Research On Detection Methods Of Foreign Objects In Power Transmission Lines

Transmission lines are the main carrier of power transmission and an important part of the power system.They play a vital role in the safe and stable operation of the system.Due to the wide distribution of power transmission lines,the complex natural environment,and the influence of people’s production and life,foreign objects such as balloons,kites,plastics,bird nests,etc.are easily attached to the lines.If these foreign objects cannot be found and cleaned up in time,they will affect the normality of the transmission line.Work,and even cause power outages or safety accidents.Transmission line image inspection can effectively detect foreign objects,eliminate hidden dangers in time,avoid foreign objects from harming the normal operation of transmission lines,and better overcome traditional manual inspection methods such as high cost,easy missed inspection,and low efficiency.Therefore,this subject has carried out a more in-depth study of foreign body detection methods on transmission lines based on classic image processing methods and deep learning methods.The main research contents are as follows:In view of the noise problem generated in the image acquisition process,the characteristics of common Gaussian noise and salt and pepper noise are analyzed.Image denoising algorithms such as median filtering,Gaussian filtering,guided image filtering and bilateral filtering are used to denoise the image.In order to objectively evaluate the pros and cons of denoising algorithms,the methods of objective evaluation of the image by means of mean square error,average absolute error,peak signal-to-noise ratio and structural similarity are used to evaluate the quality of denoised images.Finally,Gaussian filtering and bilateral bilateral filtering with better denoising effects are selected.The filtering method denoises the image and lays the foundation for the later foreign body detection.The image processing method is simple to implement,convenient and fast.Considering engineering applicability,a improve genetic algorithm Otsu method for detecting foreign objects in transmission lines is proposed based on image processing methods.Aiming at the problem that traditional threshold segmentation methods cannot completely segment the target and background,an improved genetic algorithm combined with Otsu threshold segmentation algorithm is proposed to reduce the weight coefficient of the background class,and the effectiveness of the improved method is verified by an objective evaluation method.In the aspect of power transmission line extraction,Hough line detection with added boundary conditions is adopted,and the straight lines of non-transmission lines are further screened out by analyzing the characteristics of the extracted lines based on the angle,length and coordinate points of the power lines.After that,the image is divided again by the method of region growth,and the coordinate position of the foreign object is confirmed by the method of comparing the average value of the pixels.Experiments have proved that this method can effectively detect foreign objects in transmission lines.The deep learning target detection method is used to realize the detection of foreign objects in the transmission line under complex background.Aiming at the problem of low detection accuracy of yolv3,this paper proposes YOLOv3-Rep VGG algorithm to increase the detection accuracy of foreign matters in transmission lines.According to modify the backbone network,increase the number of Rep VGG and multi-scale target detection frame.Experiments show that the improved network has a 5.8% increase in the accuracy of kite detection AP,a 10.6% increase in the detection of a bird’s nest,and an increase of 12.9% in the detection accuracy of balloons.The detection performance of small targets is significantly improved.After the improvement,the network recall rate and accuracy increased by 1.2% and 19.5% respectively,and the m AP reached 70.5%,which was an increase of 9.8% compared to the 60.7% of the YOLOv3 network m AP.Compared with other algorithms for target detection,the proposed improved method has higher detection accuracy and more accurate recognition results.