融合灰色预测和HOGI特征的红外目标跟踪方法  

Novel IR target tracking method based on the grey prediction and HOGI feature

作　　者：张齐[1] 曹琦[1] 王晓薇[2] 毕笃彦[1] 

机构地区：[1]空军工程大学工程学院,陕西西安710038 [2]空军工程大学电信工程学院,陕西西安710077

出　　处：《西安电子科技大学学报》2010年第4期751-757,共7页Journal of Xidian University

基　　金：国家863计划资助项目(2007AA701206)

摘　　要：针对红外目标跟踪时不能有效提取目标特征这一问题,提出了一种新的红外目标稳健跟踪方法.首先利用灰色模型预测目标位置,然后建立以此预测位置为中心候选目标区域的方向梯度-灰度直方图特征模型,并将此特征引入到Mean Shift算法中,以实现对红外目标的精确跟踪.仿真结果表明了该方法跟踪精度高,在目标出现部分遮挡或全部遮挡时,仍能跟踪目标,确保目标不丢失,体现出该跟踪方法良好的鲁棒性.Generally,the IR target tracking method is unable to extract the IR target＇s feature effectively during the process of IR target tracking.For solving this problem,a novel and robust IR target tracking method is proposed.The method adopts a Grey Model to predict the target＇s position firstly,and then constructs the Histogram of Oriented Gradient and Illumination（HOGI） of the candidate target area which centers on the predicted position.Finally,the method incorporates this HOGI feature into Mean Shift and realizes the precise tracking.Experimental results show that the proposed method has an excellent tracking precision.When a part or whole of the target is sheltered,the tracking goes on to make sure the target is not lost,which reflects the robustness of the method.

关 键 词：灰色模型 方向梯度-灰度直方图特征 均值漂移 目标跟踪 图像处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]