Research On Intrusion Detection Method For Large Machinery Of Transmission Line Based On Image

Transmission line is an important part of the power system,to ensure the stable operation of the power system,it is very important to prevent the transmission line from being damaged by external forces.Most of the line trip events caused by external force are caused by large-scale mechanical construction line collision,so,it is urgent to monitor the large machinery under the transmission line.At present,there are many detection methods for large machinery under transmission line,which are generally divided into two categories: the method based on traditional image processing and the method based on deep learning.Compared with traditional methods,deep learning method has the advantages of strong generalization ability and high detection accuracy.However,it is difficult to deploy on a large scale due to its large number of parameters and high computational requirements.Therefore,this paper studies the lightweight of large-scale mechanical target detection algorithm based on deep learning,which provides algorithm basis for its more efficient application in large-scale mechanical intrusion detection of transmission line.This paper first introduces the research status of transmission line intrusion detection algorithms at home and abroad in recent years.The object detection algorithm based on deep learning is selected to detect the large-scale machinery under the transmission line.In view of the specific application scenarios of this paper,the author constructs the self-made construction of three kinds of large machinery which are prone to trip line collision,such as tower crane,crane,excavator/bulldozer.The validity and rationality of the data set are verified by analyzing the data set in many aspects.Then,the object detection algorithm YOLOv3 based on deep learning is selected as the basic algorithm to complete the object detection of large machinery under the transmission line.In view of its disadvantages such as large amount of parameters,high requirements for hardware,and difficulty in large-scale deployment,this paper proposes YOLOv3-lite,which makes lightweight improvement from three aspects of feature extraction network,preset template box,and detection branch network,greatly reducing the size of the algorithm.In order to improve the detection accuracy of lightweight algorithm,this paper proposes YOLOv3-lite-SE.By adding attention module in the algorithm,the detection accuracy of the algorithm is significantly improved without adding algorithm parameters,making it more suitable for large-scale deployment in practical application scenarios.Finally,combined with the actual project,the lightweight algorithm is deployed on the embedded computing platform NVIDIA Jetson TX2,which improves the timeliness of large machinery detection under the transmission line by means of edge computing.In the case of limited computing resources,the detection of large machinery under the transmission line is completed.