Collision jitter compensation for mobile robots vision based on optical flow estimation

Collision produces the difficulty to the cooperation between multi intelligent mobile robots guided by vision. Aimed to compensate the collision jitter of robot vision, Eigen-value of the gradient matrix of image was used to detect feature points. Then this paper used optical flow constrain equations to track feature corner points and adopted the least square method to solve the kinematics equations to get 6 motion parameters, and designed Kalman observation model to smooth motion parameters. Compensation model was established based on the affine transformation kinematics. Physical test were implemented on self-develop set-up and stabilization software. Experiments were carried out under intended knocking link fixed with camera. Results show rotation jitters are less than translation jitters under collision, and compensation algorithm integrated optical flow and Kalman filter proposed is valid and real-time with computing time less than 25ms.