Towards Flow Estimation in Automotive Scenarios

Towards Flow Estimation in Automotive Scenarios
René Schuster, Oliver Wasenmüller, Georg Kuschk, Christian Bailer, Didier Stricker
Computer Science in Cars Symposium ACM Chapters Computer Science in Cars Symposium (CSCS-17), Computer Science in Cars Symposium 2017, July 6, Munich, Germany

Abstract:

In this position paper, we discuss different motion estimation methods in general and especially for automotive scenarios. The advantage of 3D scene flow over 2D optical flow is explained by a typical use case. An indirect method for scene flow reconstruction from stereo disparity and optical flow is presented along with its main issue. Further, we describe how a direct estimation can benefit the overall scene flow result. These observations support our statement that research should focus more on direct 3D motion estimation in the future.

Keywords:

Optical flow, Scene flow, Stereo