Power transmission line external damage hidden danger identification method based on vanishing point driving area

Abstract

The invention discloses a power transmission line external damage hidden danger identification method based on a vanishing point driving area. The method comprises the following steps: S1, constructing a YOLOv8 network model based on a deep learning convolutional network; s2, acquiring a target image of external damage of the power transmission line, preprocessing the target image to obtain the preprocessed target image, constructing a data set according to the preprocessed target image, and dividing the data set into a training set and a verification set; s3, the YOLOv8 model is trained through the training set, and the trained YOLOv8 model is obtained; s4, optimizing the target images in the verification set to obtain a plurality of optimized target images; s5, slicing the plurality of optimized target images to obtain a plurality of slice images, and performing target detection on the plurality of slice images based on the trained YOLOv8 model to obtain a corresponding low-missing-report-rate target detection result; and S6, carrying out validity identification detection on the low-missing-report-rate target detection result, and sending an alarm signal according to the detection result.

Description

Technical Field

[0001] The present application relates to the technical field of online monitoring of high-voltage transmission lines, and in particular to a method for identifying external damage hazards of transmission lines based on a vanishing point driving area. Background Art

[0002] In recent years, China's power industry has developed rapidly, and transmission lines have assumed the main responsibility of transmitting large amounts of electricity from power plants to distribution centers or urban substations. Due to the expansion of transmission lines, the transmission system has become more complex. Once a power outage occurs, it will cause serious consequences such as reduced industrial production, traffic congestion, and medical failure.

[0003] According to statistics, external damage has become the biggest risk to the safe operation of transmission lines, except for lightning damage. Illegal operations, illegal construction, foreign object intrusion, sawing and demolition are greater threats to the safe and stable operation of transmission lines. Therefore, the environment in which the line is located is crucial to the safe operation of the transmission line. Ensuring the safe operation of the line and reducing non-natural factors such as external damage through artificial intelligence methods are of great significance to improving the stability of the power system.

[0004] With the development of science and technology, remote monitoring equipment has been widely used in power transmission line inspections. Through the camera installed on the tower, operation and maintenance personnel can easily observe the status of the line channel. In recent years, computer vision technology based on deep learning has made significant progress. By building an external damage target recognition model for the transmission channel, it can replace maintenance personnel to monitor the status of the transmission channel in real time, greatly improving the efficiency and effectiveness of online monitoring of external damage hazards.

[0005] At present, the target recognition model for external damage hazards in power transmission channels is mainly based on the external damage monitoring algorithm that automatically extracts features through deep learning, and no longer relies on manually designed feature extraction, which significantly reduces labor costs and reduces error rates. However, the scales of targets of the same type vary greatly, and the detection effect is particularly poor for distant objects or large objects in the detection image. In fact, some instruments may be far away from the power lines and are only recognized by the algorithm in the image, but will not cause hidden dangers or substantial damage to the lines.

[0006] To address these problems, the present invention proposes a method for identifying hidden dangers of external damage to transmission lines based on vanishing point driven areas, which enhances the model's ability to identify external damage to transmission lines, accurately divides effective warning areas, and accurately predicts whether there is a risk of external damage. Summary of the invention

[0007] The embodiment of the present application provides a method for identifying external damage hazards of power transmission lines based on a vanishing point driving area, so as to solve the technical problems of high false alarm rate, low alarm accuracy and poor effect in the existing target identification and detection of external damage hazards of power transmission channels.

[0008] In view of this, the present application provides a method for identifying hidden dangers of external damage of power transmission lines based on a vanishing point driving area, comprising the following steps:

[0009] S1. Build a YOLOv8 network model based on deep learning convolutional network;

[0010] S2, collecting target images of external damage to the transmission line, and preprocessing the target images to obtain preprocessed target images, constructing a data set according to the preprocessed target images, and dividing the data set into a training set and a verification set;

[0011] S3. Training the YOLOv8 model using the training set to obtain a trained YOLOv8 model;

[0012] S4, optimizing the target image in the verification set to obtain multiple optimized target images;

[0013] S5, slicing the multiple optimized target images to obtain multiple slice images, performing target detection on the multiple slice images based on the trained YOLOv8 model to obtain corresponding low missed rate target detection results;

[0014] S6. Perform validity identification detection on the low missed alarm rate target detection result, and issue an alarm signal according to the detection result.

[0015] Optionally, the method steps for preprocessing the image data in step S2 are:

[0016] Firstly, the image data is subjected to denoising processing to obtain a target image after denoising processing;

[0017] Then, the target image after the denoising process is subjected to image enhancement process to obtain the pre-processed target image.

[0018] Optionally, the method steps of slicing the target image in the verification set in step S4 are:

[0019] Performing segmentation processing on the preprocessed target image to obtain a plurality of segmented images;

[0020] Performing preliminary image enhancement processing on the multiple segmented images to obtain multiple preliminary enhanced segmented images;

[0021] Linearly superimposing the multiple preliminary enhanced segmentation images to obtain multiple linearly superimposed preliminary enhanced segmentation images;

[0022] Based on the mosaic principle and Cutout method, multiple preliminary enhanced segmentation images after linear superposition are spliced ​​and optimized to obtain multiple optimized target images.

[0023] Optionally, in step S5, the plurality of optimized target images are sliced ​​to obtain a plurality of sliced ​​images by:

[0024] Based on the sliding window, multiple optimized target images are segmented into regions to obtain multiple slice images.

[0025] Optionally, the method steps for obtaining the corresponding low missed alarm rate target detection result in step S5 are:

[0026] Based on the non-maximum suppression method, the target detection results of all slice images are processed by removing overlapping detection frames, and several target detection frames are obtained;

[0027] Several target detection frames are spliced ​​together to generate an overall low false negative rate target detection result.

[0028] Optionally, the method steps of performing validity identification detection on the low missed alarm rate target detection result in step S6 and issuing an alarm signal according to the detection result are:

[0029] Preprocessing, Canny edge detection and Hough transform are performed on the target image of the external damage of the transmission line to obtain the pixel coordinates of the vanishing point;

[0030] Performing image screening and region sorting on the target image of the external damage of the power transmission line according to the pixel coordinates of the vanishing point to obtain a target effective region;

[0031] The target position in the low missed alarm rate target detection result is compared with the effective area by pixels to obtain a comparison result, and an alarm is issued according to the comparison result.

[0032] Optionally, the structure of the YOLOv8 network model includes a backbone network, a neck network and a detection head, the backbone network adopts a cross-stage local network, the cross-stage local network includes several layers of convolutional layers and a feature pyramid network, the neck network includes a feature pyramid network, a path aggregation network and an upsampling layer, and the detection head includes a small target detection layer.

[0033] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0034] The present application provides a method for identifying hidden dangers of external damage of power transmission lines based on vanishing point driving areas. The method constructs a YOLOv8 network model based on a deep learning convolutional network; collects target images of external damage to the transmission line, and preprocesses the target images to obtain preprocessed target images, constructs a data set based on the preprocessed target images, and divides the data set into a training set and a verification set; trains the YOLOv8 model through the training set to obtain a trained YOLOv8 model; optimizes the target image in the verification set to obtain a plurality of optimized target images; slices the plurality of optimized target images to obtain a plurality of slice images, and performs target detection on the plurality of slice images based on the trained YOLOv8 model to obtain corresponding target detection results with a low false negative rate, thereby improving the credibility of identifying hidden dangers of external damage to the transmission line and reducing the false negative rate;

[0035] Then, by performing validity identification detection on the low missed alarm rate target detection results and sending out an alarm signal according to the detection results, the number of invalid alarms is reduced and the level of online monitoring of external damage hazards is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly express the technical solutions of the embodiments of the present application, the drawings required for describing the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0037] Figure 1 A step diagram of a method for identifying external damage hazards of a power transmission line based on a vanishing point driving area provided in an embodiment of the present application. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0039] For easier understanding, see Figure 1 The present invention provides a method for identifying hidden dangers of external damage of power transmission lines based on a vanishing point driving area, comprising the following steps:

[0040] The method comprises the following steps: S1, constructing a YOLOv8 network model based on a deep learning convolutional network; specifically, constructing a YOLOv8 network model suitable for identifying hidden dangers of external damage to power transmission channels, and optimizing the backbone structure of the network by introducing the efficient feature extraction mechanism of YOLOv8 to improve the ability to identify hidden dangers of external damage to power transmission channels under complex background conditions.

[0041] S2. Collect the target image of the external damage of the transmission line, and pre-process the target image to obtain the pre-processed target image. Specifically, perform denoising on the target image, perform preliminary denoising on the target image using the two-dimensional discrete cosine transform domain, obtain the target image after preliminary denoising, obtain the energy parameter of the target image after preliminary denoising, obtain the residual of the central pixel of each unit area of ​​the target image after preliminary denoising based on the energy parameter, calculate the weight matrix of each unit area based on the residual, further denoise the target image after preliminary denoising based on the weight matrix, and obtain the denoised target image. Perform image enhancement on the denoised target image, perform frequency domain modulation on the object field of the denoised target image, perform Fourier transform on the frequency domain, take the imaginary part of the frequency domain after Fourier transform, calculate the absolute value of the imaginary part, calculate the power value based on the absolute value of the imaginary part after summing along the frequency, perform image enhancement based on the power value, and complete the image pre-processing after image enhancement to obtain the pre-processed target image, construct a data set based on the pre-processed target image, and divide the data set into a training set and a validation set.

[0042] S3. Train the YOLOv8 model using the training set to obtain a trained YOLOv8 model.

[0043] S4. Optimize the target image in the validation set to obtain multiple optimized target images; specifically, segment the target image in the validation set after preprocessing, use a neural network to assist the segmentation of the target image in the validation set after preprocessing, obtain multiple segmented images, and then classify and screen the multiple segmented images, and manually correct the segmentation errors, and select objects with obvious features and completeness to form a database. Perform preliminary image enhancement processing on multiple segmented images, and perform image change processing such as random cropping, rotation, and flipping on multiple segmented images to obtain multiple preliminary enhanced segmented images. Linearly superimpose multiple preliminary enhanced segmented images, that is, linearly superimpose every two preliminary enhanced segmented images at a random ratio, and adjust the labels accordingly to obtain multiple preliminary enhanced segmented images after linear superposition. Then, based on the mosaic principle and the Cutout method, the multiple preliminary enhanced segmented images after linear superposition are spliced ​​and optimized. The mosaic principle is an image processing technology that is often used to hide or protect sensitive information. Its principle is to divide the specific area of ​​the image into small blocks and fill each small block with the same color or blur effect, divide each preliminary enhanced segmentation image after linear superposition into multiple small blocks, and the divided small blocks are of the same size. Based on the blur filter, the recognition of each small block is removed to make it lose its recognition, and the small blocks after the recognition removal are recombined into a mosaic image, so as to achieve the purpose of hiding or confusing the target area. Cutout is a data enhancement technology, which is achieved by masking random parts of the input image, randomly covering parts of the image, simulating occlusion to enhance the situation where the target of the hidden danger of external damage to the transmission line is occluded, and enhancing the model's recognition ability when the target of the hidden danger of external damage to the transmission line is partially occluded, and obtaining multiple optimized target images.

[0044] S5. Slice multiple optimized target images to obtain multiple slice images, perform target detection on the multiple slice images based on the trained YOLOv8 model, and obtain corresponding low missed rate target detection results; specifically, slice multiple optimized target images to obtain multiple slice images, including: performing regional segmentation on multiple optimized target images based on a sliding window to obtain multiple slice images. Specifically, set the window size and step size of the sliding window, and then set the sliding window, perform regional segmentation on multiple optimized target images based on the sliding window, and obtain multiple slice images. Image slicing technology ensures that each slice contains enough detail information to avoid small targets being ignored in large-resolution images. At the same time, the slicing super-inference method decomposes the large-resolution image into several small blocks, reduces the amount of calculation for a single inference, and improves the inference efficiency. Based on the non-maximum suppression method, the target detection results of all slice images are subjected to overlapping detection frame removal processing, and several target detection frames are obtained; several target detection frames are spliced ​​to generate an overall low missed rate target detection result.

[0045] S6. Perform validity identification detection on the low missed rate target detection results, and issue an alarm signal based on the detection results; specifically, perform preprocessing, Canny edge detection and Hough transform on the target image of the external damage of the transmission line to obtain the vanishing point pixel coordinates; perform image screening and regional sorting on the target image of the external damage of the transmission line according to the vanishing point pixel coordinates to obtain the target effective area; perform pixel comparison between the target position in the low missed rate target detection result and the effective area to obtain the comparison result, and issue an alarm based on the comparison result; specifically, based on the image vanishing point calculation method of Canny edge detection and Hough line detection, according to the mapping relationship between the vanishing point and the actual width and height of the transmission line, the effective warning area of ​​the transmission channel is dynamically generated. When the camera angle or height changes, the system adjusts the warning area boundary by recalculating the position of the vanishing point, thereby achieving adaptive coverage of the hidden danger area to ensure the accuracy and reliability of the image vanishing point selection.

[0046] The Canny edge detection algorithm converts the scene image of the transmission corridor into a binary image, that is, the pixel values ​​are only 0 and 255, and only the contour features of the scene elements are retained. This narrows the feature search range of the Hough line detection and eliminates the interference of background elements on the image vanishing point. The Hough line detection algorithm maps the lines in the image to the parameter space based on the point-curve duality principle, and its polar coordinate expression is:

[0047] λ=xcosθ+ysinθ

[0048] Where λ is the vertical distance from the origin to the straight line, and θ is the angle between the vertical line and the x-axis; according to the same straight line of λ, θ is a fixed value, and when images with different coordinates (x, y) are input into the λ-θ coordinate system, the corresponding number of curves is generated, where the intersection of the curves corresponds to the number of points passing through the straight line. Based on this principle, the detection of image straight line parameters is realized. In practical applications, images of scenes with hidden dangers of external damage captured may not perfectly intersect at a certain point despite the presence of distortion; usually, multiple parallel lines in 3D will intersect at several points in a 2D image. To this end, it is necessary to calculate and determine the optimal point from these intersection points as the vanishing point of the image, which is the requirement that best reflects the degree of image distortion;

[0049] Specifically, the calculation of the vanishing point is:

[0050] First, the intersection of two lines L1 and L2 that are parallel in reality but intersecting in the image is calculated and used as a temporary vanishing point (x VP ,y VP ), as shown in the formula:

[0051]

[0052] Then, through this point (xVP ,y VP ) generates a vertical line L perpendicular to the third line L3 ⊥ The parameters of the vertical line equation are shown in the formula:

[0053]

[0054] Then, L3 and L are calculated ⊥ The intersection point is (x3,y3). The calculation method is as shown in the formula:

[0055]

[0056] Finally, the distance between the temporary vanishing point and the third line L3 is calculated, that is, the distance between the two points (x VP ,y VP ) and (x3,y3). The calculation method is as shown in the formula:

[0057]

[0058] where k1, k2, k3 and k ⊥ It is the straight line L1,L2,L3,L ⊥ The slope of b1, b2, b3 and b ⊥ are the straight lines L1, L2, L3 and L ⊥ The intercept of ; the distance is the length from the temporary vanishing point to L3;

[0059] Calculate the sum of the squares of the distances from the temporary vanishing point to more straight lines, and determine the square root as the calculation error, and compare it with the current minimum error value. If the new calculation error is smaller, redetermine the vanishing point coordinates, and finally select the point with the smallest total square distance to all lines as the vanishing point.

[0060] Furthermore, the structure of the YOLOv8 network model includes a backbone network, a neck network and a detection head, wherein the backbone network adopts a cross-stage local network, the cross-stage local network includes several layers of convolutional layers and a feature pyramid network, the neck network includes a feature pyramid network, a path aggregation network and an upsampling layer, and the detection head includes a small target detection layer; specifically, the YOLO neural network is an object detection system for computer vision tasks, and the YOLOv8 model is an object detection model based on the YOLO neural network. There are many small targets at long distances in the data set, and the model architecture of YOLOv8 is first improved. The small targets in the data set are analyzed and it is found that these small targets are caused by the long distance of the lens. Such targets are relatively complex, usually the side or front of the ship, and there are also some long-distance buoys, so it is difficult to identify such targets. In view of this situation, a detection layer is added on the basis of YOLOv8 to improve the feature extraction capability of small targets. Since the sample size of small targets is small and the sampling multiple of YOLOv8 is relatively large, it is difficult for the model to learn the feature information of small targets for deeper target feature maps. Based on the original network model of YOLOv8, the feature extraction network is modified to achieve small target detection by adding upsampling layers and convolution layers. In addition, parameters such as the size of the convolution kernel are modified according to the target features to better recognize the distant sailboats in the data set. The YOLOv8 model uses a cross-stage local network as the backbone network, which consists of a series of convolution layers and a feature pyramid network. The output feature map of the backbone network will be passed to the subsequent processing layer. The network structure design of the neck network applies the structure of the feature pyramid network plus the path aggregation network. The feature pyramid network layer conveys strong semantic features from top to bottom, while the path aggregation network tower conveys positioning features from bottom to top. The detection head is responsible for predicting the bounding box, category and confidence of the target. Detection heads of different scales are responsible for target detection prediction at different feature levels. Compared with the original model, the input resolution of the backbone network is increased to 1056, the neck network part adds an upsampling layer, sets a new convolution kernel size for small targets, and the detection head part is changed from three layers to four layers, expanding the small target detection layer. In order to improve the generalization ability of the model and enhance the detection performance of the small target detection layer, the modified model is pre-trained on the Common Objects in Context (COCO) dataset, and the complexity of the model is reduced by model pruning. The pre-trained model is transferred to the dataset for training to obtain a trained improved YOLOv8 model. Multiple slice images are input into the trained improved YOLOv8 model to obtain the target detection result corresponding to each slice image.

[0061] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0062] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for identifying hidden dangers of external damage of power transmission lines based on vanishing point driving area, characterized in that: The following steps are involved: S1. Build a YOLOv8 network model based on deep learning convolutional network; S2, collecting target images of external damage to the transmission line, and preprocessing the target images to obtain preprocessed target images, constructing a data set according to the preprocessed target images, and dividing the data set into a training set and a verification set; S3. Training the YOLOv8 model using the training set to obtain a trained YOLOv8 model; S4, optimizing the target image in the verification set to obtain multiple optimized target images; S5, slicing the multiple optimized target images to obtain multiple slice images, performing target detection on the multiple slice images based on the trained YOLOv8 model to obtain corresponding low missed rate target detection results; S6. Perform validity identification detection on the low missed alarm rate target detection result, and issue an alarm signal according to the detection result.

2. According to claim 1, a method for identifying hidden dangers of external damage of power transmission lines based on vanishing point driving area is characterized in that: The method steps for preprocessing the image data in step S2 are: Firstly, the image data is subjected to denoising processing to obtain a target image after denoising processing; Then, the target image after the denoising process is subjected to image enhancement process to obtain the pre-processed target image.

3. The method for identifying hidden dangers of power transmission line damage based on vanishing point driving area according to claim 1, characterized in that: The method steps for slicing the target image in the verification set in step S4 are as follows: Performing segmentation processing on the preprocessed target image to obtain a plurality of segmented images; Performing preliminary image enhancement processing on the multiple segmented images to obtain multiple preliminary enhanced segmented images; Linearly superimposing the multiple preliminary enhanced segmentation images to obtain multiple linearly superimposed preliminary enhanced segmentation images; Based on the mosaic principle and Cutout method, multiple preliminary enhanced segmentation images after linear superposition are spliced ​​and optimized to obtain multiple optimized target images.

4. The method for identifying hidden dangers of power transmission line damage based on vanishing point driving area according to claim 1, characterized in that: In step S5, the plurality of optimized target images are sliced ​​to obtain a plurality of sliced ​​images in the following steps: Based on the sliding window, multiple optimized target images are segmented into regions to obtain multiple slice images.

5. The method for identifying hidden dangers of power transmission line damage based on vanishing point driving area according to claim 1, characterized in that: The method steps for obtaining the corresponding low missed alarm rate target detection result in step S5 are: Based on the non-maximum suppression method, the target detection results of all slice images are processed by removing overlapping detection frames, and several target detection frames are obtained; Several target detection frames are spliced ​​together to generate an overall low false negative rate target detection result.

6. The method for identifying hidden dangers of power transmission line damage based on vanishing point driving area according to claim 1, characterized in that: In step S6, the method steps of performing validity identification detection on the low missed alarm rate target detection result and issuing an alarm signal according to the detection result are as follows: Preprocessing, Canny edge detection and Hough transform are performed on the target image of the external damage of the transmission line to obtain the pixel coordinates of the vanishing point; Performing image screening and region sorting on the target image of the external damage of the power transmission line according to the pixel coordinates of the vanishing point to obtain a target effective region; The target position in the low missed alarm rate target detection result is compared with the effective area by pixels to obtain a comparison result, and an alarm is issued according to the comparison result.

7. The method for identifying hidden dangers of power transmission line damage based on vanishing point driving area according to claim 1, characterized in that: The structure of the YOLOv8 network model includes a backbone network, a neck network and a detection head. The backbone network adopts a cross-stage local network. The cross-stage local network includes several convolutional layers and a feature pyramid network. The neck network includes a feature pyramid network, a path aggregation network and an upsampling layer. The detection head includes a small target detection layer.

Images