Title :
A Novel Motion Prediction Algorithm with Collision Detection
Author :
Tello, Tomás ; Starostenko, Oleg
Author_Institution :
Comput. Sci. Dept., Univ. de las Americas Puebla
Abstract :
The novel algorithm for estimation and prediction of objects motion in dynamic scene is presented in this paper. The proposed hybrid algorithm is based on computing the differences between consecutive frames of video stream by processing of objects time-varying edges. The prediction of objects motion and their possible collisions are computed by extrapolation taking into account proposed motion stabilization technique useful for analysis of irregular displacement. In order to avoid high error of route prediction and to detect the collisions between multiple objects in motion the algorithm operates with clustering areas of image using so-called envelopes. They permit fast motion prediction even the objects in motion have occlusions and rotation. The proposed solution has satisfactory performance for this still open problem of motion estimation and prediction
Keywords :
edge detection; extrapolation; image sequences; motion estimation; object detection; video streaming; collision detection; extrapolation computation; motion estimation; motion prediction algorithm; motion stabilization technique; object detection; time-varying edge processing; video streaming; Envelope detectors; Extrapolation; Image motion analysis; Layout; Motion analysis; Motion detection; Motion estimation; Object detection; Prediction algorithms; Streaming media;
Conference_Titel :
Electronics, Robotics and Automotive Mechanics Conference, 2006
Conference_Location :
Cuernavaca
Print_ISBN :
0-7695-2569-5
DOI :
10.1109/CERMA.2006.8