Hybrid Image Stabilization of Robotic Bionic Eyes

Traditional vision system mounted onto robots may have image blurring and shaking caused by unwanted robot motion, which makes it difficult to sense the environment precisely. This paper presents a real-time hybrid image stabilization method combining mechanical motion compensation and electronic compensation to solve the problem based on a 9-DOF vision platform called Robotic Bionic Eye developed by us. In the mechanical video stabilization stage, a distrubtance reduction control method is proposed to suppress the disturbtance. In the electronic video stabilization stage, 3D camera rotational motion is obtained from IMU and 2D image motion is detected by image motion detection algorithms. These motion is further filtered and image sequences are rewarped to remove rotational shake as well as translational shake for a second time. Experiments verified that our algorithm improves the inter-frame transformation fidelity while ensuring real-time performance.