Data enhancement method, device, equipment and medium for transmission line defect detection

Abstract

The data enhancement method, device, equipment, and medium for transmission line defect detection disclosed in the present invention cluster targets in an original image of a transmission line, crop the original image based on the clustering results, and obtain a plurality of collage material images; calculate the illumination characteristics of the original image and each collage material image, select images from the plurality of collage material images based on the calculated illumination characteristics, and embed them into the original image to obtain an enhanced image data set; select a spliced ​​image from the enhanced image data set based on size information and perform image enhancement processing on the original image to obtain the enhanced image. This application solution can enhance data samples, protect key feature details, and improve the recognition accuracy and robustness of target detection in complex power scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of power system transmission line defect detection, and in particular to a data enhancement method, device, equipment and medium for transmission line defect detection. Background Art

[0002] With the continued development of artificial intelligence and deep learning technologies, image recognition-based object detection methods have been widely applied to intelligent inspection tasks in power systems, demonstrating significant advantages in key scenarios such as identifying hidden dangers in transmission lines, autonomous drone inspections, and substation equipment status monitoring. By automatically detecting and identifying target devices, components, or defects in images of power facilities, intelligent inspection systems significantly improve operational efficiency and safety, and have gradually become a key supporting technology for the digital transformation of the power industry.

[0003] However, typical power scenarios, such as transmission lines, often contain numerous small objects with unclear texture features and significant occlusion. These include cracked insulators, loose suspension clamps, fallen bolts, and damaged cable sheaths. The features of these small objects are limited by limited information, blurred edges, and susceptibility to occlusion by complex backgrounds. This results in unclear data samples or the loss of key features, impacting the accuracy of subsequent target detection. Summary of the Invention

[0004] To address the above-mentioned defects, the present invention provides a data enhancement method, apparatus, device and medium for transmission line defect detection, which can enhance data samples and protect key feature details, thereby improving the detection accuracy of subsequent target detection.

[0005] An embodiment of the present invention provides a data enhancement method for power transmission line defect detection, the method comprising:

[0006] Clustering objects in an original image of a power transmission line, and cropping the original image according to the clustering result to obtain a plurality of collage material images;

[0007] Calculating illumination features of the original image and each collage material image, and selecting images from the plurality of collage material images based on the calculated illumination features to embed them into the original image respectively, to obtain an enhanced image data set;

[0008] A spliced ​​image is selected from the enhanced image data set according to the size information and is subjected to image enhancement processing on the original image to obtain the enhanced image.

[0009] Preferably, clustering the objects in the original image of the power transmission line and cropping the original image according to the clustering result to obtain a plurality of collage material images includes:

[0010] Extracting the center coordinates of each target in the original image and expressing them as a set of two-dimensional coordinate points;

[0011] Based on the two-dimensional coordinate point set, a clustering algorithm is used to perform spatial cluster analysis on all targets, and the targets are divided into several clusters according to the spatial distance between the targets;

[0012] For each cluster, generate a cropped region containing all targets in the cluster and their corresponding background information;

[0013] Save the cropped image corresponding to each cluster as a collage material image.

[0014] Preferably, calculating illumination features of the original image and each collage material image, and selecting images from a plurality of collage material images according to the calculated illumination features and embedding them into the original image to obtain an enhanced image dataset comprises:

[0015] Performing a lighting modeling operation on the original image, estimating the brightness of the entire image, and obtaining lighting features of the original image;

[0016] Estimating the brightness of each collage material image to obtain the illumination characteristics of each collage material image;

[0017] Selecting M collage material images with the smallest difference in illumination characteristics from the original image as candidate collage material images; M is an integer greater than zero;

[0018] Performing mask encoding on each cropped area of ​​the original image according to whether a target area exists;

[0019] The candidate collage material images are respectively embedded into the original images according to mask coding to obtain an enhanced image dataset.

[0020] Furthermore, embedding the candidate collage material images into the original images respectively according to the mask coding to obtain an enhanced image dataset includes:

[0021] Selecting material images from the candidate collage material images in sequence, determining a free area according to the mask code; randomly selecting a position of the free area as a pasting center, performing a pasting operation, embedding the material image into the original image, and updating the mask value of the material image embedding area to obtain a corresponding enhanced image;

[0022] The enhanced image corresponding to each material image is used as the enhanced image dataset.

[0023] Preferably, the illumination feature is ;

[0024] in, is the perceived brightness of the pixel in the i-th column and j-th row of the image, H and W represent the width and height of the image respectively, , 、 and are the three channel values ​​of the pixel in the i-th column and j-th row in the image.

[0025] Preferably, the selecting a spliced ​​image from the enhanced image data set according to the size information and performing image enhancement processing on the original image to obtain the enhanced image includes:

[0026] Randomly selecting a number of candidate images from the enhanced image dataset and reading the original size information of each candidate image;

[0027] Calculate the image area of ​​each candidate image and normalize the sizes of all candidate images to obtain the selection weight corresponding to each candidate image;

[0028] Using the weight of each candidate image as a selection probability to select a preset number of stitched images;

[0029] The stitched image and the original image are mosaic-joined, and a composite sample is constructed using a four-square grid layout to obtain the enhanced image.

[0030] Preferably, the selection weight is ;

[0031] in, Indicates the first The selection weight of the candidate images, and Respectively represent the first The width and height of the candidate images, and represent the minimum area and maximum area in the enhanced image dataset respectively.

[0032] An embodiment of the present invention further provides a data enhancement device for power transmission line defect detection, the device comprising:

[0033] a cropping module, configured to cluster objects in an original image of a power transmission line and crop the original image according to the clustering result to obtain a plurality of collage material images;

[0034] a fusion module, configured to calculate illumination features of the original image and each collage material image, and select images from the plurality of collage material images based on the calculated illumination features and embed them into the original image to obtain an enhanced image dataset;

[0035] The enhancement module is used to select a spliced ​​image from the enhanced image data set according to the size information and perform image enhancement processing on the original image to obtain the enhanced image.

[0036] An embodiment of the present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the data enhancement method for transmission line defect detection as described in any one of the above embodiments.

[0037] An embodiment of the present invention also provides a computer-readable storage medium, which includes a stored computer program. When the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the data enhancement method for transmission line defect detection as described in any one of the above embodiments.

[0038] The data enhancement method, device, equipment, and medium for transmission line defect detection provided by the present invention cluster targets in an original image of a transmission line, crop the original image based on the clustering results, and obtain a plurality of collage material images; calculate the illumination characteristics of the original image and each collage material image, select images from the plurality of collage material images based on the calculated illumination characteristics, and embed them into the original image to obtain an enhanced image dataset; and select a spliced ​​image from the enhanced image dataset based on size information and perform image enhancement processing on the original image to obtain the enhanced image. This application solution can enhance data samples, protect key feature details, and improve the recognition accuracy and robustness of target detection in complex power scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 1 is a flow chart of a data enhancement method for power transmission line defect detection provided by an embodiment of the present invention;

[0040] Figure 2 This is another flow chart of the data enhancement method for power transmission line defect detection provided by an embodiment of the present invention;

[0041] Figure 3 is a schematic diagram of a process for cutting collage materials based on spatial clustering according to an embodiment of the present invention;

[0042] Figure 4 is a flowchart of a resampling process based on context fusion provided by an embodiment of the present invention;

[0043] Figure 5 1 is a flow chart of a dynamically weighted mosaic enhancement process provided by an embodiment of the present invention;

[0044] Figure 62 is a schematic structural diagram of a data enhancement device for power transmission line defect detection provided by an embodiment of the present invention;

[0045] Figure 7 It is a structural diagram of a terminal device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0047] See also Figure 1 , is a flow chart of a data enhancement method for power transmission line defect detection provided by an embodiment of the present invention, the method comprising steps S1 to S3:

[0048] Step S1, clustering objects in an original image of a power transmission line, and cropping the original image according to the clustering result to obtain a plurality of collage material images;

[0049] Step S2, calculating illumination features of the original image and each collage material image, selecting images from a plurality of collage material images based on the calculated illumination features and embedding them into the original image respectively, to obtain an enhanced image dataset;

[0050] Step S3: Select a spliced ​​image from the enhanced image data set according to the size information and perform image enhancement processing on the original image to obtain the enhanced image.

[0051] When implementing this embodiment, see Figure 2 , is another flow chart of a data enhancement method for power transmission line defect detection provided by an embodiment of the present invention. The context-preserving and feature-protecting data enhancement system proposed in this invention primarily comprises three core modules: a spatial clustering-based collage material cropping module, a context-fusion-based resampling module, and a dynamically weighted mosaic enhancement module. The implementation of this method includes the following steps:

[0052] Small targets in the transmission line image are detected and their positions are extracted. A clustering algorithm is used to divide spatially adjacent small targets into clusters. A cropping region is generated based on each cluster. Image segments containing multiple small targets and their context information are cropped from the original image as collage material images. The specific process is as follows: Figure 2 As shown;

[0053] Extract the original image's illumination feature Gamma and select cropped image materials with consistent illumination. Select the appropriate collage position based on the region mask information and embed the collage image into the original image to construct an enhanced image with consistent semantics and reasonable structure.

[0054] When constructing Mosaic enhancement samples, the original size information is read from the candidate image set, the weights are calculated and images are selected for stitching to alleviate the scaling distortion of small objects.

[0055] This application constructs a collage material cropping module based on spatial clustering to effectively preserve the spatial relationship and semantic consistency between the target and its original background. This overcomes the semantic conflict problem caused by the random pasting of target positions in traditional copy-and-paste methods, significantly improving the authenticity and effectiveness of the enhanced samples. The proposed context fusion resampling module combines the illumination matching mechanism with the region mask constraint strategy to ensure the consistency of the visual style and spatial layout of the collage material, effectively alleviating the common problems of illumination inconsistency and target occlusion in existing enhancement methods, and improving data quality and model training stability.

[0056] In another embodiment of the present invention, the process of performing the cutting process in step S1 specifically includes:

[0057] See also Figure 3 , is a flowchart of a collage material cutting process based on spatial clustering provided by an embodiment of the present invention.

[0058] Extract the location information of all targets from the input original image and obtain the center coordinates of each target; given a data sample from a transmission line inspection image , extract the center coordinates of each target in the image and represent them as a set of two-dimensional coordinate points. This set records the spatial distribution of all targets in the image and serves as the basic data for subsequent spatial clustering analysis.

[0059] Based on this location information, a clustering algorithm is used to spatially cluster the objects in the image, dividing spatially adjacent objects into several clusters. A K-means clustering algorithm is then used to perform spatial cluster analysis on all objects. The algorithm divides the objects into several clusters based on their spatial distance, with each cluster containing several spatially adjacent objects.

[0060] For each cluster, a cropping region is generated to cover all objects in the cluster and their surrounding background.

[0061] Based on the cropped area, the corresponding image fragments are extracted as collage material images to ensure that multiple targets and their original background information are included at the same time; based on the extracted target center coordinates, for each cluster, its cropped area is generated, and the cropped area contains all small targets in the cluster and their corresponding background information.

[0062] The cropped images corresponding to each cluster are saved as collage material images for subsequent context fusion and stitching operations. These images not only preserve the object itself, but also fully retain its original spatial environment, effectively enhancing the semantic consistency and rationality of the object-background relationship during the data augmentation process.

[0063] The resulting cropped images are resized and annotated with attributes to serve as data input for the subsequent context fusion and mosaic enhancement modules.

[0064] This case uses spatial clustering-based collage material cropping to extract the spatial positions of small targets in the original image. A clustering algorithm is then used to spatially cluster the small targets to generate regional clusters containing multiple targets. A cropping area is then generated for each cluster, and an image area containing multiple small targets and their original background is extracted from the original image as collage material, ensuring that the semantic consistency and contextual information between the target and the background are preserved during the enhancement process.

[0065] In another embodiment of the present invention, the process of determining the enhanced image set in step S2 specifically includes:

[0066] See also Figure 4 , is a flowchart of the context fusion-based resampling process provided by an embodiment of the present invention.

[0067] Model the illumination characteristics of the original image and use the Gamma brightness calculation formula to obtain the global average perceived brightness of the image to characterize the overall illumination distribution;

[0068] Based on the illumination distribution, several cropped images with illumination conditions close to the original image are selected from the cropped image material set to ensure the brightness consistency between the target and the background in the subsequent collage process; the extracted collage material image set is input into the illumination matching module, the brightness value Gamma of each image is calculated respectively, and the collage materials are sorted according to the brightness difference with the original image, and the one with the smallest difference is selected. The image is used as a candidate collage material to ensure that the target area and the background match in terms of lighting.

[0069] The extracted cropped area is used to construct a region mask of the original image, where the position with a value of 1 indicates the existing target area and cannot be covered, and the position with a value of 0 indicates the free area and can be used to paste the enhanced target;

[0070] The candidate collage material images are respectively embedded into the original images according to the mask coding to enhance the image dataset.

[0071] A context-fusion-based resampling module is used to coordinate the lighting and spatial position between the collage image and the original image during the data augmentation process. First, the gamma brightness value of the original image is calculated to quantify the lighting distribution, and the image with the closest brightness characteristics to the original image is screened from the collage material set based on lighting similarity. Then, a region mask is used to construct an available region map of the original image to constrain the target pasting position. The collage image is embedded in the original image without obstructing the existing target, ensuring the coordination of the enhanced sample in terms of lighting, structure and spatial layout.

[0072] In another embodiment of the present invention, the candidate collage material images are respectively embedded into the original images according to mask coding to obtain an enhanced image dataset, specifically:

[0073] According to the location of the existing target in the original image, a region mask is generated to mark the target area and mark the remaining area as an idle area;

[0074] The target cropped area is selected from the illumination-matched material, and the position that does not overlap with the original target is randomly selected in the mask image as the collage point, and the filtered cropped image is embedded into the original image; the position where the region mask is 0 is randomly selected as the pasting center, and the pasting operation is performed to embed the material image into the original image to construct an enhanced image with semantic consistency; after each collage operation, the region mask is updated to mark the newly occupied area to avoid multiple coverage.

[0075] After each pasting is completed, the area mask is updated in real time, and the mask value of the area covered by the newly added target is updated from 0 to 1 to avoid overlap or occlusion between target areas and ensure the rationality of the pasting operation in spatial layout;

[0076] The enhanced images corresponding to each material image are used as the enhanced image dataset, and the generated fused image is used as the intermediate enhancement result for subsequent dynamic weighted mosaic splicing operations to further improve data diversity.

[0077] In another embodiment of the present invention, in the context fusion resampling module, the illumination modeling operation is first performed on the original image, the brightness of the entire image is estimated, and its global illumination distribution characteristics are obtained, which are recorded as , used to measure the overall brightness level of the image, the lighting characteristics are ;

[0078] in, is the perceived brightness of the pixel in the i-th column and j-th row of the image, H and W represent the width and height of the image respectively, , 、 and are the three channel values ​​of the pixel in the i-th column and j-th row in the image.

[0079] In another embodiment provided by the present invention, the step S3 specifically includes the following steps:

[0080] See also Figure 5 , is a flowchart of the dynamic weighted Mosaic enhancement process provided by an embodiment of the present invention.

[0081] Randomly preselect several candidate images from the enhanced image dataset and read the original size information of each image;

[0082] Calculate the image area based on the width and height of the image, and normalize the sizes of all candidate images to obtain the selection weight corresponding to each image;

[0083] The image is selected as the material image for Mosaic stitching based on the weight as the selection probability, which reduces the size compression during the scaling process and reduces the risk of feature loss of small objects.

[0084] The three selected images are geometrically transformed together with the current processing image and spliced ​​in a four-square grid layout to construct a composite image sample containing diverse targets.

[0085] Through dynamic weighted Mosaic enhancement, the feature loss of small targets caused by scaling is reduced during the multi-image stitching process. The image area is calculated by reading the original size information of the candidate image, and a weight is assigned to each image based on the size, with the smaller the size, the higher the weight. Then, based on the normalized weights, three images are selected from the candidate set to participate in the Mosaic stitching together with the current image, and a composite sample is constructed using a four-square grid layout. By prioritizing smaller images and reducing their scaling, this module effectively preserves the edge and texture details of small targets, enhancing the model's perception and detection performance for tiny targets.

[0086] In another embodiment of the present invention, in the dynamic weighted Mosaic enhancement module, N candidate images are first pre-selected from the enhanced image sample pool. As candidate stitching materials, and read the original size information of each image, including width and height , and then calculate the selection weight of each image:

[0087] ;

[0088] in, Indicates the first The selection weight of the candidate images, and Respectively represent the first The width and height of the candidate images, and represent the minimum area and maximum area in the enhanced image dataset respectively.

[0089] The image weights are normalized and used as the selection probability of the image. Then, three images are selected from the N candidate images according to the selection probability and used together with the current image to participate in the mosaic data enhancement.

[0090] The present invention constructs a collage material cropping based on spatial clustering to effectively preserve the spatial relationship and semantic consistency between the target and its original background, overcoming the semantic conflict problem caused by the random pasting of the target position in the traditional copy-and-paste method, and significantly improving the authenticity and effectiveness of the enhanced samples; context fusion resampling, combined with the illumination matching mechanism and the regional mask constraint strategy, ensures the coordination and consistency of the visual style and spatial layout of the collage material, effectively alleviating the common problems of illumination inconsistency and target occlusion in existing enhancement methods, and enhancing data quality and model training stability; dynamic weighted mosaic enhancement, while preserving data diversity, reduces the edge blur and feature degradation problems of small targets caused by scaling by prioritizing the selection of smaller images for stitching operations, effectively improving the model's detection performance for small targets;

[0091] A large number of experimental verifications were carried out based on real datasets, and the results showed that the method of the present invention is significantly better than the existing technology in target detection tasks on multiple public datasets.

[0092] The embodiment of the present invention also provides a data enhancement device for power transmission line defect detection, see Figure 6 , is a schematic structural diagram of a data enhancement device for power transmission line defect detection provided by an embodiment of the present invention, the device comprising:

[0093] a cropping module, configured to cluster objects in an original image of a power transmission line and crop the original image according to the clustering result to obtain a plurality of collage material images;

[0094] a fusion module, configured to calculate illumination features of the original image and each collage material image, and select images from the plurality of collage material images based on the calculated illumination features and embed them into the original image to obtain an enhanced image dataset;

[0095] The enhancement module is used to select a spliced ​​image from the enhanced image data set according to the size information and perform image enhancement processing on the original image to obtain the enhanced image.

[0096] It should be noted that the data enhancement device for transmission line defect detection provided in the embodiment of the present invention can execute the data enhancement method for transmission line defect detection described in any of the above embodiments, and the specific functions of the data enhancement device for transmission line defect detection are not described here.

[0097] See also Figure 7 , is a schematic diagram of the structure of a terminal device provided by an embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data enhancement program for power transmission line defect detection. When the processor executes the computer program, the steps of each of the above-mentioned data enhancement methods for power transmission line defect detection are implemented, such as Figure 1 Alternatively, the processor executes the computer program to implement the functions of the modules in the above-mentioned device embodiments.

[0098] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing functions, which describe the execution process of the computer program in the terminal device. For example, the computer program may be divided into various modules, and the specific functions of each module are not described in detail here.

[0099] The terminal device may be a computing device such as a desktop computer, laptop, PDA, or cloud server. The terminal device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the schematic diagram is merely an example of a terminal device and does not limit the terminal device. The terminal device may include more or fewer components than shown, or a combination of certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, etc.

[0100] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0101] The memory can be used to store the computer programs and / or modules. The processor implements the various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and accessing the data stored in the memory. The memory may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function or an image playback function); the data storage area may store data generated based on the use of the mobile phone (such as audio data, a phone book, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0102] If the module / unit integrated into the terminal device is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can also implement all or part of the process steps in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, software distribution medium, etc.

[0103] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A data enhancement method for transmission line defect detection, characterized in that: The method comprises: Clustering objects in an original image of a power transmission line, and cropping the original image according to the clustering result to obtain a plurality of collage material images; Calculating illumination features of the original image and each collage material image, and selecting images from the plurality of collage material images based on the calculated illumination features to embed them into the original image respectively, to obtain an enhanced image data set; Selecting a spliced ​​image from the enhanced image dataset according to size information and performing image enhancement processing on the original image to obtain the enhanced image; The step of calculating illumination features of the original image and each collage material image, and selecting images from the plurality of collage material images according to the calculated illumination features and embedding them into the original image to obtain an enhanced image dataset comprises: Performing a lighting modeling operation on the original image, estimating the brightness of the entire image, and obtaining lighting features of the original image; Estimating the brightness of each collage material image to obtain the illumination characteristics of each collage material image; Selecting M collage material images with the smallest difference in illumination characteristics from the original image as candidate collage material images; M is an integer greater than zero; Performing mask encoding on each cropped area of ​​the original image according to whether a target area exists; The candidate collage material images are respectively embedded into the original images according to mask coding to obtain an enhanced image dataset.

2. The data enhancement method for power transmission line defect detection according to claim 1, characterized in that: The objects in the original image of the power transmission line are clustered, and the original image is cropped according to the clustering result to obtain a plurality of collage material images, including: Extracting the center coordinates of each target in the original image and expressing them as a set of two-dimensional coordinate points; Based on the two-dimensional coordinate point set, a clustering algorithm is used to perform spatial cluster analysis on all targets, and the targets are divided into several clusters according to the spatial distance between the targets; For each cluster, generate a cropped region containing all targets in the cluster and their corresponding background information; Save the cropped image corresponding to each cluster as a collage material image.

3. The data enhancement method for power transmission line defect detection according to claim 1, wherein: The step of embedding the candidate collage material images into the original images according to the mask coding to obtain an enhanced image dataset includes: Selecting material images from the candidate collage material images in sequence, determining a free area according to the mask code; randomly selecting a position of the free area as a pasting center, performing a pasting operation, embedding the material image into the original image, and updating the mask value of the material image embedding area to obtain a corresponding enhanced image; The enhanced image corresponding to each material image is used as the enhanced image dataset.

4. The data enhancement method for power transmission line defect detection according to claim 1, wherein: The illumination characteristics are ; in, is the perceived brightness of the pixel in the i-th column and j-th row of the image, H and W represent the width and height of the image respectively, , 、 and are the three channel values ​​of the pixel in the i-th column and j-th row in the image.

5. The data enhancement method for power transmission line defect detection according to claim 1, wherein: The step of selecting a spliced ​​image from the enhanced image dataset according to the size information and performing image enhancement processing on the original image to obtain the enhanced image includes: Randomly selecting a number of candidate images from the enhanced image dataset and reading the original size information of each candidate image; Calculate the image area of ​​each candidate image and normalize the sizes of all candidate images to obtain the selection weight corresponding to each candidate image; Using the weight of each candidate image as a selection probability to select a preset number of stitched images; The stitched image and the original image are mosaic-joined, and a composite sample is constructed using a four-square grid layout to obtain the enhanced image.

6. The data enhancement method for power transmission line defect detection according to claim 1, wherein: The selection weight is ; in, Indicates the first The selection weight of the candidate images, and Respectively represent the first The width and height of the candidate images, and represent the minimum area and maximum area in the enhanced image dataset respectively.

7. A data enhancement device for power transmission line defect detection, characterized in that: The device comprises: a cropping module, configured to cluster objects in an original image of a power transmission line and crop the original image according to the clustering result to obtain a plurality of collage material images; a fusion module, configured to calculate illumination features of the original image and each collage material image, and select images from the plurality of collage material images based on the calculated illumination features and embed them into the original image to obtain an enhanced image dataset; An enhancement module, configured to select a spliced ​​image from the enhanced image dataset according to size information and perform image enhancement processing on the original image to obtain the enhanced image; The step of calculating illumination features of the original image and each collage material image, and selecting images from the plurality of collage material images according to the calculated illumination features and embedding them into the original image to obtain an enhanced image dataset comprises: Performing a lighting modeling operation on the original image, estimating the brightness of the entire image, and obtaining lighting features of the original image; Estimating the brightness of each collage material image to obtain the illumination characteristics of each collage material image; Selecting M collage material images with the smallest difference in illumination characteristics from the original image as candidate collage material images; M is an integer greater than zero; Performing mask encoding on each cropped area of ​​the original image according to whether a target area exists; The candidate collage material images are respectively embedded into the original images according to mask coding to obtain an enhanced image dataset.

8. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for data enhancement of power transmission line defect detection according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the data enhancement method for power transmission line defect detection according to any one of claims 1 to 6.

Images