Detection Of Foreign Objects On Tower And Transmission Line Galloping Based On Image Sequence And Video Analysis

Electricity is the basis for maintaining the normal operation of modern society.Ensuring the normal supply of electricity is of vital importance to the entire society.However,transmission lines erected in the wild are extremely vulnerable to various safety hazards during operation.Among them,tower foreign objects and transmission line galloping are relatively common major hazards that seriously endanger the stable operation of the power transmission system.Foreign objects appearing on the tower or near the tower can easily be blown onto the transmission line by the wind and cause electrical accidents such as line short circuit.Once the transmission line galloping for a long time,it is very easy to cause the interphase discharge,resulting in a large-scale blackout accident.In order to discover these hidden dangers in time,it is an effective inspection way to monitor the transmission system in real time by installing cameras on the tower.Therefore,the main research content is the detection of tower foreign objects and transmission line galloping.Aiming at the detection of tower foreign objects,the common method is to detect tower foreign objects by extracting the features of foreign objects.However,the number of pixels presented by tower foreign objects in the image is very small,and the types,shapes and positions of tower foreign objects are very easy to change,it is difficult to extract effective features by single image.Considering that the color and texture of the location of the foreign objects will change before and after it appears,we proposed a foreign objects detection method based on deep background difference model.The main work is as follows:(1)A deep background difference model based on Deep Labv3+ is proposed.The image difference method which only uses pixel-level information has poor effect in image sequence.Therefore,we proposed a deep background differencing model by combining deep learning with image difference,which can learn deep difference features and semantic information between different images and has better robustness to disturbances such as illumination changes and dynamic backgrounds.(2)A method of tower foreign objects detection based on deep background difference model is proposed.First,the combination of YOLOv4 and image sequence is used to detect the tower region of the image,so as to narrow the detection range of the tower foreign objects and improve the detection speed and accuracy;then,the tower area in the background image is aligned with the tower region in the current image to reduce the impact of camera shake on the differential effect;finally,the deep background differential model is called to achieve the tower foreign objects detection.Focusing on the transmission line galloping detection,in high voltage or ultra-high voltage transmission lines,spacers are installed on the wires to ensure that a certain distance is maintained between different wires to prevent interphase discharge.Therefore,a method of transmission line galloping detection based on significant target point tracking is proposed,which takes the spacer as the significant target point to reflect the transmission line galloping.The main work is as follows:(1)Video pre-processing.The camera installed on tower due to the impact of high winds and other factors will inevitably shake resulting in unstable captured video images,the error in the jittery video for transmission line galloping is large.So,a video de-jittering method based on local feature point matching is used to get a stable video.(2)A transmission line galloping detection method based on significant target point tracking is proposed.First,the method of rotational projection is used to locate the transmission line;then,the spacer is detected by traversing the line and analyzing the Laws energy of the projection curve in the sliding window;finally,by tracking the spacer to obtain its trajectory and and calculating the amplitude,frequency,and other parameters to achieve the transmission line galloping detection.Previously,our research group cooperated with a company in Shandong to develop a transmission line channel hidden danger detection system.At present,some of the algorithms in this paper have been integrated into the system and applied in several provinces and cities,which greatly reduces the workload of power staff and improves inspection efficiency.