基于多线索融合跟踪算法的录播系统设计  

Design of Recording and Broadcasting System Based on Multi-cue-fusion Tracing Algorithm

作　　者：戴雯惠[1] 

机构地区：[1]苏州经贸职业技术学院信息系,苏州215009

出　　处：《煤炭技术》2012年第12期158-160,共3页Coal Technology

摘　　要：针对传统Camshift跟踪算法在目标运动速度过快和受到大面积相似区域情况下,容易引起跟踪目标丢失的问题,提出了一种基于多线索融合的跟踪算法来提高运动目标跟踪的精度。首先利用运动目标具有连续性的特点,采用Kalman位置预测原理实施有效跟踪,更好的解决了Camshift跟踪算法在跟踪过程中环境因素对跟踪目标的影响。然后通过加入Adaboost算法有效提高了教师人脸跟踪的准确性。改进的算法分别通过了阻挡、变形和相似肤色实验验证。实验表明改进后的算法提高了检测时间和识别率,实现了对快速运动目标的稳定跟踪。For the traditional Camshift tracing algorithm is likely to loss its tracking target when the target moves too fast or is interfered by large scale of similar background areas, this article introduces a new tracing algorithm based on multi-cue-fusion to improve precision of moving target tracking. First, this tracing algorithm uses Kalman＇s position prediction theory to carry out effective tracking based on the continuity of moving target which effectively avoids the interference of background area on tracing target existed in Camshift tracing algorithm. Then, by adding in Adaboost algorithm, the new tracing algorithm remarkably improves the accuracy of face tracing. And this revised algorithm has passed blocking, morphing and similar complexion experimental verification separately. The improved the detecting time and recognition experiments show that the ratio while achieving steady target.

关 键 词：目标跟踪 CAMSHAFT KALMAN滤波 ADABOOST 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]