摘要
为了获取农业物料在风筛式清选筛面的实际运动规律,通过对多颗粒散体中的目标颗粒进行着色处理,提出采用基于颜色特征向量的Meanshift算法,实现对目标运动轨迹的跟踪。算法根据Bhattacharyya系数的大小判定跟踪目标是否被遮挡,并引入了Kalman滤波器设计。在正常跟踪过程中,利用Kalman滤波器预测每帧Mean shift算法的起始位置,然后利用Mean shift算法对目标进行精确定位,当目标被遮挡时,将其运动视为时不变系统,并通过Kalman滤波器估算目标近似位置。试验结果表明,该方法在复杂背景和光照变化条件下,实现了对快速运动目标的稳定持续跟踪,具有很好的鲁棒性,为散体颗粒运动规律的研究提供了一种图像检测方法。
For tracking target particles＇ motion on air-and-screen cleaning sieve, staining method was used and Mean shift algorithm utilizing the color eigenvector was proposed. Bhattacharyya coefficient was exploited to decide whether the target was occluded. In the frames of regular tracking, the initial point of Mean shift algorithm was predicted by Kalman filter and then the precise position of the target was calculated with Mean shift algorithm. When the occlusion appears, the target motion was regared as a time-invariant system and the position was estimated by Kalman filter. Experiment results show that the algorithm is robust and can track the fast motion target stably with complex baekground and variant light. The usefull image detection technology was provided for motion law research of granular materials.
出处
《农业工程学报》
EI
CAS
CSCD
北大核心
2009年第5期119-122,共4页
Transactions of the Chinese Society of Agricultural Engineering
基金
国家自然科学基金资助项目(50875113)
高等学校博士学科点专项科研基金项目(20060299004)
江苏省六大人才高峰资助项目
