摘要
针对传统Mean Shift跟踪算法在进行目标跟踪时,由背景因素带来的定位偏差和缺乏相应的模型更新策略而易陷入局部最小值的情况,提出了两方面的改进措施。一方面在建立目标模型时,对背景像素建立新的模型以弱化对目标模型的影响,另一方面在跟踪过程中,融合目标颜色特征和连续两帧目标中心的欧氏距离动态的决定目标模型更新策略。实验结果表明,该算法在目标姿态、环境光照变化强烈时均能取得较好的跟踪效果。
Considering the defects of traditional mean shift in object tracking, that is the locating accuracy may decrease caused by background and converged to local optimum due to no corresponding model update strategy, two kinds of improved methods were proposed. One is to create a new feature model for background pixels to weaken the influence on target model, the other is model update strategy, which fusion the color features of target and Euclidean distance between the target center of present and last frame. The results indicated that the improved algorithm can robustly track the target under complex scenes.
出处
《软件导刊》
2012年第9期30-32,共3页
Software Guide
