| With the development of computer vision,image processing and signal processing technology have been widely used in many fields,such as navigation,intelligent driving and automatic control.In recent years,many scholars and researchers have done thorough research on video tracking,and study on further innovation based on the original classical algorithm.Although the algorithms which is Mean shift and particle filter are widely used in practical engineering,the two algorithms have their own advantages and disadvantages.Mean Shift algorithm and particle filter are combined with improved based on their advantages in this paper,so it can still have real-time and robustness in complex background.The specific research carried out as follows:This paper first introduces the application of Mean Shift algorithm in target tracking,and proves that the effect of Mean Shift is not good when the size of target is changed or the target is similar with background or a part of the target is blocked.Mean shift is used in target tracking for the advantages of Mean Shift is less computation and high real-time performance.According to the shortcomings of Mean Shift's,the original Mean Shift algorithm is improved from three aspects: bandwidth window,target modeling and model updating strategy.Then,the experiment is carried out,and found that the improved Mean Shift algorithm is good at the situations that the size of target is changed or the target is similar with background or a part of the target is blocked,but the rapid movement of the target or a large change in shape,the effect is still not ideal.Then,the basic principle of particle filter is expounded,and a multiple feature that color,texture and structure feature in particle filter algorithm is proposed to improve the ability of particle filter to solve non-linear problem.Combining the Mean Shift and particle filter,a new target tracking algorithm based on improved Mean Shift and multiple feature particle filter is proposed.Compare to the particle filter algorithm and mean shift,the fusion algorithm is as good as particle filter,but the average running time is greatly shortened through the three group video test,and proved that the fusion algorithm has high accuracy and real-time performance. |
