a first infrared image of the electrical equipment to be identified, which is acquired by an infrared camera;

a temperature obtaining module 702, configured to obtain temperature information of the electrical device to be identified according to the first infrared image;

the area determining module 703 is configured to determine, according to the temperature information, an abnormal temperature area in the infrared image;

the feature extraction module 704 is used for extracting features of the temperature abnormal region through the trained defect determination model, generating a plurality of first masks corresponding to the temperature abnormal region according to the features, and determining first mask coefficients of defective components in the temperature abnormal region;

a mask generation module 705, configured to generate a second mask of the temperature abnormal region according to the plurality of first masks and the first mask coefficients through the defect determination model;

and a defect determining module 706, configured to determine, according to the second mask, an image position of the defective component of the electrical device to be identified in the first infrared image through the defect determining model.

It can be seen that, with the electrical device defect determining apparatus described in the foregoing embodiment, an image corresponding to a temperature abnormal region in a first infrared image of an electrical device to be identified can be determined according to temperature information acquired from the first infrared image, a trained defect determining model is used to extract features of the corresponding image, and then a plurality of first masks having the same size as the corresponding image are generated according to the features, and a second mask capable of indicating a defective component in the temperature abnormal region is generated according to the plurality of first masks and the first mask coefficients according to a first mask coefficient of the defective component in the temperature abnormal region, and an image position of the defective component in the first infrared image can be accurately determined according to the second mask, so that the defective component can be accurately positioned, and an accuracy of identifying the defective portion in the electrical device can be effectively improved.

Referring to fig. 8, fig. 8 is a block diagram of another electrical device defect determining apparatus according to an embodiment of the present disclosure. The electrical equipment defect determining apparatus shown in fig. 8 is optimized by the electrical equipment defect determining apparatus shown in fig. 7. Compared with the electrical equipment defect determining apparatus shown in fig. 7, the electrical equipment defect determining apparatus shown in fig. 8 may further include:

the image comparison module 801 is configured to compare the first infrared subimage and the temperature information of the defective component with a second infrared image of each component of the electrical device to be identified in the database and the temperature information of each component, respectively, where the second infrared image is an infrared image of each component of the electrical device to be identified in a normal state;

and an information determining module 802, configured to determine component information of the defective component according to a comparison result.

The area determining module 703 may be further configured to intercept a first infrared sub-image corresponding to the defective component from the first infrared image according to the image position.

The temperature obtaining module 702 may be further configured to obtain temperature information of the defective component according to the first infrared sub-image.

Therefore, the electrical equipment defect determining device described in the above embodiment can determine the related information of the defective component, and is beneficial to the replacement or maintenance of the subsequent defective component, thereby improving the safety of the electrical equipment.

As an optional implementation manner, the feature extraction module 704 may be further configured to:

processing the temperature abnormal area through a plurality of convolution modules of the depth residual error network to obtain a plurality of first characteristic graphs with different output sizes; processing the first characteristic diagram with the minimum output size through a convolution layer to obtain a second characteristic diagram; performing convolution and down-sampling operation on the second feature map to obtain a deep network feature map; performing convolution processing on the first characteristic diagram with the second smallest output size to obtain a third characteristic diagram; and amplifying the second characteristic diagram, and summing the amplified second characteristic diagram and the third characteristic diagram to obtain a shallow network characteristic diagram.

It can be seen that, by using the electrical equipment defect determining apparatus described in the above embodiment, the feature maps of different sizes can be obtained, and the obtained feature maps of different sizes can better represent the global information of the temperature abnormal region and the information of the defective component in the temperature abnormal region, thereby improving the accuracy of defect determination.

As an optional implementation manner, the defect determination model further includes a mask generation network and a coefficient generation network, and the mask generation network and the coefficient generation network are parallel networks.

The feature extraction module 704 may be further configured to:

inputting the shallow network characteristic diagram into a mask generation network, and generating a plurality of first masks corresponding to the temperature abnormal areas through the mask generation network based on the shallow network characteristic diagram; and inputting the second characteristic diagram, the shallow layer network characteristic diagram and the deep layer network characteristic diagram into a coefficient generation network together, and determining a first mask coefficient of the defective component in the temperature abnormal area through the coefficient generation network based on the second characteristic diagram, the shallow layer network characteristic diagram and the deep layer network characteristic diagram.

It can be seen that, by using the electrical device defect determining apparatus described in the above embodiment, parallel operation of a plurality of first mask obtaining processes and a plurality of first mask coefficient obtaining processes can be achieved, and the rate of obtaining a defective part prediction frame can be effectively increased, so that the rate of the whole defect determining process is increased, and the timeliness of safety detection is enhanced.

As an optional implementation manner, the defect determining module 706 may further be configured to:

dividing the second mask of the temperature abnormal region through a defect determining model to obtain a third mask of each component in the temperature abnormal region; performing image binarization processing on the third mask according to a preset threshold value through a defect determining model to obtain a fourth mask; and determining the image position of the defective part of the electrical equipment to be identified in the first infrared image according to the fourth mask.

Therefore, by adopting the electrical equipment defect determining device described in the above embodiment, display optimization of the mask corresponding to the electrical equipment defective component can be realized, so that the defective component can be better displayed and determined.

in the defect determining model, generating a plurality of prediction frames of the defective part in the temperature abnormal region according to the characteristics, and determining classification confidence coefficients corresponding to the prediction frames one by one and second mask coefficients corresponding to the prediction frames one by one; and screening the plurality of prediction frames according to the classification confidence degrees corresponding to the prediction frames, determining a target prediction frame uniquely corresponding to the defective part, and determining a second mask coefficient corresponding to the target prediction frame as a first mask coefficient.

Therefore, by using the electrical equipment defect determining device described in the above embodiment, the categories of the defective components and the prediction frames possibly corresponding to the categories can be screened, and the accuracy of the obtained prediction frames of the defective components can be improved.

Referring to fig. 9, fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure.

As shown in fig. 9, the electronic device may include:

a memory 901 in which executable program code is stored;

a processor 902 coupled to a memory 901;

the processor 902 calls the executable program code stored in the memory 901 to execute one of the electrical device defect determining methods of fig. 1, 4 or 6.

An embodiment of the application discloses a computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute one of the electrical device defect determining methods of fig. 1, 4 or 6.

The embodiment of the present application also discloses an application publishing platform, wherein the application publishing platform is used for publishing a computer program product, and when the computer program product runs on a computer, the computer is caused to execute part or all of the steps of the method in the above method embodiments.

It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearances of the phrases "in some embodiments" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. Those skilled in the art should also appreciate that the embodiments described in this specification are all alternative embodiments and that the acts and modules involved are not necessarily required for this application.

In various embodiments of the present application, it should be understood that the size of the serial number of each process described above does not mean that the execution sequence is necessarily sequential, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated units, if implemented as software functional units and sold or used as a stand-alone product, may be stored in a computer accessible memory. Based on such understanding, the technical solution of the present application, which is a part of or contributes to the prior art in essence, or all or part of the technical solution, may be embodied in the form of a software product, stored in a memory, including several requests for causing a computer device (which may be a personal computer, a server, a network device, or the like, and may specifically be a processor in the computer device) to execute part or all of the steps of the above-described method of the embodiments of the present application.

It will be understood by those skilled in the art that all or part of the steps in the methods of the embodiments described above may be implemented by hardware instructions of a program, and the program may be stored in a computer-readable storage medium, where the storage medium includes Read-Only Memory (ROM), Random Access Memory (RAM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), One-time Programmable Read-Only Memory (OTPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Compact Disc Read-Only Memory (CD-ROM), or other Memory, such as a magnetic disk, or a combination thereof, A tape memory, or any other medium readable by a computer that can be used to carry or store data.

The method for determining the defect of the electrical device and the terminal device disclosed in the embodiment of the present application are described in detail above, a specific example is applied in the description to explain the principle and the implementation of the present application, and the description of the above embodiment is only used to help understanding the method and the core idea of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. A method for determining defects in an electrical device, the method comprising:

acquiring a first infrared image of the electrical equipment to be identified;

2. The method according to claim 1, wherein the defect-determining model is trained by a training image set, the training image set includes infrared images corresponding to the components of the electrical device as defective components, and the infrared images are marked with information about the components of the electrical device as defective components.

3. The method according to claim 1, characterized in that after said determining, by means of said defect determination model, the image position of the defective component of the electrical device to be identified in the first infrared image according to the second mask, the method further comprises:

4. The method of claim 1, wherein the defect determination model comprises a deep residual network having a plurality of convolution modules, and wherein extracting the features of the temperature anomaly region through the trained defect determination model comprises:

5. The method of claim 4, wherein the defect determination model further comprises a mask generation network and a coefficient generation network, the mask generation network and the coefficient generation network being parallel networks;

6. The method according to claim 1, wherein said determining, by said defect determination model, an image position of a defective component of said electrical device to be identified in said first infrared image from said second mask comprises:

7. The method of claim 1, wherein determining a first mask coefficient for a defective component in the temperature anomaly region comprises:

8. An electrical equipment defect identification device, comprising:

the area determining module is used for determining an abnormal temperature area in the infrared image according to the temperature information;

the characteristic extraction module is used for extracting the characteristics of the temperature abnormal region through a trained defect determination model, generating a plurality of first masks corresponding to the temperature abnormal region according to the characteristics, and determining first mask coefficients of defective components in the temperature abnormal region;

a mask generation module, configured to generate a second mask of the temperature abnormal region according to the plurality of first masks and the first mask coefficient through the defect determination model;

9. An electronic device, comprising: memory and a processor, on which a computer program is stored, characterized in that the computer program is executable by the processor to cause the processor to implement the method according to any of claims 1 to 7.

10. A computer-readable storage medium, in which a computer program is stored which is adapted to be loaded and executed by a processor such that having the processor implement the method according to any of claims 1-7.