Visual Perception based Pipeline Tracking System using a Multirotor

• Main Authors: HUANG, PIN-HAN, 黃品翰
• Other Authors: SHIEH,MING-YUAN
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/r522pe

碩士 === 南臺科技大學 === 電機工程系 === 106 === The pipeline system is still the main transportation method for transporting liquids, gases, and others such as petroleum and natural gas. It is widely used in many buildings and factories. In order to maintain the safety of use, regular inspections and maintenance are indispensable. Compared with in-pipe inspection, out-pipe inspection is having the advantages of easy installation and non-destructive. However, since manual inspections will consume many workers and time costs, how to improve the efficiency of pipe maintenance and reducing the burden on operators has become an important issue in the development of automation systems in recent years. To adopt out-pipe inspection robots is a general solution especially for thickness and corrosion inspection. Since most pipelines are fixed on the walls of buildings and factories, such working situation is not suitable the out-piping mobile robot. Therefore, the out-pipe inspection with an automatic pipeline tracking system by a multirotor can perform many tasks in place of workers, more important is with features such as high security, good economy, and strong functionality. In the process of pipeline inspection, the most critical is to achieve the automatic tracking of pipelines. Based on a monocular webcam, the detection and tracking system for an Arduino multirotor that is proposed to realize autonomous cruise of pipelines. Firstly, according to the imaging characteristics of pipe images, every image of pipeline is processing by image binarization and canny edge detection algorithm to collect the boundary information of object. Second, shape characteristics of pipeline are obtained by Hough transform; besides, the tracking of pipelines are carried out according to the feature of pipeline by Kalman filter. Finally, some experiments are executed under different scenes whose results illustrate the effectiveness of the proposed pipeline tracking system. It demonstrates that the proposed pipeline tracking algorithm has good robustness and real-time performance.