Sectional insulator anomaly identification method, device, medium and equipment based on improved UNet3+ network

Abstract

The present invention discloses a method for identifying abnormalities of segmented insulators based on an improved UNet3+ network, comprising: obtaining each frame of image information in a surveillance video to be identified; using target detection technology to roughly locate the segmented insulator in the image information to obtain the position information of the segmented insulator; segmenting the image information according to the position information to obtain image blocks of the segmented insulator; using an improved UNet3+ network to extract detailed features of the image blocks to obtain feature information of the segmented insulator; and performing abnormal warning of the segmented insulator according to the feature information. The present invention solves the problems of insufficient robustness and lack of universality in the prior art when performing abnormal identification of segmented insulators, and realizes real-time monitoring of equipment.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, medium and equipment for identifying abnormalities of a segmented insulator based on an improved UNet3+ network. Background Art

[0002] Subway is a form of railway transportation. It is a rail transit system that mainly operates underground. It has the advantages of high capacity, high density, and exclusive right of way. At present, domestic subway construction is booming. With the continuous improvement of urbanization, more and more people are pouring into cities, which makes the daily average passenger flow of urban rail transit and the number of electric locomotives running on the line increase year by year, and the working intensity of the contact network is greatly improved. The segment insulator is an insulating device used for electrical segmentation of the contact network, which plays a very important role in the electrified railway.

[0003] As an important electrical equipment in the traction power supply system, the segmented insulator is expensive and difficult to replace. It is also installed on the top of the train. Conventional manual inspection is time-consuming and labor-intensive, lacks accuracy, and has no unified inspection standards. It is easily affected by the subjective factors of maintenance personnel. The existing technology uses an abnormality recognition method based on traditional image processing. However, the recognition model has the disadvantages of insufficient robustness, lack of universality, and the need for a large number of hyperparameter tuning. It is impossible to perform real-time monitoring. It mainly indirectly realizes the tracking of equipment operation status by increasing the frequency of maintenance and patrols. The repair efficiency is low and the labor cost is high. Summary of the invention

[0004] The embodiments of the present invention provide a method, device, medium and equipment for identifying abnormalities of segmented insulators based on an improved UNet3+ network, so as to solve the problems of insufficient robustness and lack of universality in the prior art when identifying abnormalities of segmented insulators, and realize real-time monitoring of the equipment.

[0005] A method for identifying abnormality of a segmented insulator based on an improved UNet3+ network, the method comprising:

[0006] Obtain each frame of image information in the surveillance video to be identified;

[0007] Using target detection technology to roughly locate the segmented insulator in the image information to obtain position information of the segmented insulator;

[0008] Segmenting the image information according to the position information to obtain image blocks of segmented insulators;

[0009] Using an improved UNet3+ network to extract detail features from the image block, and obtaining feature information of the segmented insulator;

[0010] An abnormality warning of the segment insulator is performed according to the characteristic information.

[0011] Optionally, the encoder of the improved UNet3+ network replaces the pooling layer with a hole convolution layer with a dilation rate of 2.

[0012] Optionally, in the encoder of the improved UNet3+ network, a convolution operation is performed on the image block according to a first convolution kernel, and then a normalization process and a Relu activation function operation are performed to obtain a first feature map;

[0013] Performing a convolution operation on the first feature map according to a second convolution kernel, and then performing a normalization process and a Relu activation function operation to obtain a second feature map;

[0014] The second convolution kernel is updated to a convolution kernel with a dilation rate of 2, and the updated convolution kernel is used to perform a convolution operation on the first feature map, and then normalizes and performs a Relu activation function operation to obtain a third feature map after tensor processing;

[0015] Performing an add operation on the second feature map and the third feature map to obtain a fourth feature map;

[0016] The fourth feature map is subjected to feature extraction using a dilation rate of 2 dilated convolutional layer, and the representation capability is increased through the SENet channel attention mechanism to obtain the encoder output.

[0017] Optionally, the second convolution kernel is a 3*3 convolution operation.

[0018] Optionally, the characteristic information of the segmented insulator includes bolt line drawing information;

[0019] The abnormal warning of the segment insulator according to the characteristic information includes:

[0020] If there is only one bolt line information, no abnormal warning is issued;

[0021] If there are multiple bolt line information, the angle information of the bolt line information is calculated, and an abnormal warning is performed according to the angle information.

[0022] Optionally, the angle information for calculating the bolt line information includes:

[0023] Get the first two bolt lines as the target lines;

[0024] Calculate the slope between the target lines and get the cosine value according to the tangent formula;

[0025] The inverse cosine is taken according to the cosine value to obtain the angle information of the bolt line information.

[0026] Optionally, the characteristic information of the segmented insulator includes a region of interest of a guide plate;

[0027] The abnormal warning of the segment insulator according to the characteristic information includes:

[0028] Extracting width information of the region of interest of the guide plate;

[0029] Comparing the width information with a preset width threshold range;

[0030] If the width information exceeds the width threshold range, an abnormal warning is performed.

[0031] A sectional insulator abnormality identification device based on an improved UNet3+ network, the device comprising:

[0032] An acquisition module is used to acquire each frame of image information in the surveillance video to be identified;

[0033] A positioning module, used for roughly positioning the segmented insulator in the image information by using a target detection technology to obtain position information of the segmented insulator;

[0034] A segmentation module, used for segmenting the image information according to the position information to obtain image blocks of segmented insulators;

[0035] An extraction module, used for extracting detail features of the image block using an improved UNet3+ network to obtain feature information of the segmented insulator;

[0036] The early warning module is used to issue an abnormal early warning of the segmented insulator according to the characteristic information.

[0037] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the segmented insulator abnormality identification method based on the improved UNet3+ network as described above is implemented.

[0038] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the segmented insulator abnormality identification method based on the improved UNet3+ network as described above is implemented.

[0039] The embodiment of the present invention obtains each frame of image information in the monitoring video to be identified; uses target detection technology to roughly locate the segmented insulator in the image information to obtain the position information of the segmented insulator; segments the image information according to the position information to obtain image blocks of the segmented insulator; uses an improved UNet3+ network to extract detail features of the image blocks to obtain feature information of the segmented insulator; and performs abnormal warning of the segmented insulator based on the feature information; thereby introducing deep learning into the abnormal monitoring of the segmented insulator, completing target detection and image segmentation through deep learning, effectively solving the problems of insufficient robustness and lack of universality in the prior art when performing abnormal identification of the segmented insulator, and realizing real-time monitoring of the equipment, effectively improving the repair efficiency and saving labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0041] Figure 1 It is a flow chart of a method for identifying abnormality of a segmented insulator based on an improved UNet3+ network provided by an embodiment of the present invention;

[0042] Figure 2 It is a schematic diagram of the structure of an encoder of an improved UNet3+ network provided by an embodiment of the present invention;

[0043] Figure 3 A schematic diagram of bolt line drawing and target line drawing provided by an embodiment of the present invention, wherein Figure 3 (a) indicates a straight line pre-drawn on the bolt and fastener. Figure 3 (b) indicates target drawing line;

[0044] Figure 4 It is a structural schematic diagram of a sectional insulator abnormality identification device based on an improved UNet3+ network provided by an embodiment of the present invention;

[0045] Figure 5 is a schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] The embodiment of the present invention introduces deep learning in the step of identifying abnormalities of segmented insulators, uses target detection to roughly locate key positions, and then uses the improved UNet3+ network to extract the required abnormality detection and discrimination information, thereby improving the accuracy of abnormality identification, effectively solving the problems of insufficient robustness and lack of universality in the prior art when identifying abnormalities of segmented insulators, realizing real-time monitoring of equipment, improving emergency repair efficiency and saving labor costs.

[0048] The following is a detailed description of the segmented insulator abnormality identification method based on the improved UNet3+ network provided in this embodiment. Figure 1 As shown, the segmented insulator abnormality identification method based on the improved UNet3+ network includes:

[0049] In step S101, each frame of image information in the surveillance video to be identified is obtained.

[0050] Here, the embodiment of the present invention uses a camera, a color video camera and other equipment to monitor the segmented insulator and capture the monitoring video as the monitoring video to be identified. After obtaining the monitoring video to be identified, the monitoring video to be identified is converted into image information according to the frame number.

[0051] Optionally, as a preferred embodiment of the present invention, after obtaining the image information, the method further includes: performing data preprocessing on each frame of the image information, wherein the data preprocessing includes but is not limited to image noise reduction, enhancement and cropping.

[0052] In step S102, the segmented insulator in the image information is roughly located using target detection technology to obtain position information of the segmented insulator.

[0053] The target detection technology is a YOLO network, which is an object recognition and positioning algorithm based on a deep neural network. The embodiment of the present invention uses a YOLO network to roughly locate the segmented insulator and preliminarily obtain the location information of the segmented insulator.

[0054] In step S103, the image information is segmented according to the position information to obtain image blocks of the segmented insulator.

[0055] The embodiment of the present invention crops the image information according to the position information to obtain image blocks one by one, and then inputs the blocks into the improved UNet3+ network to extract detail features.

[0056] In step S104, an improved UNet3+ network is used to extract detail information of the image block to obtain feature information of the segmented insulator.

[0057] Optionally, the encoder of the improved UNet3+ network replaces the pooling layer with a hole convolution layer with a dilation rate of 2. In neural networks, the most used pooling layer is the maximum pooling layer. Maximum pooling provides a small amount of rotational invariance and slight scale invariance, but maximum pooling will lose a lot of original information. In semantic segmentation, the translation invariance brought by maximum pooling is not needed. What semantic segmentation requires is equivariance, not invariance. A small change in the input will bring about a corresponding small change in the output. In view of this, the embodiment of the present invention improves the UNet3+ network and uses a hole convolution layer with a dilation rate of 2 to replace the pooling layer in the encoder. While reducing the input resolution, it will not cause too much loss of spatial information and implicitly combines multi-scale information.

[0058] Further, as a preferred embodiment of the present invention, Figure 2 A schematic diagram of the structure of the encoder of the improved UNet3+ network provided by an embodiment of the present invention. Figure 2 It can be obtained that the encoder of the improved UNet3+ network first performs a convolution operation on the image block according to the first convolution kernel, and then performs normalization processing and Relu activation function operation to obtain the first feature map. Next, the first feature map is convolved according to the second convolution kernel, and then normalized and Relu activation function operation is performed to obtain the second feature map; synchronously, the second convolution kernel is updated to a convolution kernel with an expansion rate of 2, and the updated convolution kernel is used to perform a convolution operation on the first feature map, and normalization processing and Relu activation function operation are performed to obtain the third feature map after tensor processing. Finally, the second feature map and the third feature map are added to obtain the fourth feature map; the fourth feature map is extracted using a dilation rate of 2 hollow convolution layer, and the representation capability is increased through the SENet channel attention mechanism to obtain the encoder output.

[0059] Optionally, the second convolution kernel is a 3*3 convolution operation.

[0060] In each stage of the encoder, the embodiment of the present invention performs two branch operations, one using a 3*3 ordinary convolution and the other using a dilation rate of 2. Then, the add operation is used to fuse the two scale information and then output. Here, UNet3+ adopts the architecture concept of the combination of encoder and decoder. The multi-scale information fusion in UNet3+ spans the encoder and decoder stages. There is no multi-scale information fusion in the encoder. Drawing on the architecture of the atrous spatial pyramid pooling (ASPP), according to the characteristics of actual needs, the embodiment of the present invention proposes to use the fusion of two scale information, one is the 3*3 convolution used by UNet3+, and the other is the 3*3 convolution with a dilation rate of 2. Since the baseline used is UNet3+, and the multi-scale information has the same number of feature map channels, it has similar feature semantic information. According to the project deployment requirements and the requirements of deployment on lightweight devices, the add operation is selected as the fusion means of multi-scale information. Under the premise of reducing the number of parameters and the amount of calculation, the improved UNet3+ network can achieve better segmentation performance.

[0061] The embodiment of the present invention also adds an attention mechanism to UNet3+. The attention mechanism can be interpreted as a method of allocating available computing resources to the part with the largest amount of information in the signal. The embodiment of the present invention selects channel attention, which can be understood as specifying the content of the neural network. In the architecture of the neural network, each layer of the convolutional network has many convolution kernels according to the input size and the number of channels. Each convolution kernel corresponds to a feature channel. The channel attention is to allocate resources between each convolution channel. Among them, the representative of the channel attention mechanism is SENet, which improves performance by only capturing the most significant attributes of the image for a given task. The embodiment of the present invention adds the SENet module to UNet3+ to introduce the attention mechanism. The SENet module is also lightweight in computation and enhances the representation ability of the network by modeling channel relationships in an efficient computational way.

[0062] Optionally, in the SENet module, the dimension of the input data X is C 1 ×H 1 ×W 1 First, the input data is convolved to obtain the output U with a dimension of C×H×W. The branch operation is performed. One branch first performs a compression operation F(·) on U, and uses the global pooling layer to embed global information with a compression dimension of 1×1×C. Then it is input into the sigmoid function to learn the channel weight, and the obtained weight is assigned to the original information U, adding an attention mechanism to the channel.

[0063] In step S105, an abnormality warning of the segment insulator is performed according to the characteristic information.

[0064] Here, abnormal conditions of the segmented insulator mainly include bolt loosening and wear detection.

[0065] Optionally, as a preferred example of the present invention, the characteristic information of the segmented insulator includes bolt line drawing information.

[0066] The abnormal warning of the segmented insulator according to the characteristic information in step S105 includes:

[0067] In step S201, if there is only one piece of bolt line information, no abnormal warning is performed.

[0068] In step S202, if there are multiple bolt line information, angle information of the bolt line information is calculated, and abnormal warning is performed according to the angle information.

[0069] Here, the embodiment of the present invention draws a straight line on the bolts and fasteners in advance, and the color is based on being distinguishable from the background. Through the above steps S101 to S104, each frame of image information in the surveillance video to be identified also includes the above straight line. First, the image information is input into the target detection network to locate the position of the fasteners and bolts, and then input into the improved Unet3+ network to segment the position information of the straight line. If the bolt is not loose, there is only one segmented straight line, and there is no need to perform an abnormal warning at this time. Otherwise, if there are multiple segmented straight lines, an abnormal warning needs to be performed. Optionally, as a preferred example, the embodiment of the present invention performs an abnormal warning by calculating the angle information of the bolt line drawing information. The angle information of the bolt line drawing information calculated in step S202 includes:

[0070] In step S2021, two drawn lines obtained by splitting the same bolt drawn line are obtained as target drawn lines.

[0071] Here, if the bolt is loose, the straight line drawn on the bolt and the fastener will be split into two lines. The embodiment of the present invention selects a bolt line, and then uses the two lines after the front bolt line is split as the target line. For ease of understanding, Figure 3 A schematic diagram of bolt line drawing and target line drawing provided by an embodiment of the present invention. Figure 3 (a) indicates a straight line pre-drawn on the bolt and fastener. Figure 3 (b) represents the target drawing line.

[0072] In step S2022, the slope between the target lines is calculated, and the cosine value is obtained according to the tangent formula. The formula for calculating the slope of a straight line is: Where (x1, y1) and (x2, y2) represent the coordinates of non-overlapping points on the same target line. A system is established based on the line drawing results and the slopes corresponding to the two lines are calculated, which are denoted as k1 and k2.

[0073] Then according to the tangent formula Get the tangent and cosine values ​​between the two target lines mentioned above

[0074] In step S2023, the arc cosine is taken according to the cosine value to obtain the angle information of the bolt line information.

[0075] The inverse cosine of the cosine value obtained in the above step S2023 is taken, and the obtained angle information is used as the target drawing line, that is, the angle information of the bolt drawing line information, and the angle information is output for abnormal warning.

[0076] Optionally, as a preferred example of the present invention, the characteristic information of the segmented insulator includes a region of interest of a guide plate.

[0077] The abnormal warning of the segmented insulator according to the characteristic information in step S105 includes:

[0078] In step S301, the width information of the region of interest of the guide plate is extracted.

[0079] In step S302, the width information is compared with a preset width threshold range.

[0080] In step S303, if the width information exceeds the width threshold range, an abnormality warning is issued.

[0081] Here, the embodiment of the present invention first inputs the image information into the target detection network to locate the position of the guide plate through the above steps S101 to S104, and then inputs it into the improved Unet3+ network to segment the region of interest of the guide plate, extract the width of the region of interest of the guide plate, and compare it with the preset width threshold range. The preset width threshold range refers to the width range of the guide plate under normal circumstances. If the width of the extracted region of interest of the guide plate exceeds the width threshold range, it is considered that wear occurs and an abnormal warning is issued.

[0082] The segmented insulator anomaly recognition method based on the improved UNet3+ network provided in an embodiment of the present invention introduces deep learning in the segmented insulator anomaly recognition, adopts target detection technology to roughly locate the key position, and adopts the improved Unet3+ network to segment and extract feature information. Among them, the improved Unet3+ network adopts a hole convolution layer with an expansion rate of 2 to replace the pooling layer of the original encoder, which will not cause too much loss of spatial information while reducing the input resolution, and implicitly combines multi-scale information; secondly, at each stage of the encoder, two branch operations are performed, one using a 3* ordinary convolution, and the other using a hole convolution with an expansion rate of 2. The two scale information is fused through the add operation, which can reduce the number of parameters and reduce the amount of calculation, so that the UNet3+ network can achieve better segmentation performance. Finally, the SENet module is added to introduce the attention mechanism, and the channel relationship is modeled in an efficient calculation method to enhance the representation ability of the network. The feature information extracted by the improved UNet3+ network is more accurate, and the abnormal monitoring of the segmented insulator is performed based on the extracted feature information, which effectively solves the problems of insufficient robustness and lack of universality in the prior art when identifying abnormalities of segmented insulators, realizes real-time monitoring of the equipment, improves the efficiency of emergency repairs, saves labor costs, reduces the workload of maintenance staff, and makes the abnormal identification task have clear detection standards, so as to timely discover abnormal information of key parts, and will not affect the safe operation of trains even in bad weather.

[0083] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0084] In one embodiment, the present invention further provides a segmented insulator abnormality identification device based on an improved UNet3+ network, and the segmented insulator abnormality identification device based on an improved UNet3+ network corresponds one-to-one to the segmented insulator abnormality identification method based on an improved UNet3+ network in the above embodiment. Figure 4 As shown, the segmented insulator abnormality identification device based on the improved UNet3+ network includes an acquisition module 31, a positioning module 32, a segmentation module 33, an extraction module 34, and an early warning module 35. The detailed description of each functional module is as follows:

[0085] The acquisition module 31 is used to acquire each frame of image information in the surveillance video to be identified;

[0086] A positioning module 32, configured to roughly locate the segmented insulator in the image information by using a target detection technology to obtain position information of the segmented insulator;

[0087] A segmentation module 33, used to segment the image information according to the position information to obtain image blocks of the segmented insulator;

[0088] An extraction module 34, configured to extract detail features of the image block using an improved UNet3+ network to obtain feature information of the segmented insulator;

[0089] The early warning module 35 is used to issue an abnormal early warning of the segmented insulator according to the characteristic information.

[0090] Optionally, the encoder of the improved UNet3+ network replaces the pooling layer with a hole convolution layer with a dilation rate of 2.

[0091] Optionally, in the encoder of the improved UNet3+ network, a convolution operation is performed on the image block according to a first convolution kernel, and then a normalization process and a Relu activation function operation are performed to obtain a first feature map;

[0092] Performing a convolution operation on the first feature map according to a second convolution kernel, and then performing a normalization process and a Relu activation function operation to obtain a second feature map;

[0093] The second convolution kernel is updated to a convolution kernel with a dilation rate of 2, and the updated convolution kernel is used to perform a convolution operation on the first feature map, and then normalizes and performs a Relu activation function operation to obtain a third feature map after tensor processing;

[0094] Performing an add operation on the second feature map and the third feature map to obtain a fourth feature map;

[0095] The fourth feature map is subjected to feature extraction using a dilation rate of 2 dilated convolutional layer, and the representation capability is increased through the SENet channel attention mechanism to obtain the encoder output.

[0096] Optionally, the second convolution kernel is a 3*3 convolution operation.

[0097] Optionally, the characteristic information of the segmented insulator includes bolt line drawing information;

[0098] The early warning module 35 includes:

[0099] The bolt warning unit is used for not issuing an abnormal warning if there is only one piece of bolt line information; if there are multiple pieces of bolt line information, calculating the angle information of the bolt line information and issuing an abnormal warning according to the angle information.

[0100] Optionally, the bolt warning unit is used to:

[0101] Get the first two bolt lines as the target lines;

[0102] Calculate the slope between the target lines and get the cosine value according to the tangent formula;

[0103] The inverse cosine is taken according to the cosine value to obtain the angle information of the bolt line information.

[0104] Optionally, the characteristic information of the segmented insulator includes a region of interest of a guide plate;

[0105] The early warning module 35 includes:

[0106] A guide plate warning unit, used for extracting width information of an area of ​​interest of the guide plate;

[0107] Comparing the width information with a preset width threshold range;

[0108] If the width information exceeds the width threshold range, an abnormal warning is performed.

[0109] For the specific limitations of the segmented insulator abnormality identification device based on the improved UNet3+ network, please refer to the limitations of the segmented insulator abnormality identification method based on the improved UNet3+ network above, which will not be repeated here. Each module in the above-mentioned segmented insulator abnormality identification device based on the improved UNet3+ network can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0110] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a segmented insulator abnormality identification method based on an improved UNet3+ network is implemented.

[0111] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0112] Obtain each frame of image information in the surveillance video to be identified;

[0113] Using target detection technology to roughly locate the segmented insulator in the image information to obtain position information of the segmented insulator;

[0114] Segmenting the image information according to the position information to obtain image blocks of segmented insulators;

[0115] Using an improved UNet3+ network to extract detail features from the image block, and obtaining feature information of the segmented insulator;

[0116] An abnormality warning of the segment insulator is performed according to the characteristic information.

[0117] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0118] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0119] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for identifying abnormality of a segmented insulator based on an improved UNet3+ network, characterized in that: include: Obtain each frame of image information in the surveillance video to be identified; Using target detection technology to roughly locate the segmented insulator in the image information to obtain position information of the segmented insulator; Segmenting the image information according to the position information to obtain image blocks of segmented insulators; An improved UNet3+ network is used to extract detail features of the image block to obtain feature information of the segmented insulator; the encoder of the improved UNet3+ network replaces the pooling layer with a hole convolution layer with a dilation rate of 2; In the encoder of the improved UNet3+ network, a convolution operation is performed on the image block according to a first convolution kernel, and then a normalization process and a Relu activation function operation are performed to obtain a first feature map; Performing a convolution operation on the first feature map according to a second convolution kernel, and then performing a normalization process and a Relu activation function operation to obtain a second feature map; The second convolution kernel is updated to a convolution kernel with a dilation rate of 2, and the updated convolution kernel is used to perform a convolution operation on the first feature map, and then normalizes and performs a Relu activation function operation to obtain a third feature map after tensor processing; Performing an add operation on the second feature map and the third feature map to obtain a fourth feature map; Using a dilated convolutional layer with a dilation rate of 2 to extract features from the fourth feature map, increasing the representation capability through the SENet channel attention mechanism, and obtaining an encoder output; An abnormality warning of the segment insulator is performed according to the characteristic information.

2. The segmented insulator abnormality identification method based on the improved UNet3+ network according to claim 1 is characterized in that: The second convolution kernel is a 3*3 convolution operation.

3. The segmented insulator abnormality identification method based on the improved UNet3+ network according to claim 1 is characterized in that: The characteristic information of the segmented insulator includes bolt line information; The abnormal warning of the segment insulator according to the characteristic information includes: If there is only one bolt line information, no abnormal warning is issued; If there are multiple bolt line information, the angle information of the bolt line information is calculated, and an abnormal warning is performed according to the angle information.

4. The segmented insulator abnormality identification method based on the improved UNet3+ network according to claim 3 is characterized in that: The angle information for calculating the bolt line information includes: Get the first two bolt lines as the target lines; Calculate the slope between the target lines and get the cosine value according to the tangent formula; The inverse cosine is taken according to the cosine value to obtain the angle information of the bolt line information.

5. The segmented insulator abnormality identification method based on the improved UNet3+ network according to claim 1 is characterized in that: The characteristic information of the segmented insulator includes an area of ​​interest of the guide plate; The abnormal warning of the segment insulator according to the characteristic information includes: Extracting width information of the region of interest of the guide plate; Comparing the width information with a preset width threshold range; If the width information exceeds the width threshold range, an abnormal warning is performed.

6. A sectional insulator abnormality identification device based on an improved UNet3+ network, characterized in that: include: An acquisition module is used to acquire each frame of image information in the surveillance video to be identified; A positioning module, used for roughly positioning the segmented insulator in the image information by using a target detection technology to obtain position information of the segmented insulator; A segmentation module, used for segmenting the image information according to the position information to obtain image blocks of segmented insulators; An extraction module, used for extracting detail features of the image block using an improved UNet3+ network to obtain feature information of the segmented insulator; the encoder of the improved UNet3+ network replaces the pooling layer with a hole convolution layer with a dilation rate of 2; In the encoder of the improved UNet3+ network, a convolution operation is performed on the image block according to a first convolution kernel, and then a normalization process and a Relu activation function operation are performed to obtain a first feature map; Performing a convolution operation on the first feature map according to the second convolution kernel, and then performing normalization processing and Relu activation function operation to obtain a second feature map; updating the second convolution kernel to a convolution kernel with a dilation rate of 2, using the updated convolution kernel to perform a convolution operation on the first feature map, and then performing normalization processing and Relu activation function operation to obtain a third feature map after tensor processing; Performing an add operation on the second feature map and the third feature map to obtain a fourth feature map; Using a dilated convolutional layer with a dilation rate of 2 to extract features from the fourth feature map, increasing the representation capability through the SENet channel attention mechanism, and obtaining an encoder output; The early warning module is used to issue an abnormal early warning of the segmented insulator according to the characteristic information.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for identifying abnormalities of a segmented insulator based on an improved UNet3+ network as described in any one of claims 1 to 5 is implemented.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for identifying abnormalities of a segmented insulator based on an improved UNet3+ network as described in any one of claims 1 to 5 is implemented.

Images