Calculating Camera Orientation using Optical flow for ADAS Applications
Hyunjun Kim1, Hwanyong Lee2
1Hyunjun Kim Department of Software and Computer Engineering, Ajou University, Suwon, Rep. of Korea, East Asian.
2Hwanyong Lee Department of Software and Computer Engineering, Ajou University, Suwon, Rep. of Korea, East Asian.
Manuscript received on 08 June 2019 | Revised Manuscript received on 14 June 2019 | Manuscript Published on 22 June 2019 | PP: 89-93 | Volume-8 Issue-8S2 June 2019 | Retrieval Number: H10170688S219/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open-access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: ADAS applications using optical camera should have accurate view orientation for correct computation. We need calibration of view orientation which is changed from initial installed status for impact, deformation, etc. In this research, we proposed automatic computation methods for camera orientation with using optical flow captured movie and vehicle motion data without longtime task at a vehicle workshop. We used small set captured images for about 4seconds without correction for lens distortion and collected vectors of optical flow. Collected vectors of optical flow modified by using camera lens intrinsic in formation and using mathematical regression we calculated camera calibration data. We eliminated part of vectors of optical flows far from regression curve for we considered them as outlier data. We implemented using OpenCV API, made cases using equipment of 3degree of freedom camera motion control and tested on our implementation. We found that we can get camera orientation parameters, with accuracy range less than 5 pixels. Part of this error may come from initial configuration error and computation error. Therefore, fine tuning of calibration may be required after our proposed method however, fine tuning process can be more efficient with using calculated result from our method. Our experimental results show proposed method are fulfilled for fast detection of camera calibration data in ADAS application like 360 around view system with reasonable accuracy, performance. Furthermore, method for exclude outlier data work efficiently.
Keywords: Optical Flow, Camera Calibration, ADAS, Around View, Camera Transform.
Scope of the Article: Systems and Software Engineering