Localization of Generator Inspection Vehicle using Visual Odometry
| Autor: | Toshiaki Kondo, H. Shoichi, Waree Kongprawechnon, Teesid Leelasawassuk, T. Pattanapisont |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
0209 industrial biotechnology
Generator (computer programming) Rotor (electric) Orientation (computer vision) business.industry Computer science 020208 electrical & electronic engineering Optical flow 02 engineering and technology Translation (geometry) law.invention 020901 industrial engineering & automation law 0202 electrical engineering electronic engineering information engineering Computer vision Artificial intelligence Visual odometry business Rotation (mathematics) Slip (vehicle dynamics) |
| Zdroj: | SICE |
| DOI: | 10.23919/sice.2019.8859867 |
| Popis: | This research presents a method to determine orientation and translation of Generator Inspection Vehicle (GIV). The GIV is developed to be an inspecting assistant for generator inspection staff to examine the internal condition of a power generator. It is inserted into a space between rotor and stator of a power generator, and used for examining the condition inside the power generator. The main purpose of the GIV is to reduce cost and time spent for rotor removal and insertion. The GIV version one was attached with encoders to the vehicle wheels for measuring distance, and its accuracy was only 72%. The accuracy is not as good as it should be due to slip between the GIV's wheels and the surface inside the generator. The slip caused the encoders to miscount rotation of GIV's wheels. Therefore, a computer vision technique, namely Visual Odometry, is applied to alleviate the issue. Visual Odometry (VO) is a process to determine position and orientation of a robot in motion. The mentioned method is applied for localizing of the GIV's position inside a power generator. This study investigates two VO methods. The first method determines the orientation and translation using optical flow extracted from FAST feature detection. The second method uses an adaptation of PTAM method [8] to extract the GIV motion. The GIV's extracted translation of both methods are plotted to see the trend of motion. The results are evaluated by comparing with ground-truth acquired from a motion tracker system. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|