Translator Disclaimer
2 February 2009 Correspondence of feature points on moving object in tracking system based on stabilization
Author Affiliations +
Camera systems are often unsteady on platform of airborne, car borne and ship borne. Stabilization algorithm can be used to eliminate impact of vibration. But image sequence after processing is different from original sequence. If there is a moving target in camera field, feature points on the target must be indentified and made sure corresponding relationship in processed sequence. To solve the problem that moving target features position and correspondence are difficult to identify in image sequence after image stabilization processing, background updating difference moving target detection algorithm based on motion analysis is proposed. It uses subsample mean and subsample variance and introduces the concept of background gray probability to identify feature points of moving target in the steady image sequence. In addition, to solve the problem of incomplete motion track of feature points caused by obstruction or weak target detection algorithm, partial limit incomplete smooth track algorithm is proposed. It is used to identify correspondence of feature points on the moving target, and to solve temporary occlusion of moving object. Experimental results show that moving target features position and correspondence can be identified quickly through the two algorithms. Single-frame processing speed can reach an average of 27 ms with DSP6416 processor. Image stabilization algorithm and the two algorithms can be combined to realize real-time tracking based on image stabilization.
© (2009) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Bin Wang and Yue-jin Zhao "Correspondence of feature points on moving object in tracking system based on stabilization", Proc. SPIE 7157, 2008 International Conference on Optical Instruments and Technology: Advanced Sensor Technologies and Applications, 715719 (2 February 2009);

Back to Top