A Real-time Extraction And Tracking For Moving Trolleybus Power Line Based On The Embedded System

Trolleybus is a kind of public transport means with energy conservation environmental protection,comfort,health,low price and agility.Trolley buses should move along the power lines' track strictly.Therefore,the trolleybus' s poles will fall off from the power lines when trolleybuses overtake or change the lane.At present the main methods to solve the problem depend on manual connection.As for using manual connection,it will cause the traffic jams.To overcome problems mentioned above,the paper combines embedded system with the target tracking algorithm and proposes a real-time automatic connection platform based on embedded system and a real-time extraction and tracking algorithm for the power line positions.1.Design and implement the automatic connection platform based on the embedded system platform for the real-time automatic connection,including hardware module,the selection of sensors and cameras,the construction of the Linux system environment and the generation of boot files and loading the driver module,etc.2.Put forward a complete set of extraction and tracking algorithm for the automatic connection platform.This algorithm includes three stages.Image preprocessing,the paper presents a novel method to preprocess the original images based on the Otsu.Line extraction,the Probabilistic Hough Transform(PHT)is applied to obtain the line set.Clustering,in the Hough space,system clustering algorithm is used to cluster and filter and get the line set.3.Design experiments for the accuracy of the filter template,the reliability and instantaneity of the power lines tracking algorithm,the feasibility and safety of the automatic connection platform.The results show that the filter template keep the realtime performance,at the same time improve the detection precision.The accuracy of the power lines tracking algorithm can reach 95%,the average run time is 18.6ms per frame,which meet the real-time requirements.At the same time,the embedded platform experiments show that after the algorithm is transplanted to the hardware platform,operating time is 100 ms per frame,the detection accuracy is 95%,which confirmed the safety and feasibility.