摘要
针对在目标与背景颜色相近的场景中,Cam Shift算法不能实现目标准确跟踪的问题,提出改进Cam Shift的目标跟踪算法.通过将背景颜色设置为初始搜索窗口颜色直方图模型中最小分量值对应的色彩值,增大目标与背景区域之间的颜色差值,解决相似背景的干扰问题.利用曲线拟合方法判断预测窗口的有效性,实现对预测窗口的及时校正.实验表明,在目标与背景颜色相近的场景中,改进算法与传统算法相比具有更高的跟踪精度,并能纠正由跟踪目标逐渐变小、目标遮挡等情况引起的跟踪失效.
Due to the limitation of CamShift algorithm in the scene where the target＇ s color is similar to the background, this paper proposes an improved-CamShift algorithm to achieve an accurate object tracking. Setting the background color as the minimal value of the color histogram of the initial search window, we widen the difference between the target and background region and solve the inter- ference of similar background. The curve fitting method is used to judge and revise the validation of the prediction window timely. Fi- nally, for the situation whose target and background color is similar, the experimental results show that our improved algorithm has a higher tracking accuracy through the comparison with traditional algorithm. It is also proved that the improved-CamShift can overcome the tracking failure which is caused by the diminishing tracking target, the target occlusion, etc.
出处
《小型微型计算机系统》
CSCD
北大核心
2015年第6期1325-1330,共6页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61271304)资助
北京市教委科技发展计划重点项目暨北京市自然科学基金B类重点项目(KZ201311232037)资助
北京市属高等学校创新团队建设与教师职业发展计划项目(IDHT20130519)资助
