Abstract: In order to simplify image processing and improve the real-time performance for binocular vision-based pose measurement of parallel mechanism,a method for predicting marker imaging window is proposed.It makes parallel mechanism pose prediction based on dynamic model of the mechanism and Unscented Kalman filter(UKF) algorithm,establishes error propagation model between marker center image and parallel mechanism pose according to error theory,and consequently realizes the prediction of marker imaging window.Finally,simulation experiments are conducted to verify the feasibility of the prediction for marker imaging window.