This paper presents an optical flow-based novel technique to perceive the instant motion velocity of a smart wheelchair robot. The primary focus of this study is to determine the wheelchair’s ego-motion using a displacement field in temporally consecutive image pairs. In contrast to most previous approaches for estimating velocity, the proposed strategy has two main innovations. Firstly, the proposed tilted overlooking camera set-up instead of conventional downward-looking camera and the corresponding ego-motion model is presented for compact indoor mobile robots. Secondly, by virtue of the graphic processing unit-accelerated TV-L¹ algorithm, which is coupled with motion priors-based pixel prediction, we are permitted to improve the accuracy and efficiency of the optical flow estimation significantly. In order to render our method more robust with respect to noise and outliers, we propose a quadratic motion model-based random sample consensus (RANSAC) refinement of flow fields. Advantages of our proposal are validated by real experimental results carried on our smart wheelchair platform and contrast evaluations conducted on Pioneer robot.