Method and device for removing duplication of components in inspection image in power transmission line gear

Abstract

The invention discloses a method and a device for de-weighting components in an inspection image in a power transmission line file. The method comprises the following steps: acquiring the inspection image in the power transmission line file; constructing an adjacent image pair according to the name of the inspection image in the power transmission line file; inputting the adjacent image pairs into a pre-trained identification network, and outputting an identification result; performing global motion compensation on adjacent image pairs by adopting an optical flow method; calculating the similarity according to the recognition result and the adjacent image pairs after global motion compensation; performing layered step-by-step judgment on the components in the adjacent image pairs, and determining repeated components in the adjacent image pairs; according to the repeated parts, the adjacent image pairs are subjected to duplicate removal, so that an accurate, efficient and lightweight duplicate removal technology capable of adapting to the characteristics of the inspection image can be provided, redundant data in the inspection image in the power transmission line gear is reduced, and the cost is reduced.

Description

TECHNICAL FIELD

[0001] The embodiment of the present application relates to the technical field of computer vision and digital image processing, and particularly relates to a component deduplication method and device for transmission line section inspection images. BACKGROUND

[0002] With the wide application of unmanned aerial vehicles, high-altitude cameras and other devices in transmission line inspection, a large amount of sequence image data is generated. These images usually take large-format pictures of line section components, and the following technical pain points exist: (1) Redundant data caused by repeated shooting: In order to ensure coverage, there are a large number of transmission line components that are repeatedly shot between adjacent images, resulting in the same component being captured multiple times and generating a large amount of redundant data, which greatly increases the burden and cost of storage, transmission and subsequent intelligent analysis.

[0003] (2) Existing deduplication methods are not adaptable: Traditional image deduplication methods are sensitive to content and cannot adapt to changes in the same component under different angles, lighting and slight occlusion. General object tracking algorithms are too heavy and are not suitable for jump, non-continuous video frames, i.e. sequences composed of a series of independently shot high-definition pictures.

[0004] (3) Serious interference in complex scenes: The background of the inspection image is complex, including buildings, vegetation, mountains, etc., and the camera moves and shakes, resulting in changes in the apparent features and positions of the same component in different images, further increasing the difficulty of accurate deduplication.

[0005] Therefore, there is an urgent need for a specialized deduplication technology that is accurate, efficient, lightweight and can adapt to the characteristics of inspection images to solve the above problems. SUMMARY

[0006] The embodiment of the present application provides a component deduplication method for transmission line section inspection images, which provides an accurate, efficient, lightweight and adaptable deduplication technology to the characteristics of inspection images, reduces the redundant data in transmission line section inspection images and reduces costs. The method comprises: Obtaining a transmission line section inspection image; According to the name of the transmission line section inspection image, constructing an adjacent image pair; Inputting the adjacent image pair into a pre-trained recognition network to output a recognition result; the recognition result includes a component region recognition result and a background region recognition result; the recognition network is obtained by training a convolutional neural network using a global image carrying a component region mask and a component region image corresponding to the global image; the recognition network is used for feature extraction, feature decoupling and image reconstruction of the adjacent image pair; The global motion compensation module is configured to perform global motion compensation on the adjacent image pair by using an optical flow method to generate a global motion compensated adjacent image pair. The similarity calculation module is configured to calculate a part similarity, a background similarity and a global motion compensated full image similarity according to the recognition result and the global motion compensated adjacent image pair by using a cosine similarity algorithm. The repeated part determination module is configured to determine a repeated part in the adjacent image pair according to the part similarity, the background similarity, the global motion compensated full image similarity and a preset threshold by performing hierarchical and step-by-step judgment on the part in the adjacent image pair. The deduplication module is configured to perform deduplication on the adjacent image pair according to the repeated part.

[0007] Another aspect of the present application also provides a device for removing parts in inspection images in transmission line towers, which provides a precise, efficient, lightweight and adaptive deduplication technology for inspection images, reduces redundant data in the inspection images in transmission line towers and reduces costs. The image acquisition module is configured to acquire the inspection images in transmission line towers. The adjacent image pair construction module is configured to construct an adjacent image pair according to the name of the inspection images in transmission line towers. The region identification module is configured to input the adjacent image pair into a pre-trained identification network to output a recognition result, wherein the recognition result includes a part region identification result and a background region identification result; the identification network is obtained by training a convolutional neural network using a global image carrying a part region mask and a part region image corresponding to the global image; and the identification network is used for feature extraction, feature decoupling and image reconstruction on the adjacent image pair. The global motion compensation module is configured to perform global motion compensation on the adjacent image pair by using an optical flow method to generate a global motion compensated adjacent image pair. The similarity calculation module is configured to calculate a part similarity, a background similarity and a global motion compensated full image similarity according to the recognition result and the global motion compensated adjacent image pair by using a cosine similarity algorithm. The repeated part determination module is configured to determine a repeated part in the adjacent image pair according to the part similarity, the background similarity, the global motion compensated full image similarity and a preset threshold by performing hierarchical and step-by-step judgment on the part in the adjacent image pair. The deduplication module is configured to perform deduplication on the adjacent image pair according to the repeated part.

[0008] The present application also provides a computer device, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the part deduplication method in the inspection images in transmission line towers.

[0009] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the component deduplication method in the transmission line inspection image.

[0010] The embodiment of the present application also provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to realize the component deduplication method in the transmission line inspection image.

[0011] Compared with the image deduplication technical solution in the prior art, the embodiment of the present application obtains the transmission line inspection image, constructs an adjacent image pair according to the name of the transmission line inspection image, inputs the adjacent image pair into a pre-trained recognition network, and outputs a recognition result; the recognition result comprises a component region recognition result and a background region recognition result; the recognition network is obtained by training a convolutional neural network by using a global image carrying a component region mask and a component region image corresponding to the global image; the recognition network is used for feature extraction, feature decoupling and image reconstruction on the adjacent image pair; the optical flow method is used for global motion compensation on the adjacent image pair to generate a globally motion compensated adjacent image pair; the cosine similarity algorithm is used to calculate a component similarity, a background similarity and a globally motion compensated full image similarity according to the recognition result and the globally motion compensated adjacent image pair; and the component in the adjacent image pair is determined to be a repeated component through hierarchical and step-by-step judgment according to the component similarity, the background similarity, the globally motion compensated full image similarity and a preset threshold; and the adjacent image pair is deduplicated according to the repeated component, so that a deduplication technology which is accurate, efficient, light and adaptive to the characteristics of the inspection image can be provided, the redundant data in the transmission line inspection image is reduced, and the cost is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings. In the drawings: Figure 1 A flowchart of the component deduplication method in the transmission line inspection image in the embodiment of the present application; Figure 2 A flowchart of a specific example of the component deduplication method in the transmission line inspection image in the embodiment of the present application; Figure 3 A flowchart of a specific example of the component deduplication method in the transmission line inspection image in the embodiment of the present application; Figure 4 A structural diagram of a component deduplication device in a transmission line section inspection image in an embodiment of the present application; Figure 5 A computer device structural schematic diagram in an embodiment of the present application. DETAILED DESCRIPTION

[0013] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer and more apparent, the embodiments of the present application are further described in detail below with reference to the drawings. Herein, the illustrative embodiments of the present application and their descriptions are used to explain the present application, but not as a limitation of the present application.

[0014] The transmission line inspection image is divided into two parts, section and tower. The section is the part of overhead line between two towers. The embodiments of the present application design a component deduplication method for transmission line section inspection image according to the characteristics of the section image.

[0015] In recent years, with the wide application of unmanned aerial vehicles, high-altitude cameras and other equipment in transmission line inspection, a large amount of sequence image data has been generated. These images are usually taken in large format to shoot the line section equipment. There are several technical pain points. First, repeated shooting leads to data redundancy. In order to ensure coverage, there is a lot of overlapping area between adjacent images, which leads to the same component being captured multiple times, resulting in a large amount of redundant data, greatly increasing the burden and cost of storage, transmission and subsequent intelligent analysis. Second, the existing deduplication method is not adaptable. Traditional image deduplication methods are sensitive to content and cannot adapt to changes in the same component under different angles, lighting and slight occlusion. General target tracking algorithms are too heavy and not suitable for jumping, non-continuous video frames, i.e. sequences composed of a series of independently shot high-definition pictures. Finally, the interference in complex scenes is serious. The inspection image background is complex, containing buildings, vegetation, mountains, etc., and the camera may move or shake, causing the apparent features and positions of the same component in different images to change, further increasing the difficulty of accurate deduplication. Therefore, there is an urgent need for a precise, efficient, lightweight and special deduplication technology that can adapt to the characteristics of inspection images to solve the above problems.

[0016] The application belongs to the technical field of computer vision and digital image processing, and relates to intelligent processing of large-format time sequence images in power system transmission line inspection, in particular to a component repeated detection and data deduplication method and device, electronic equipment and storage medium. The application takes feature decoupling and cross reconstruction network as the core: in the training stage, based on the data set of the subject mask built by the applicant and manually labeled, mask guided consistency, cross reconstruction consistency, cross-path subject consistency and semantic supervision are used for multi-loss collaborative optimization to obtain explicit decoupling representation of the subject and the background; in the reasoning stage, without external mask, stable subject and context feature representation can be automatically formed. Meanwhile, for time sequence jitter and view angle change, global motion estimation and compensation based on feature points are introduced to realize robust alignment of adjacent images; on this basis, a multi-source consistency fusion and deduplication determination process is constructed to comprehensively determine the subject consistency, context cross matching and whole image similarity to complete reliable determination of the same component and effective elimination of repeated samples.

[0017] The application aims to provide a transmission line inspection image component deduplication method for feature decoupling and cross reconstruction network of large-format file image, which realizes accurate identification and deduplication of the target component by introducing dynamic scene perception and compensation technology and multi-source consistency fusion determination. The method first extracts the subject feature and background feature of the component through the feature decoupling and cross reconstruction network, and decouples the target region to avoid the interference of background information on component identification; then, the optical flow method is used to perform global motion compensation on adjacent images to eliminate image differences caused by camera motion and object displacement and ensure alignment of the component region; finally, the multi-source consistency fusion determination strategy calculates the subject feature, context background feature and whole image similarity information, etc. to gradually filter and accurately determine whether the target is the same component, which significantly improves the accuracy and robustness of component deduplication. The application can be widely applied to image deduplication, target detection and other fields, and is especially suitable for complex inspection environments, solving the problem of repeated component determination caused by changes such as light, view angle and occlusion.

[0018] Figure 1 For the flowchart of the component deduplication method in the transmission line inspection image in the embodiment of the application, as shown in Figure 1 , the method comprises: Step 101, acquiring a transmission line inspection image; Step 102, constructing an adjacent image pair according to the name of the transmission line inspection image; Step 103, inputting the adjacent image pair into a pre-trained recognition network to output a recognition result; the recognition result comprises a component region recognition result and a background region recognition result; the recognition network is obtained by training a convolutional neural network using a global image carrying a component region mask and a component region image corresponding to the global image; the recognition network is used for feature extraction, feature decoupling and image reconstruction of the adjacent image pair; Step 104, performing global motion compensation on the adjacent image pair using an optical flow method to generate a global motion compensated adjacent image pair; Step 105, calculating a component similarity, a background similarity and a global motion compensated full image similarity according to the recognition result and the global motion compensated adjacent image pair using a cosine similarity algorithm; Step 106, performing hierarchical and step-by-step judgment on components in the adjacent image pair according to the component similarity, the background similarity, the global motion compensated full image similarity and a preset threshold to determine a repeated component in the adjacent image pair; Step 107, performing deduplication on the adjacent image pair according to the repeated component.

[0019] As shown in the flowchart, Figure 1 It can be known from the flowchart that, compared with the image deduplication technical solution in the prior art, the embodiment of the present application acquires a transmission line file inspection image; constructs an adjacent image pair according to the name of the transmission line file inspection image; inputs the adjacent image pair into a pre-trained recognition network to output a recognition result; the recognition result comprises a component region recognition result and a background region recognition result; the recognition network is obtained by training a convolutional neural network using a global image carrying a component region mask and a component region image corresponding to the global image; the recognition network is used for feature extraction, feature decoupling and image reconstruction of the adjacent image pair; global motion compensation is performed on the adjacent image pair using an optical flow method to generate a global motion compensated adjacent image pair; a component similarity, a background similarity and a global motion compensated full image similarity are calculated according to the recognition result and the global motion compensated adjacent image pair using a cosine similarity algorithm; hierarchical and step-by-step judgment is performed on components in the adjacent image pair according to the component similarity, the background similarity, the global motion compensated full image similarity and a preset threshold to determine a repeated component in the adjacent image pair; deduplication is performed on the adjacent image pair according to the repeated component, which can provide a precise, efficient, lightweight and inspection image feature-adaptive deduplication technology, reduce redundant data in the transmission line file inspection image and reduce costs.

[0020] Figure 2 A flowchart of a specific example of the component deduplication method in the transmission line file inspection image in the embodiment of the present application is shown in Figure 2As shown, the component deduplication method for the transmission line inspection image in the embodiment of the application can include: input: a piece of inspection line shooting data; time sequence adjacent frame intelligent matching; feature decoupling and cross reconstruction network: background and main feature decoupling, multi-level feature reconstruction; dynamic scene perception and compensation: calculating the optical flow between images, estimating the global motion matrix by RANSAC (Random Sample Consensus Algorithm), and motion compensation; multi-source consistency fusion and deduplication determination: main feature matching, context feature matching, and global feature matching after motion compensation; output: the same component encoding mapping table.

[0021] Figure 3 The flowchart of the specific example of the component deduplication method for the transmission line inspection image in the embodiment of the application is shown as Figure 3 As shown, in one embodiment, in step 102, constructing adjacent image pairs according to the names of the transmission line inspection images can include: step 201, parsing the names of the transmission line inspection images to extract the serial numbers in the names; step 202, sorting the transmission line inspection images according to the serial numbers; and step 203, constructing adjacent image pairs according to the sorted transmission line inspection images.

[0022] In this embodiment, after obtaining the transmission line inspection images, time sequence adjacent frame intelligent pairing is performed. By parsing the file name of the image to be processed, the digital serial number in the picture shooting number is extracted, and adjacent image pairs are automatically constructed according to the continuity of the serial numbers to ensure that the comparison only occurs between time sequence adjacent images, greatly improving the image processing efficiency.

[0023] In one embodiment, the recognition network can include a feature extraction module, a feature decoupling module, a feature reconstruction module, and a loss function; the feature extraction module is a backbone network constructed based on a convolutional neural network, used to extract features from adjacent image pairs; the feature decoupling module is used to decouple the features into component features and background features; the feature reconstruction module is used to reconstruct the image through a decoder according to the features, the component features, and the background features to generate a reconstructed image; the loss function is the result of weighted summation of multiple sub-loss functions; and the multiple sub-loss functions are used to optimize the loss function of each part in the recognition network.

[0024] In this embodiment, the recognition network is a feature decoupling and cross reconstruction network for large-format aerial image, aiming to decouple the subject feature and the background feature in the input image, and restore the complete image through the reconstruction module. The feature decoupling and cross reconstruction network for large-format aerial image is used to automatically predict the subject component region and the context background region from the input photographed image. The core innovation lies in that, first, the explicit decoupling representation of the subject feature and the background feature is simultaneously learned in the same framework, and through the cross reconstruction mechanism of shared background and double-path subject, the subject representation is constrained to remain stable and consistent under different viewing angles, scales and brightness conditions; second, a multi-loss collaborative optimization system composed of mask consistency, cross reconstruction consistency, cross-path consistency and semantic supervision is constructed, so that the decoupled subject feature can not only accurately restore the global content, but also retain the component semantics and discriminability, thereby providing a robust and interpretable feature basis for subsequent similarity fusion and deduplication judgment.

[0025] The recognition network includes multiple modules: an input layer, a feature extraction module, a feature decoupling module, a feature reconstruction module and a loss function. During training, the input of the recognition network includes an original image with a component subject region mask annotation. The mask is used to mark the region where the component is located in the image, and through this mask information, the recognition network can concentrate on processing the features of the target component and its surrounding background region; a trained convolutional neural network is used as the backbone network for feature extraction, and the features extracted from the original image are decoupled through the subject feature encoder and the background feature encoder, respectively, to obtain the background feature and the subject feature, which are aligned with the actual image through back propagation learning to restore the details of the image; the network introduces multiple loss functions to ensure that the network learns effective component feature and background feature decoupling and strengthens the reconstruction quality of the image.

[0026] In the training phase, based on the self-built and self-labeled dataset, the input is composed of a global original image with a subject region mask annotation and a corresponding subject region image. The subject mask is used as a strong supervision signal for end-to-end optimization of multiple losses, which is used to accurately constrain the feature decoupling and cross reconstruction of the subject or the background, so that the model learns a restorable and discriminable subject representation under the condition of sharing the background. In the inference phase, without providing any external mask, the input is only the original image to be processed. The network relies on the decoupling ability formed in the training to automatically form the subject component representation and the context background representation, and outputs stable multi-source features for subsequent fusion and deduplication judgment. This process can maintain robustness and generalization to different viewing angles, scales and lighting conditions without relying on external annotations, and provides a more accurate and interpretable feature basis for component deduplication tasks.

[0027] The input data of the input layer includes a global image I global , which contains the input of the entire image; a subject region mask M mask, which marks the location of the target component in the image, usually a binary mask where the target component region is 1 and the rest is 0; the main component image I component The main component region image cropped from the global image. The input image and its mask are processed by a feature extraction network, and the mask is used to constrain the network to learn features in a specific region.

[0028] The feature extraction module is a backbone network based on a convolutional neural network (CNN) and is responsible for extracting high-dimensional features from the global image I global The backbone network uses a pre-trained ResNet-50 model and removes the last fully connected layer to obtain high-dimensional convolutional features. The global image I global and the main region mask M mask , output the extracted feature map F global , whose size is HxWxC, where H and W are the height and width of the image, respectively, and C is the number of channels of the convolutional output.

[0029] In one embodiment, the feature decoupling module includes a component feature encoder and a background feature encoder; the component feature encoder includes a max pooling layer, a convolutional layer, a fully connected layer, and a feature extractor; the component feature encoder is used to decouple the features through the max pooling layer, the convolutional layer, and the fully connected layer to obtain first component features, and extract second component features from the features through the feature extractor; the background feature encoder includes a max pooling layer, a convolutional layer, and a fully connected layer; the background feature encoder is used to decouple the features to obtain background features.

[0030] The feature decoupling module decouples the features of the global image into two parts: component features and background features. This module consists of two independent encoders: a component feature encoder E component , which receives the main region mask M mask of the global image, combines it with the feature map F global , and extracts the first component features F component1 of the main component; a background feature encoder E background , which separates the background region in the global image through M mask , and extracts the background features F backgroundThe component feature encoder uses four convolutional layers to extract features from the target component. The first convolutional layer has a 3×3 kernel size, 64 kernels, and a stride of 1, followed by a max-pooling layer (2×2, stride 2). The next three convolutional layers have 3×3 kernels, with kernel numbers of 128, 256, and 512 respectively, all with a stride of 1, followed by a max-pooling layer. Finally, a fully connected layer outputs a 1024-dimensional feature vector representing the global features of the target component. The background feature encoder has a similar structure to the component feature encoder, but focuses on background region extraction. It also contains four convolutional layers with a 3×3 kernel size, starting with 64 kernels and decreasing to 128, 256, and 512 kernels respectively, with a stride of 1. The pooling layer settings are the same as the component feature encoder. After these convolutional and pooling layers, a fully connected layer outputs a 1024-dimensional feature vector representing the features of the background region surrounding the target component. Note that the main component (Figure I)... component The second component feature F is obtained directly through the feature extractor. component2 They will subsequently participate in feature reconstruction and loss function learning.

[0031] In one embodiment, the feature reconstruction module is used to reconstruct a first reconstructed image by means of a decoder based on the features of a first component and background features; the feature reconstruction module is also used to reconstruct a second reconstructed image by means of a decoder based on the features of a second component and background features; the decoder includes a deconvolution layer, a batch normalization function and an activation function, and the deconvolution layer is used to restore the spatial resolution of the image.

[0032] The feature reconstruction module performs a feature reconstruction on the first decoupled component, F. component1 and background features F background and main component diagram I component The second component feature F is obtained directly through the feature extractor. component2 Reconstruction is performed using a separate decoder module to generate a reconstructed image. Where F... component1 and F background Reconstructed as the first reconstructed image I construct1 F component2 and F background Reconstructed into a second reconstructed image I construct2 The decoder structure includes deconvolutional layers to progressively restore the spatial resolution of the image; batch normalization functions and ReLU activation functions to increase the network's stability and non-linear expressive power. The reconstructed image is subsequently compared with the original image to calculate the loss function.

[0033] The recognition network includes 6 sub-loss functions, each of which is used to optimize different network parts to ensure effective decoupling of part features and background features and high quality of image reconstruction. Ultimately, the 6 sub-loss functions will be combined into one loss function to optimize the training process of the entire network, as follows: (1) Mask-guided subject consistency loss L mask , which constrains the subject features extracted by the subject encoder to align with the semantic of the mask region, avoiding the mixing of subject information and background information, and ensuring that the network learns accurate subject features within the subject region (marked by the mask). This sub-loss function constrains the feature learning of the network in the part region by calculating the difference between the subject features of the mask region and the features of the subject region in the original image. The specific formula is as follows: ; where M mask [i] is the mask region; F global [i] is the feature of the global image; F component1 [i] is the feature of the decoupled subject part.

[0034] (2) Self-reconstruction consistency loss L construct1 , which is the constraint loss of reconstructing the first image (one of the image pairs in the adjacent image pair) and the global image, verifying the reconstructability of F component1 and F background to the global image content, promoting the complementary of the decoupled subject features and background features. It ensures that the decoupled features of the network can be reconstructed into an image consistent with the original image. During training, the network generates a reconstructed image I construct1 , which is compared with the original image to optimize the recovery effect of the image. The specific formula is as follows: ; where I construct1 [i] is the reconstructed image of the first image, and I global [i] is the global image.

[0035] (3) Cross-reconstruction consistency loss L construct2 , which is the constraint loss of reconstructing the second image (the other image in the adjacent image pair) and the global image, ensuring that the background decoupled features of the network can be reconstructed into an image consistent with the original image with the subject features obtained by the feature extractor. During training, the network generates a reconstructed image I construct2 , which is compared with the original image to optimize the recovery effect of the image. The specific formula is as follows: ; where I construct2 [i] is the reconstructed image of the second image.

[0036] (4) Cross-path subject consistency loss L component This loss constrains the subject features of the two paths to remain subject-invariant under different views or scales by calculating the difference between subject feature 1 and subject feature 2, avoiding over-decoupling. The specific formula is as follows: ; Where F conponent1 [i] is the subject feature 1, and F conponent2 [i] is the subject feature 2.

[0037] (5) Subject semantic supervision loss L component1_class , subject feature 1 classification loss, used to optimize the classification accuracy of subject feature 1, and the network learns the class information of the part through cross-entropy loss. The specific formula is as follows: ; Where y[i] is the true label, is the predicted class probability.

[0038] (6) Subject semantic supervision loss L component2_class , subject feature 2 classification loss, used to optimize the classification accuracy of subject feature 2, and the network learns the class information of the part through cross-entropy loss. The specific formula is as follows: ; Where y[i] is the true label, is the predicted class probability.

[0039] Finally, the loss function L total is the result of weighting and combining the above six sub-loss functions. By assigning a weight coefficient a i to each sub-loss function, the contribution of each part can be adjusted according to the task requirements. The expression of the total loss function is as follows: .

[0040] The training process is an end-to-end training through the above loss functions by inputting images with subject region masks and extracted part region maps. During the training process, these loss functions are optimized to gradually improve the accuracy of feature decoupling and the quality of the reconstructed image. In the inference stage, the original image is input, and the network can automatically predict the subject region features and background region features in the image, and reconstruct the target part and background image through the decoder module, and finally output the subject part region and context background region in the original image.

[0041] In an embodiment, in step 104, the global motion compensation for the adjacent image pair is generated by using the optical flow method, which can include: using an ORB (Oriented FAST and Rotated BRIEF) feature point detector using an ORB algorithm to extract feature points in the adjacent image pair; using a brute force matcher to match the feature points by calculating the Hamming distance between the descriptors of the feature points to determine the matching feature points; using a random sample consensus algorithm to calculate the geometric relationship between the matching feature points to obtain an affine transformation matrix; and compensating the coordinates of one image in the adjacent image pair according to the affine transformation matrix, aligning the coordinates after transforming the coordinates of the image into the coordinate system of the other image in the adjacent image pair, and determining the adjacent image pair after the coordinate alignment as the global motion compensated adjacent image pair.

[0042] In the present embodiment, the dynamic scene perception and compensation accurately estimates the motion between adjacent images by using a global motion estimation method based on feature points, so as to perform motion compensation in the image comparison process and eliminate the image differences caused by camera motion or background changes. First, sufficient local feature points are extracted from the input adjacent images for subsequent matching. The ORB algorithm is used for feature point detection and description; and the BFMatcher (brute force matcher) is used to calculate the matching of the feature points in the images according to the ORB descriptors. This step matches by calculating the Hamming distance between the feature points; and the RANSAC (random sample consensus) algorithm is used to estimate the transformation matrix between the matching points to obtain an affine transformation or a similarity transformation matrix. The RANSAC algorithm can effectively eliminate the mismatched points and calculate the transformation matrix that can truly reflect the global motion between the images; and the target in the images is compensated by using the obtained global transformation matrix, the target coordinates are mapped from the second image to the coordinate system of the first image, the offset caused by the camera motion or the background changes is compensated, and the alignment of the target in the two images is ensured. In this way, the subsequent similarity calculation can be performed in the same coordinate system, thereby improving the accuracy of image matching.

[0043] The dynamic scene perception and compensation are used to accurately estimate the motion between adjacent images, so as to perform motion compensation in the image comparison process and eliminate the image differences caused by camera motion or background changes. The feature point detection and matching by using the optical flow method, this step uses the optical flow method to perform global motion estimation by detecting and matching the feature points in the adjacent images to calculate the motion between the images. The matching of the feature points can help to identify the common parts in the images, thereby providing data support for the subsequent motion compensation.

[0044] The feature point detection uses the ORB feature point detector to extract feature points in the image. The ORB algorithm can maintain robustness under rotation, scale transformation and illumination change, and has high computational efficiency, and the specific formula is as follows: kp1, des1 = ORB(I1); kp2, des2 = ORB(I2); Where I1 and I2 are the input adjacent image pair, kp1 and kp2 are the feature point coordinates of the first image and the second image, and des1 and des2 are the descriptors corresponding to the feature point coordinates of the first image and the second image.

[0045] The feature point matching uses the BFMatcher (brute force matcher) to perform feature point matching by calculating the Hamming distance between the descriptors. Through the cross-validation strategy, it is ensured that the matching points are consistent in both directions, and the specific formula is as follows: matches = BFMMatcher(des1, des2); Where des1 and des2 are the descriptors corresponding to the feature point coordinates of the first image and the second image, and matches is the matching point set.

[0046] Global motion estimation and transformation matrix calculation, the RANSAC algorithm is used to calculate the global motion between images, that is, the affine transformation matrix M, which can describe the translation, rotation and scaling of the second image relative to the first image. RANSAC estimates the transformation matrix: by calculating the geometric relationship between the matching feature points, the RANSAC algorithm is used to estimate the affine transformation matrix M: ; Where a, b, c, d are the rotation and scale parameters of the affine transformation, and tx, ty are the translation parameters. The RANSAC algorithm calculates the optimal transformation matrix by randomly sampling and removing outliers, and the formula is as follows: M = RANSAC(matches).

[0047] Coordinate transformation and motion compensation, the affine transformation matrix M is used to transform the target coordinates in the second image to the coordinate system of the first image, and the motion compensation is performed to ensure that the target regions of the images are aligned. Given the target point p2=(x2, y2) in the second image, the affine transformation matrix M is used to transform it to get the new coordinates of the target in the coordinate system of the first image p1=(x1, y1), and the transformation process is as follows: ; Apply the affine transformation matrix M to each target point in the second image to compensate for the coordinates consistent with the first image, and apply it to the subsequent target comparison and similarity calculation.

[0048] In one embodiment, the component region recognition result can include component feature vectors of the adjacent image pair; the background region recognition result can include context background feature vectors of each component of the adjacent image pair; the context background feature vector is a feature vector of a neighboring region located above, below, left and right of the component.

[0049] In the embodiment, in step 105, the component similarity, the background similarity and the global motion compensated full image similarity are calculated according to the recognition result and the global motion compensated adjacent image pair by using the cosine similarity algorithm, which can include: calculating the cosine similarity between the component feature vectors of the two images in the adjacent image pair by using the cosine similarity algorithm, determining the cosine similarity between the component feature vectors of the two images in the adjacent image pair as the component similarity; calculating the cosine similarity between the context background feature vector of one image and the context background feature vector of the other image in the adjacent image pair, determining the maximum value of the calculation result as the background similarity; calculating the cosine similarity between the two images in the global motion compensated adjacent image pair, determining the cosine similarity between the two images in the global motion compensated adjacent image pair as the global motion compensated full image similarity.

[0050] First, the component similarity sim component of the first image and the second image is calculated, which measures the similarity between the component main feature vectors extracted by the deep feature decoupling and reconstruction network. The formula is as follows: ; Where F component1 is the first component feature; F component2 is the second component feature.

[0051] Secondly, the background similarity (context background feature similarity) is determined: The background feature is directly obtained by the deep feature decoupling network prediction, and the background information representation extracted in the four expansion directions (up, down, left and right) of the component position. Due to the forward and upward movement of the camera in the two actual adjacent repeated shooting pictures, the context information of the component in the front and back two pictures changes relative to the position of the component. The same context information with obvious features around the component in the previous shooting picture may appear in the upper, lower, left and right positions of the component in the next shooting picture. Therefore, the context background feature matching is no longer limited to point-to-point matching, but the background features of the same component in the two pictures are matched two by two in the upper, lower, left and right four regions, the maximum value of the matching result is taken, and a higher determination threshold is set. The specific formula is as follows: ; Where the background feature of the first image is F i1 , and the background feature of the second image is Fj2 , i, j ∈ {up, down, left, right}, represents the combination of direction pairs, for example sim up，left is the similarity between the part top background feature of the first image and the part left background feature of the second image. The final similarity of each direction (up, down, left, right) is obtained by selecting the maximum similarity value of each direction matched with other directions from all the calculated similarities: sim contex_max_i = max(sim i，j ).

[0052] For each direction, the maximum similarity of the corresponding direction with other directions is selected, and finally the maximum similarity value of the four directions is obtained, that is, the background similarity: sim contex_max = max(sim contex_max_i ).

[0053] The global motion compensated full image similarity formula is as follows: ; The image contrast after global motion compensation is to further verify whether the targets in the two images belong to the same physical part by calculating the similarity between the first image and the second image after global motion compensation. The cosine similarity is used to measure the similarity between the features of the motion compensated images. Among them, I compensated_1 and I compensated_2 are the first image and the second image after global motion compensation. The similarity value obtained by calculation is compared with the preset threshold value, if the similarity is greater than the threshold value, it is considered that the targets of the two images belong to the same physical part. Through the global motion estimation of the optical flow method, the accuracy of the image contrast can be improved, especially in the case of camera movement.

[0054] In one embodiment, in step 106, according to the part similarity, the background similarity, the global motion compensated full image similarity and the preset threshold, the parts in the adjacent image pair are judged layer by layer, and the determination of the repeated parts in the adjacent image pair can include: when the part similarity is not less than the first preset threshold, the part is determined as a repeated part; when the part similarity is less than the first preset threshold, and the background similarity is not less than the second preset threshold, the part is determined as a repeated part; when the part similarity is less than the first preset threshold, the background similarity is less than the second preset threshold, and the global motion compensated full image similarity is not less than the third preset threshold, the part is determined as a repeated part.

[0055] In this embodiment, multi-source consistency fusion judgment is used to comprehensively integrate various feature similarity information, and different thresholds are set to accurately determine whether the targets in the image pair are the same part. The subject features and context background features obtained through the feature decoupling and cross-reconstruction network are used to determine whether the targets in adjacent images are the same physical part step by step. The multi-level feature similarity collaborative decision module finally combines multiple similarity measures (subject features, context features, full-image features, etc.) to make a part deduplication decision through layer-by-layer screening. This process is determined through three main stages. The step-by-step determination process is as follows: First-level determination: part similarity feature judgment. If sim component ≥ the first preset threshold T1, it is directly determined as a duplicate part.

[0056] Second-level determination: background similarity determination. If the part feature similarity does not reach the threshold, the background similarity sim context_max around the target part is calculated, and whether the targets in the two images are the same part is determined according to the background similarity and the second preset threshold T2.

[0057] Third-level determination: global similarity judgment after global motion compensation. If the similarity of the part features and the background features does not reach the preset threshold, the full-image similarity sim global after global motion compensation is used to further evaluate the similarity.

[0058] Determination rules and threshold settings: The multi-level feature similarity collaborative decision module uses three similarity judgment stages, each of which sets different thresholds: the first preset threshold T1 is set to 0.90, which is used to determine whether the part features are similar enough. If the part similarity sim component of the two images exceeds this threshold, it is determined to be the same part; the second preset threshold T2 is set to 0.85, which is used to determine the background similarity of the background region around the part with significant distinguishable features; the third preset threshold T3 is set to 0.7, which is used for auxiliary judgment. When the part feature and background feature similarity do not pass the threshold, the full-image similarity is used for supplementary judgment.

[0059] Through the hierarchical and step-by-step determination logic, different feature similarities can be gradually screened, improving the accuracy of the decision. At the same time, it can efficiently process large-scale data, avoid redundant calculation, improve the efficiency of part deduplication, and effectively reduce false positives and omissions. By setting adjustable thresholds, the decision engine can flexibly adapt to different image features and is suitable for complex situations such as changes in lighting, occlusion, and angle differences.

[0060] The step integrates multi-source evidences such as subject consistency, context matching and global image similarity, and adopts a process of 'hierarchical threshold + fusion judgment': firstly, subject consistency is taken as the main criterion, and double thresholds for passing or rejecting are set to quickly confirm whether the samples are obviously the same component or obviously different; for the image pairs in the gray area, further combined with context and global image similarity, a comprehensive score is formed, and the final judgment is made according to the calibrated threshold. For the samples still in doubt, multi-scale and expanded context review is started. The mechanism improves the determination robustness under different viewing angles, scales and lighting conditions under the constraint of multi-source information, and realizes accurate identification of the same component and effective rejection of repeated samples.

[0061] The embodiment of the present application also provides a component deduplication device for transmission line patrol images, as described in the following embodiment. Since the principle of the device solves the problem is similar to the component deduplication method for transmission line patrol images, the implementation of the device can be referred to the implementation of the component deduplication method for transmission line patrol images, and the repeated parts will not be described again.

[0062] Figure 4 The structural diagram of the component deduplication device for transmission line patrol images in the embodiment of the present application is shown in Figure 4 The device comprises: An image acquisition module 401 is configured to acquire transmission line patrol images. An adjacent image pair construction module 402 is configured to construct adjacent image pairs according to the names of the transmission line patrol images. A region identification module 403 is configured to input the adjacent image pairs into a pre-trained identification network to output identification results. The identification results include component region identification results and background region identification results. The identification network is obtained by training a convolutional neural network using a global image carrying a component region mask and a component region image corresponding to the global image. The identification network is used for feature extraction, feature decoupling and image reconstruction of the adjacent image pairs. A global motion compensation module 404 is configured to perform global motion compensation on the adjacent image pairs by using an optical flow method to generate globally motion-compensated adjacent image pairs. A similarity calculation module 405 is configured to calculate component similarity, background similarity and globally motion-compensated global image similarity according to the identification results and the globally motion-compensated adjacent image pairs by using a cosine similarity algorithm. A repeated component determination module 406 is configured to determine the components in the adjacent image pairs by hierarchical and step-by-step judgment according to the component similarity, the background similarity, the globally motion-compensated global image similarity and a preset threshold, and determine the repeated components in the adjacent image pairs. A deduplication module 407 is configured to deduplicate the adjacent image pairs according to the repeated components.

[0063] In one embodiment, the adjacent image pair construction module 402 is specifically configured to: resolve the name of the inspection image in the power line file, and extract the serial number in the name; sort the inspection images in the power line file according to the serial number; construct the adjacent image pair according to the sorted inspection images in the power line file.

[0064] In one embodiment, the recognition network comprises a feature extraction module, a feature decoupling module, a feature reconstruction module, and a loss function; The feature extraction module is a backbone network constructed based on a convolutional neural network, configured to extract features from the adjacent image pair; The feature decoupling module is configured to decouple the features into part features and background features; The feature reconstruction module is configured to reconstruct the image through a decoder according to the features, the part features, and the background features, to generate a reconstructed image; The loss function is the result of weighted summation of a plurality of sub-loss functions; the plurality of sub-loss functions are configured to optimize the loss function of each part in the recognition network.

[0065] In one embodiment, the feature decoupling module comprises a part feature encoder and a background feature encoder; The part feature encoder comprises a max-pooling layer, a convolutional layer, a fully connected layer, and a feature extractor; the part feature encoder is configured to decouple the features through the max-pooling layer, the convolutional layer, and the fully connected layer to obtain first part features, and extract second part features from the features through the feature extractor; The background feature encoder comprises a max-pooling layer, a convolutional layer, and a fully connected layer; the background feature encoder is configured to decouple the features to obtain background features.

[0066] In one embodiment, the feature reconstruction module is configured to reconstruct through a decoder according to the first part features and the background features, to generate a first reconstructed image; The feature reconstruction module is further configured to reconstruct through the decoder according to the second part features and the background features, to generate a second reconstructed image; The decoder comprises an inverse convolutional layer, a batch normalization function, and an activation function; the inverse convolutional layer is configured to restore the spatial resolution of the image.

[0067] In one embodiment, the global motion compensation module 404 is specifically configured to: extract feature points in the adjacent image pair using an ORB feature point detector adopting an ORB algorithm; determine the matching feature points by calculating the Hamming distance between the descriptors of the feature points using a brute-force matcher to perform feature point matching; The random sample consensus algorithm is used to calculate the geometric relationship between the matched feature points to obtain an affine transformation matrix. According to the affine transformation matrix, the coordinates of one image in the adjacent image pair are compensated, and after the coordinates of the image are transformed into the coordinate system of the other image in the adjacent image pair, the coordinates are aligned. The adjacent image pair after the coordinate alignment is determined as the adjacent image pair after the global motion compensation.

[0068] In one embodiment, the component region recognition result includes a component feature vector of the adjacent image pair; the background region recognition result includes a context background feature vector of each component in the adjacent image pair; and the context background feature vector is a feature vector of a neighboring region located above, below, left and right of the component. The similarity calculation module 405 is specifically configured to: The cosine similarity algorithm is used to calculate the cosine similarity between the component feature vectors of the two images in the adjacent image pair, and the cosine similarity between the component feature vectors of the two images in the adjacent image pair is determined as the component similarity. The cosine similarity between the context background feature vector of one image in the adjacent image pair and the context background feature vector of the other image is calculated, and the maximum value of the calculation result is determined as the background similarity. The cosine similarity between the two images in the adjacent image pair after the global motion compensation is calculated, and the cosine similarity between the two images in the adjacent image pair after the global motion compensation is determined as the full image similarity after the global motion compensation.

[0069] In one embodiment, the repeated component determination module 406 is specifically configured to: When the component similarity is not less than a first preset threshold, it is determined that the component is a repeated component; When the component similarity is less than the first preset threshold and the background similarity is not less than a second preset threshold, it is determined that the component is a repeated component; When the component similarity is less than the first preset threshold, the background similarity is less than the second preset threshold, and the full image similarity after the global motion compensation is not less than a third preset threshold, it is determined that the component is a repeated component.

[0070] Based on the foregoing inventive concept, as Figure 5 The present application also proposes a computer device 500, which includes a memory 510, a processor 520, and a computer program 530 stored in the memory 510 and executable on the processor 520, wherein the processor 520 implements the foregoing power transmission line inspection image component deduplication method when executing the computer program 530.

[0071] Based on the foregoing inventive concept, the present application proposes a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the foregoing power transmission line file in-patrol image component deduplication method.

[0072] Based on the foregoing inventive concept, the present application proposes a computer program product, which comprises a computer program, and the computer program is executed by a processor to realize a power transmission line file in-patrol image component deduplication method.

[0073] The acquisition, storage, use, processing, etc. of data in the technical solution of the present application comply with relevant regulations.

[0074] Compared with the image deduplication technical solution in the prior art, the present application acquires the power transmission line file in-patrol image; constructs an adjacent image pair according to the name of the power transmission line file in-patrol image; inputs the adjacent image pair into a pre-trained recognition network to output a recognition result; the recognition result comprises a component region recognition result and a background region recognition result; the recognition network is obtained by training a convolutional neural network using a global image carrying a component region mask and a component region image corresponding to the global image; the recognition network is used for feature extraction, feature decoupling and image reconstruction of the adjacent image pair; a global motion compensation method is used to perform global motion compensation on the adjacent image pair to generate a globally motion-compensated adjacent image pair; a cosine similarity algorithm is used to calculate a component similarity, a background similarity and a globally motion-compensated full image similarity according to the recognition result and the globally motion-compensated adjacent image pair; and the component in the adjacent image pair is determined to be a duplicate component through hierarchical and step-by-step judgment according to the component similarity, the background similarity, the globally motion-compensated full image similarity and a preset threshold; and the adjacent image pair is deduplicated according to the duplicate component, which can provide a deduplication technology that is accurate, efficient, lightweight and suitable for the characteristics of the in-patrol image, reduces the redundant data in the power transmission line file in-patrol image, and reduces the cost.

[0075] The method of the present application can significantly improve the accuracy and efficiency of the in-patrol image interval deduplication in the power transmission line file. Its core advantages are: The application of feature decoupling and cross-reconstruction network, which can effectively decouple the main features of the target component from the background features, thereby avoiding the interference of background information and ensuring the accurate identification of the target component. Through this decoupling technology, the network can extract clearer and more representative component features, avoiding the negative impact of background changes or irrelevant areas on the recognition result. This technology greatly improves the robustness and accuracy of target component identification, especially in complex backgrounds and dynamic scenarios.

[0076] In addition, the introduction of dynamic scene perception and compensation technology makes the differences between adjacent images due to camera movement or object displacement can be effectively compensated, thereby ensuring the accurate alignment of the target components in the images, eliminating the errors caused by motion and perspective changes. Through global motion estimation, the application of the optical flow method further improves the accuracy of dynamic compensation, avoiding the motion mismatch problem in traditional methods.

[0077] Finally, the multi-source consistency fusion decision strategy considers the subject features, context background features, and whole image similarity information, and accurately judges whether the target is the same component through hierarchical screening. The advantage of this engine is its hierarchical judgment mechanism, which can improve the accuracy and reliability of the final decision through subsequent supplementary judgment even if the similarity of a certain layer is not high enough, thereby effectively solving the misjudgment caused by complex background and different perspectives. Overall, the present application combines the depth feature decoupling and reconstruction network, dynamic scene perception and compensation technology, and multi-level feature similarity collaborative decision engine, significantly improves the accuracy and robustness of component deduplication, especially in complex and dynamic inspection environments, can effectively solve the problem of component repeated judgment caused by factors such as light, perspective, and occlusion, and can improve the identification accuracy and recall rate of target components, providing reliable support for subsequent component matching and defect detection.

[0078] This method has wide application prospects, especially in the fields of power transmission line inspection and equipment detection, and has important significance for the intelligent operation and maintenance of the power system.

[0079] Those skilled in the art will appreciate that embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0080] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the computer or other programmable data processing apparatus produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The function of one flow or multiple flows and / or blocks Figure 1 The function of one flow or multiple flows and / or blocks

[0081] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more flow or blocks and / or blocks Figure 1 one or more flow or blocks and / or blocks

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions that are executed on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flow or blocks and / or blocks Figure 1 one or more flow or blocks and / or blocks

[0083] The above specific embodiments are described to further explain the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for deduplicating components in inspection images of transmission line spans, characterized in that, include: Acquire inspection images of transmission line files; Based on the names of the inspection images in the transmission line files, adjacent image pairs are constructed; The adjacent image pairs are input into a pre-trained recognition network, which outputs recognition results. The recognition results include component region recognition results and background region recognition results. The recognition network is obtained by training a convolutional neural network using a global image labeled with component region masks and component region images corresponding to the global image. The recognition network is used to perform feature extraction, feature decoupling, and image reconstruction on the adjacent image pairs. The optical flow method is used to perform global motion compensation on the adjacent image pairs to generate globally motion-compensated adjacent image pairs. Using the cosine similarity algorithm, the component similarity, background similarity, and global image similarity after global motion compensation are calculated based on the recognition results and the adjacent image pairs after global motion compensation. Based on the component similarity, the background similarity, the global motion-compensated full-image similarity, and a preset threshold, the components in the adjacent image pairs are judged hierarchically and level by level, and duplicate components are identified in the adjacent image pairs; Based on the repeating component, the adjacent image pairs are deduplicated.

2. The method as described in claim 1, characterized in that, The recognition network includes a feature extraction module, a feature decoupling module, a feature reconstruction module, and a loss function; The feature extraction module is a backbone network built on a convolutional neural network, used to extract features from the adjacent image pairs; The feature decoupling module is used to decouple the features into component features and background features; The feature reconstruction module is used to reconstruct the image using a decoder based on the features, the component features, and the background features, to generate a reconstructed image. The loss function is the result of a weighted sum of multiple sub-loss functions; the multiple sub-loss functions are used to optimize the loss function of each part in the recognition network.

3. The method as described in claim 2, characterized in that, The feature decoupling module includes a component feature encoder and a background feature encoder; The component feature encoder includes a max pooling layer, a convolutional layer, a fully connected layer, and a feature extractor; the component feature encoder is used to decouple the features through the max pooling layer, the convolutional layer, and the fully connected layer to obtain a first component feature, and to extract a second component feature from the features through the feature extractor; The background feature encoder includes a max pooling layer, a convolutional layer, and a fully connected layer; the background feature encoder is used to decouple the features to obtain background features.

4. The method as described in claim 3, characterized in that, The feature reconstruction module is used to reconstruct the first reconstructed image by using a decoder based on the features of the first component and the background features; The feature reconstruction module is further configured to reconstruct the second reconstructed image by means of the second component features and the background features using a decoder. The decoder includes a deconvolution layer, a batch normalization function, and an activation function. The deconvolution layer is used to restore the spatial resolution of the image.

5. The method as described in claim 1, characterized in that, The component region identification result includes the component feature vectors of the adjacent image pairs; the background region identification result includes the context background feature vector of each component in the adjacent image pairs; The context background feature vector is the feature vector of the adjacent areas located above, below, to the left and to the right of the component; Using a cosine similarity algorithm, based on the recognition results and the adjacent image pairs after global motion compensation, the component similarity, background similarity, and global image similarity after global motion compensation are calculated, including: The cosine similarity algorithm is used to calculate the cosine similarity between the component feature vectors of the two images in the adjacent image pair, and the cosine similarity between the component feature vectors of the two images in the adjacent image pair is determined as the component similarity. Calculate the cosine similarity between the context background feature vector of one image and the context background feature vector of the other image in the adjacent image pair, and determine the maximum value of the calculation result as the background similarity. Calculate the cosine similarity between two images in a pair of adjacent images after global motion compensation, and determine the cosine similarity between two images in a pair of adjacent images after global motion compensation as the global image similarity after global motion compensation.

6. The method as described in claim 1, characterized in that, Based on the component similarity, the background similarity, the global motion-compensated full-image similarity, and a preset threshold, components in adjacent image pairs are judged hierarchically and level by level. Duplicate components are identified in the adjacent image pairs, including: If the similarity of the components is not less than a first preset threshold, the components are determined to be duplicate components. If the similarity of the component is less than a first preset threshold and the similarity of the background is not less than a second preset threshold, the component is determined to be a duplicate component. If the similarity of the component is less than a first preset threshold, the similarity of the background is less than a second preset threshold, and the similarity of the whole image after global motion compensation is not less than a third preset threshold, the component is determined to be a duplicate component.

7. A device for removing duplicate components from inspection images of transmission line spans, characterized in that, include: The image acquisition module is used to acquire inspection images of transmission line sections; The adjacent image pair construction module is used to construct adjacent image pairs based on the names of the inspection images in the transmission line file; The region recognition module is used to input the adjacent image pairs into a pre-trained recognition network and output recognition results; the recognition results include component region recognition results and background region recognition results; the recognition network is obtained by training a convolutional neural network using a global image with component region mask annotations and component region images corresponding to the global image; the recognition network is used to perform feature extraction, feature decoupling, and image reconstruction on the adjacent image pairs; The global motion compensation module is used to perform global motion compensation on the adjacent image pairs using optical flow, and generate globally motion-compensated adjacent image pairs. The similarity calculation module is used to calculate the component similarity, background similarity, and global image similarity after global motion compensation by using the cosine similarity algorithm based on the recognition result and the adjacent image pairs after global motion compensation. The duplicate component determination module is used to perform hierarchical and step-by-step judgment on components in the adjacent image pairs based on the component similarity, the background similarity, the global motion-compensated full-image similarity, and a preset threshold, and to determine duplicate components in the adjacent image pairs. The deduplication module is used to deduplicat adjacent image pairs based on the repeating components.

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, it implements the method of any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.

Images