Research on Scale-Adaptive Object Tracking Algorithm by Improving KCF
The SMAKCF(Scale-Adaptive Multiple-Feature Anti-Occlusion KCF)is proposed to solve the problem that the KCF algorithm can not adapt to the object scale and occlusion when tracking an object.The SMAKCF algorithm optimizes simultane-ously several problems including scale response,feature extraction,and update strategy.To be specific,a fusion feature is put for-ward combing HOG features and CN features efficiently,then a scale estimation filter is added and the APCE criterion is introduced to improve the updating method of the position estimation filter.Besides,an extra detection module is designed for re-detecting the object which is unreliably detected.Experiments are conducted on 50 test video sequences of Benchmark to evaluate the algorithm performance.It is indicated that SMAKCF algorithm can overcome difficulty in the scale change and occlusion of the object,the tracking ability in the long-term object tracking process is enhanced significantly.
国家科技期刊平台
NSTL
万方数据
