基于多特征融合的Camshift结合Kalman预测的实时性视觉跟踪算法研究  

CamShift Based on Multi-Feature Fusion Combines Kalman Prediction Algorithm for Real-Time Visual Tracking

作　　者：刘继承 孙慧琳[1] 赵洪涛 

机构地区：[1]东北石油大学油气田控制动态监测重点实验室,黑龙江大庆163318 [2]大庆市第三采油厂第一油矿测试队,黑龙江大庆163256

出　　处：《自动化技术与应用》2014年第5期54-57,共4页Techniques of Automation and Applications

摘　　要：根据传统CamShift算法的研究,由于它仅采用颜色作为特征,容易产生错误跟踪,本文提出了一种基于多特征融合的camshift结合kalman预测的实时性视觉跟踪算法,采用颜色直方图与灰度梯度直方图,进行加权处理,有效解决背景干扰,实时性跟踪的问题,实验表明本文算法可以实时性跟踪并且在迭代次数上有所下降,提高了跟踪的准确性和效率。According to the tradition CamShift algorithm,it prone to tracking error only by color feature.This paper proposes a visual tracking algorithm that is CamShift based on multi-feature fusion combining Kalman prediction.The algorithm with color weighted histogram and gray gradient weighted histogram is an effective solution to the background interference and real-time tracking.Experiments show that the proposed algorithm can complete real-time tracking and reduce the number of iterations,so it can improve preciseness and efficiency in the tracking.

关 键 词：视觉跟踪 CAMSHIFT 多特征融合 颜色直方图 灰度梯度直方图 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]