ABSTRACT
In the field of automotive engineering, the estimation of ego-motion plays an important role in robot control. In this review, we showed a general roadmap of camera ego-motion estimation, introduced the methods of optical flow estimation from image sequences, and compared three methods of ego-motion estimation from estimated optical flow in detail. We discussed the specialties of the mentioned methods and analyzed the error sources. We concluded that the optical flow-based ego-motion estimation methods can be used for velocity estimation but are not suitable for position or pose estimation. We encouraged researchers to focus on finding approaches to error elimination and finding less complicated methods.
