車載単眼カメラを用いた自車両の運動推定

抄録

This paper proposes a method for estimating the ego-motion of the vehicle by using a vehicle mounted monocular camera. Our proposed method is based on Structure from Motion (SFM). There are two problems with applying a SFM algorithm in a real road environment. Firstly, a typical road scene contains moving objects, such as other vehicles. Secondly, roads display fewer feature points, compared to the number associated with background structures. These problems cause false estimations or reduce the accuracy of the estimated motion. Our proposed method detects moving objects in every frame and selects feature points used for motion estimation from all extracted feature points in the image, so that those feature points are dispersedly distributed in various regions except those containing moving objects. The ego-motion of the vehicle can be accurately estimated from correspondences of these selected feature points. In our experiments, it has been shown that the proposed method is able to estimate the ego-motion in real road scenes.