摘要
针对Mean Shift算法中存在的不足,提出了结合均值偏移和灰色马尔可夫预测模型的目标跟踪算法。该方法利用灰色马尔可夫模型预测目标在当前时刻的中心位置,以此点作为均值偏移算法进行目标搜索的起始位置。同时,提取目标的几何特征,根据目标的面积来改善跟踪窗口的大小,利用"目标模型"和"候选模型"之间目标特征的变化产生模型更新策略。经实验得,该方法能实时稳健地进行跟踪。
According to the existing shortcomings of Mean Shift algorithm, a combination of Mean Shift and the grey Markov forecasting model of the target tracking algorithm is presented. The method utilizes the gray Markov model to predict the center of the target at the present time,and this point is used for a starting position of the tracking window in the Mean Shift algorithm. Meanwhile, a geometric characteristics of the target is extracted. And then, the size of the tracking window is resized on the basis of the target area. The target characteristics＇ change between the target model and the candi- date model is used to produce the model update strategy. By means of experiments,the method is validated to have a good real-time and robustness.
出处
《电视技术》
北大核心
2013年第3期44-46,61,共4页
Video Engineering
