Infrared detection method and system for abnormal heating of composite insulated cross arm

Abstract

The invention discloses an infrared detection method and system for abnormal heating of a composite insulation cross arm, and the method comprises the steps: obtaining the infrared image data of the composite cross arm, and carrying out the enhancement processing of the infrared image data of the composite cross arm, and generating a first input image; performing feature extraction on the first input image to obtain first feature data, and performing extraction and conversion on the first feature data to generate second feature data; performing sampling processing on the second feature data, and performing multi-scale feature extraction on the sampled second feature data to obtain target features under different scales; and performing decoupling processing on the target features to obtain an infrared detection result of abnormal heating of the composite insulated cross arm. According to the invention, the safety and economy of the power grid can be improved, and the operation and maintenance work of the power grid becomes safer and more efficient.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment fault detection, and in particular to an infrared detection method and system for abnormal heating of a composite insulating cross arm. Background Art

[0002] Composite insulated crossarms are an important part of the power system. They are used between transmission lines and transmission towers to play the role of insulator strings. They can effectively reduce the width of the transmission line corridor and the height of the tower, and reduce the cost of line construction. Their safety and stability determine the safe and stable operation of the power system. Timely detection of abnormal heating of composite crossarms can effectively detect abnormalities of composite crossarms in advance. At present, the inspection mode of composite crossarms is that line patrol personnel use drones and handheld infrared cameras to shoot composite crossarms at a distance, and use manual judgment to determine abnormal heating points from infrared images. However, this method has low inspection efficiency and consumes a lot of personnel resources. In the case of more infrared images, misjudgment and missed judgment may occur. At present, with the development of deep learning and artificial intelligence technology, the detection method of using deep learning technology to identify and judge infrared images has gradually become the mainstream.

[0003] At present, there are few studies on the detection algorithm of abnormal heating of composite crossarms, but there are many studies on the algorithm for identifying abnormal heating of power equipment. Since the infrared cameras mounted on drones and the infrared cameras used by inspection personnel have low resolution and most composite crossarms are far away from the cameras, the pixels of the abnormal heating points of the composite crossarms are small, which poses a great challenge to the algorithm's recognition ability of small targets and low-resolution targets. The algorithm for detecting abnormal heating of other power equipment cannot be directly used. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides an infrared detection method and system for abnormal heating of a composite insulating crossarm to solve the problems mentioned in the background technology.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, an embodiment of the present invention provides an infrared detection method for abnormal heating of a composite insulating crossarm, comprising: acquiring infrared image data of a composite crossarm, and performing enhancement processing on the infrared image data of the composite crossarm to generate a first input image;

[0009] Extracting features from the first input image to obtain first feature data, and extracting and converting the first feature data to generate second feature data;

[0010] Performing sampling processing on the second feature data, and performing multi-scale feature extraction on the second feature data after the sampling processing to obtain target features at different scales;

[0011] The target features are decoupled to obtain infrared detection results of abnormal heating of composite insulation crossarms.

[0012] As a preferred solution of the infrared detection method for abnormal heating of composite insulating crossarms described in the present invention, wherein: enhancing the infrared image data of the composite crossarm to generate a first input image includes: randomly scaling each target composite crossarm infrared image according to a predetermined scaling ratio, and randomly cropping the scaled image, and splicing the scaled and cropped images to generate the first input image.

[0013] As a preferred scheme of the infrared detection method for abnormal heating of composite insulating crossarms described in the present invention, wherein: feature extraction of the first input image to obtain first feature data includes: feature extraction of the first input image by a space-to-depth convolution module consisting of a space-to-depth layer and a strideless convolution layer; after the first input image is converted in the space-to-depth layer, the conversion output result is convolved through a strideless convolution layer to generate second feature data.

[0014] As a preferred solution of the infrared detection method for abnormal heating of composite insulating crossarms described in the present invention, the first input image is converted in the space-to-depth layer, including:

[0015] The space-to-depth layer converts the first input image from the space domain to the depth domain, and adjacent pixels of the first input image are grouped into the same depth channel to form a new three-dimensional matrix;

[0016] The three-dimensional matrix is ​​downsampled to merge adjacent pixels into one pixel, thereby reducing the resolution of the image; the space-to-depth layer converts the three-dimensional matrix with reduced image resolution into a two-dimensional matrix to obtain downsampled image data.

[0017] As a preferred solution of the infrared detection method for abnormal heating of composite insulating crossarms described in the present invention, it also includes: a funnel ReLU activation function, which is expressed as:

[0018] f(x c ,i,j)=max(x c,i,j .T(x c,i,j )

[0019]

[0020] in, represents a window centered at position (i, j) in the 2D image on the cth channel, Indicates the parameter information shared by this window in the same channel.

[0021] As a preferred solution of the infrared detection method for abnormal heating of composite insulating crossarms described in the present invention, wherein: the second feature data is sampled and processed, and multi-scale feature extraction is performed on the second feature data after the sampling and processing, including: after the second feature data is down-sampled, cyclic extraction and fusion of features at different scales, and the detection capability of the model is enhanced through multi-scale processing;

[0022] In each cycle, after the feature map is extracted and fused, the feature maps of different scales are spliced ​​together through feature fusion operation to generate a multi-scale target feature map.

[0023] As a preferred scheme of the infrared detection method for abnormal heating of composite insulating crossarms described in the present invention, the target features are decoupled to obtain the infrared detection results of abnormal heating of composite insulating crossarms, including: using a decoupling head structure and an improved loss function to perform target detection and classification, and outputting the detection results.

[0024] In a second aspect, the present invention provides an infrared detection system for abnormal heating of a composite insulating crossarm, comprising:

[0025] A data processing module, used for acquiring composite cross-arm infrared image data, and performing enhancement processing on the composite cross-arm infrared image data to generate a first input image;

[0026] A feature extraction and fusion module, used to extract features from the first input image to obtain first feature data, and extract and transform the first feature data to generate second feature data;

[0027] A multi-scale feature extraction and fusion module, used to perform sampling processing on the second feature data, and perform multi-scale feature extraction on the second feature data after the sampling processing to obtain target features at different scales;

[0028] The target feature decoupling module is used to decouple the target features to obtain infrared detection results of abnormal heating of composite insulation crossarms.

[0029] In a third aspect, the present invention provides a computing device, comprising:

[0030] Memory and processor;

[0031] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the infrared detection method for abnormal heating of composite insulating crossarms are implemented.

[0032] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the infrared detection method for abnormal heating of a composite insulating crossarm.

[0033] Compared with the prior art, the present invention has the following beneficial effects: the present invention adopts Mosaic-9 to improve the input end of the YOLOv8 network, increases the number of input features of the algorithm, and enhances the robustness of the algorithm; then introduces SPD-Conv and FReLU activation functions to replace some convolution and activation function layers in the algorithm, significantly improving the recognition accuracy of the algorithm for small targets and low-resolution targets; finally, a small target detection path is added to the network, further improving the detection accuracy of the algorithm for targets of different scales. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0035] Figure 1 A schematic diagram of a method flow of an infrared detection method and system for abnormal heating of a composite insulating crossarm according to an embodiment of the present invention;

[0036] Figure 2 This is an AHDS-YOLO network structure diagram of an infrared detection method and system for abnormal heating of a composite insulating crossarm according to an embodiment of the present invention;

[0037] Figure 3 This is a SPD-Conv principle diagram of an infrared detection method and system for abnormal heating of a composite insulating crossarm according to an embodiment of the present invention;

[0038] Figure 4 It is a ReLU and FReLU principle diagram of a composite insulating cross-arm abnormal heating infrared detection method and system according to an embodiment of the present invention;

[0039] Figure 5 A schematic diagram of the receptive field of a method and system for infrared detection of abnormal heating of a composite insulating crossarm according to an embodiment of the present invention;

[0040] Figure 6 This is a detection effect diagram of a method and system for infrared detection of abnormal heating of a composite insulating crossarm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0042] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0043] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0044] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0045] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0046] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0047] Example 1

[0048] Reference Figure 1 to Figure 5 , is an embodiment of the present invention, which provides an infrared detection method for abnormal heating of a composite insulating cross arm, such as Figure 1 As shown, including:

[0049] S100: Acquire composite cross-arm infrared image data, and perform enhancement processing on the composite cross-arm infrared image data to generate a first input image;

[0050] S200: extracting features from a first input image to obtain first feature data, and extracting and converting the first feature data to generate second feature data;

[0051] S300: performing sampling processing on the second feature data, and performing multi-scale feature extraction on the second feature data after the sampling processing to obtain target features at different scales;

[0052] S400: Decoupling the target features to obtain infrared detection results of abnormal heating of composite insulation crossarms.

[0053] It should be noted that the YOLO series is a representative of the one-stage target detection algorithm. Many algorithms in the series and their improved versions are widely used in the power industry. YOLOv8 is the latest improved version of the YOLO series. Based on the YOLOv5 algorithm, it refers to the ELAN module of the YOLOv7 algorithm in the Backbone part, designs and introduces the C2F module with rich gradient flow; in the Neck part, the PANet structure is lightened and the head and tail convolution connection layers are removed; in the Head part, the decoupled head structure is introduced to realize the separation of the detection head and the classification structure, and the Anchor-Based structure is replaced by the Anchor-Free structure, and the classification focus loss (Distribution Focal Loss, DFL) is introduced in the Loss function to cooperate with the CIoU loss function to calculate the loss function of the model. Through a series of improvements, YOLOv8 has achieved a significant improvement in accuracy compared with YOLOv5, and its excellent performance is suitable for composite cross-arm operation and maintenance detection.

[0054] In order to meet the demand for accurate detection of infrared images of abnormal heating of composite crossarms, this application improves multiple components of YOLOv8 to form an AHDS-YOLO network, whose network structure is as follows: Figure 2 shown.

[0055] In an embodiment of the present application, enhancing the composite cross-arm infrared image data to generate a first input image includes: randomly scaling each target composite cross-arm infrared image according to a predetermined scaling ratio, and randomly cropping the scaled images, and splicing the scaled and cropped images to generate the first input image.

[0056] It should be noted that the Mosaic-4 data enhancement method used in the original YOLOv8 refers to the CutMix data enhancement method in principle and improves it. The principle is to randomly select four input images, randomly scale, crop, arrange, etc., and then splice them into one image before inputting it into the network, thereby reducing the amount of input calculations while improving the background detail information of the image, thereby achieving the effect of improving the robustness of the network.

[0057] In view of the limitation of hardware resources of edge devices and the demand for identifying small target defects or complex background defects in transmission lines, the embodiment of the present application introduces a Mosaic-9 data enhancement method to process input data at the input end. The Mosaic-9 data enhancement method optimizes the Mosaic-4 data enhancement method of the original YOLOv8, and introduces more input features, background information and target scales, so as to solve the problem that the hardware resources of the edge are limited and a large number of images cannot be processed efficiently and quickly. At the same time, more background and detail information is input during training to more effectively improve the generalization ability of the algorithm.

[0058] For example, 9 images are randomly selected from the input data at the input end of the network, randomly scaled and cropped, and then spliced ​​into one image and scaled to a standard size, and finally sent to the network. This operation method allows the GPU to process 9 images at the same time, which can better solve the problem of limited hardware resources at the edge and the inability to process a large number of images efficiently and quickly, while inputting more background and detail information during training, more effectively improving the generalization ability of the algorithm.

[0059] It should be noted that the current convolutional neural network design system is aimed at scenes with high resolution and moderately sized target objects. The introduction of convolution step and pooling convolution layers can skip redundant image information and will not affect the model's learning of image features. However, when the image resolution is low or the target object is extremely small, the introduction of convolution step for skip convolution or pooling will lead to serious information loss and insufficient features for learning, resulting in a significant drop in recognition accuracy. Therefore, a new module is needed to extract feature information.

[0060] The YOLOv8 algorithm is an improvement on the YOLOv5 algorithm. As mentioned above, it is optimized for large data sets such as COCO and VOC. For the low-resolution and small-target situations in the composite cross-arm infrared images collected at the inspection site, there is a problem of low recognition accuracy, and it is difficult to achieve accurate recognition and positioning. To address the above problems, this application introduces the SPD-Conv module in the network, replaces some convolutional layers and pooling layers in the network, and downsamples the features of the image data, which can reduce the size of the spatial dimension and increase the size of the channel dimension.

[0061] The SPD-Conv module consists of a space-to-depth layer and a strideless convolutional layer. The input image is first converted in the SPD layer, and then the output result is convolved through the strideless convolutional layer. The SPD layer uses Figure 3 The spatial mapping method (Space-to-Depth) [Frame-recurrent video super-resolution] of the inverse Sub-Pixel method shown downsamples the features of the image data, which can reduce the size of the spatial dimension and increase the size of the channel dimension.

[0062] In an embodiment of the present application, feature extraction of a first input image to obtain first feature data includes: feature extraction of the first input image by a space-to-depth convolution module consisting of a space-to-depth layer and a strideless convolution layer; after the first input image is converted in the space-to-depth layer, a convolution operation is performed on the conversion output result through the strideless convolution layer to generate second feature data.

[0063] In an embodiment of the present application, converting the first input image in the space-to-depth layer includes:

[0064] The space-to-depth layer converts the first input image from the space domain to the depth domain. Adjacent pixels of the first input image are grouped into the same depth channel to form a new three-dimensional matrix.

[0065] The three-dimensional matrix is ​​downsampled to merge adjacent pixels into one pixel, thereby reducing the resolution of the image; the space-to-depth layer converts the three-dimensional matrix with reduced image resolution into a two-dimensional matrix to obtain the downsampled image data.

[0066] It should be noted that convolutional neural networks are generally composed of convolutional layers and activation function layers. The convolutional layers are responsible for obtaining spatial dependencies, while the activation function layers are responsible for adding nonlinear capabilities to the model. The activation functions used in the current common activation function layers are generally scalar activation functions or context conditional activation functions. The former refers to activation functions whose input and output are both single quantities. Among them, the widely used ones include: ReLU, Leaky ReLU, PReLU and other activation functions whose negative semi-axis completely depends on samples for adaptation; the latter refers to activation functions with multiple input quantities and a single output quantity, which is represented by the MaxOut function with a large number of parameters. Although there are many types of activation functions currently, they are all simple nonlinear activation functions and do not have the ability to adaptively obtain the context of 2D images. This leads to the spatial insensitivity of the activation function, making it difficult to improve the accuracy of the algorithm model.

[0067] In order to solve the above problems and improve the detection accuracy of the algorithm, this application introduces the funnel ReLU (Funnel ReLU, FReLU). The FReLU function can provide the ability to extract pixel-level spatial structure, solve the problem that the activation function is insensitive to space, and can effectively improve the feature extraction ability of the model with very little increase in the number of parameters, thereby improving the detection accuracy of the model. It uses the space condition with very little computational consumption to extend the ReLU and PReLU activation functions to the activation function with the ability to adaptively obtain the local context of the image. The difference between it and ReLU is as follows Figure 4 shown.

[0068] In the embodiment of the present application, it also includes: a funnel ReLU activation function, expressed as:

[0069] f(x c ,i,j)=max(x c,i,j .T(x c,i,j )

[0070]

[0071] in, represents a window centered at position (i, j) in the 2D image on the cth channel, Indicates the parameter information shared by this window in the same channel.

[0072] It should be noted that the calculation amount of the normal ReLU activation function is shown in the following formula (1), and the calculation amount of the ReLU activation function is shown in the following formula (2):

[0073] ReLU = C × C × K × K = C 2 K 2 (1)

[0074] FReLU=C×C×K×K+1×K×K×1×C=C 2 K 2 +CK 2 (2)

[0075] Among them, C is the number of channels of the feature map of the input and output activation function, and K is the size of the convolution kernel in the convolution layer.

[0076] As shown in equations (1) and (2), the increase in the number of parameters generated by the use of FReLU in the model is only CK 2 , where C is much larger than K, and the added parameters are relative to the original parameters C 2 K 2 The FReLU activation function, through the T(x) spatial condition and the characteristics of the max function, can provide the ability to extract the spatial structure at the pixel level, solving the problem that the activation function is insensitive to space, and can effectively improve the feature extraction ability of the model with very little increase in the number of parameters, thereby improving the detection accuracy of the model.

[0077] In the embodiment of the present application, the second feature data is sampled and processed, and multi-scale feature extraction is performed on the sampled second feature data, including: after the second feature data is down-sampled, cyclically extracting and fusing features at different scales, and enhancing the detection capability of the model through multi-scale processing;

[0078] In each cycle, after the feature map is extracted and fused, the feature maps of different scales are spliced ​​together through feature fusion operation to generate a multi-scale target feature map.

[0079] It should be noted that when taking infrared images of composite crossarms, different angles and distances may lead to great differences in target scales, and in most cases, abnormal hot spots account for a very small proportion of the image. Their features will be lost as the feature map is gradually downsampled, the resolution becomes smaller, and the receptive field gradually increases. The concept of receptive field is as follows: Figure 5 Based on the current three detection heads, this application adds a low-level, high-resolution, small receptive field detection branch to alleviate the disappearance of small-size features caused by target scale changes and improve the model's ability to detect small targets.

[0080] In an embodiment of the present application, the target features are decoupled to obtain infrared detection results of abnormal heating of composite insulating crossarms, including: using a decoupling head structure and an improved loss function to perform target detection and classification, and outputting the detection results.

[0081] The above is a schematic scheme of a composite insulating cross-arm abnormal heating infrared detection method of this embodiment. It should be noted that the technical scheme of this composite insulating cross-arm abnormal heating infrared detection system and the technical scheme of the composite insulating cross-arm abnormal heating infrared detection method belong to the same concept. For details not described in detail in the technical scheme of the composite insulating cross-arm abnormal heating infrared detection system of this embodiment, please refer to the description of the technical scheme of the composite insulating cross-arm abnormal heating infrared detection method.

[0082] In this embodiment, a composite insulating cross-arm abnormal heating infrared detection system includes:

[0083] A data processing module, used for acquiring composite cross-arm infrared image data, and performing enhancement processing on the composite cross-arm infrared image data to generate a first input image;

[0084] A feature extraction and fusion module, used to extract features from the first input image to obtain first feature data, and extract and transform the first feature data to generate second feature data;

[0085] A multi-scale feature extraction and fusion module is used to perform sampling processing on the second feature data, and perform multi-scale feature extraction on the sampled second feature data to obtain target features at different scales;

[0086] The target feature decoupling module is used to decouple the target features and obtain the infrared detection results of abnormal heating of the composite insulation crossarm.

[0087] This embodiment also provides a computing device, which is applicable to the case of the infrared detection method for abnormal heating of a composite insulating crossarm, including:

[0088] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the infrared detection method for abnormal heating of composite insulating crossarms as proposed in the above embodiment.

[0089] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the infrared detection method for abnormal heating of a composite insulating cross-arm as proposed in the above embodiment is implemented.

[0090] The storage medium proposed in this embodiment and the infrared detection method for abnormal heating of composite insulating crossarms proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0091] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccess Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform the methods of various embodiments of the present invention.

[0092] Example 2

[0093] Reference Figure 6 , is an embodiment of the present invention. This embodiment is different from the first embodiment in that this embodiment verifies the beneficial effects of the present invention through implementation examples.

[0094] like Figure 6 As shown, this embodiment uses Mosaic-9 to improve the input end of the YOLOv8 network, increases the number of input features of the algorithm, and enhances the robustness of the algorithm; then introduces SPD-Conv and FReLU activation functions to replace some convolution and activation function layers in the algorithm, which significantly improves the algorithm's recognition accuracy for small targets and low-resolution targets; finally, a small target detection path is added to the network, which further improves the algorithm's detection accuracy for targets of different scales.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting abnormal heating of composite insulating crossarms by infrared, characterized in that: include: Acquire composite cross-arm infrared image data, and perform enhancement processing on the composite cross-arm infrared image data to generate a first input image; Extracting features from the first input image to obtain first feature data, and extracting and converting the first feature data to generate second feature data; Performing sampling processing on the second feature data, and performing multi-scale feature extraction on the second feature data after the sampling processing to obtain target features at different scales; The target features are decoupled to obtain infrared detection results of abnormal heating of composite insulation crossarms.

2. The infrared detection method for abnormal heating of composite insulation crossarms according to claim 1, characterized in that: The step of enhancing the composite cross-arm infrared image data to generate the first input image includes: randomly scaling each target composite cross-arm infrared image according to a predetermined scaling ratio, randomly cropping the scaled images, and splicing the scaled and cropped images to generate the first input image.

3. The infrared detection method for abnormal heating of composite insulation crossarms according to claim 2, characterized in that: Performing feature extraction on the first input image to obtain first feature data includes: performing feature extraction on the first input image using a space-to-depth convolution module consisting of a space-to-depth layer and a strideless convolution layer; after the first input image is converted in the space-to-depth layer, convolution operation is performed on the conversion output result through the strideless convolution layer to generate second feature data.

4. The infrared detection method for abnormal heating of composite insulation crossarms according to claim 3 is characterized in that: The first input image is converted in a space-to-depth layer including: The space-to-depth layer converts the first input image from the space domain to the depth domain, and adjacent pixels of the first input image are grouped into the same depth channel to form a new three-dimensional matrix; The three-dimensional matrix is ​​downsampled to merge adjacent pixels into one pixel, thereby reducing the resolution of the image; the space-to-depth layer converts the three-dimensional matrix with reduced image resolution into a two-dimensional matrix to obtain downsampled image data.

5. The infrared detection method for abnormal heating of composite insulation crossarms according to claim 4, characterized in that: Also includes: The funnel ReLU activation function is expressed as: f(x c ,i,j)=max(x c,i,j .T(x c,i,j ) in, represents a window centered at position (i, j) in the 2D image on the cth channel, Indicates the parameter information shared by this window in the same channel.

6. The infrared detection method for abnormal heating of composite insulation crossarms according to claim 5, characterized in that: Sampling the second feature data, and performing multi-scale feature extraction on the second feature data after the sampling process includes: after downsampling the second feature data, cyclically extracting and fusing features at different scales, and enhancing the detection capability of the model through multi-scale processing; In each cycle, after the feature map is extracted and fused, the feature maps of different scales are spliced ​​together through feature fusion operation to generate a multi-scale target feature map.

7. The infrared detection method for abnormal heating of composite insulating crossarms according to claim 6, characterized in that: The target features are decoupled to obtain the infrared detection results of abnormal heating of composite insulation crossarms, including: using a decoupling head structure and an improved loss function to perform target detection and classification, and outputting the detection results.

8. An infrared detection system for abnormal heating of composite insulation crossarms, characterized in that: include: A data processing module, used for acquiring composite cross-arm infrared image data, and performing enhancement processing on the composite cross-arm infrared image data to generate a first input image; A feature extraction and fusion module, used to extract features from the first input image to obtain first feature data, and extract and transform the first feature data to generate second feature data; A multi-scale feature extraction and fusion module, used to perform sampling processing on the second feature data, and perform multi-scale feature extraction on the second feature data after the sampling processing to obtain target features at different scales; The target feature decoupling module is used to decouple the target features to obtain infrared detection results of abnormal heating of composite insulation crossarms.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the infrared detection method for abnormal heating of composite insulating crossarms as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the infrared detection method for abnormal heating of a composite insulating crossarm as claimed in any one of claims 1 to 7.

Images