ToFA: Tower frame abstraction from transmission line inspection visible video

摘要

Transmission Line Inspection by helicopter inspection obtains a large amount of video data, industrial applications are in urgent need of locating target tower rapidly within these video. However, using traditional object detection algorithm to extract video frame including target tower makes complete tower extraction difficulty: In ? background and face with changeable natural factors, the abstraction speed and accuracy can't meet the business needs. In this paper, a tower frame abstraction method based on transmission line inspection visible video is proposed to locate target tower rapidly. First, according to the known tower running number, obtain the corresponding latitude and longitude data of the target tower in the device and ledgers systems. Then match latitude and longitude with the synchronous telemetry data, get video segment containing the target tower. On this basis, based on the video segment, changing the color space and selecting feature channel to extract connected components, use a linear feature to locate the tower rapidly. Actual transmission line videos are used to test algorithm performance, test data includes far-focus video data, near-focus and far-focus blended video data and data with complex farm background. Detection accuracy rate is 88.2% in average, it verifies the effectiveness of the algorithm.