基于BRISK和CamShift的鲁棒目标跟踪研究

摘要

在CamShift目标跟踪的框架下,提出了基于BRISK特征匹配和CamShift的目标跟踪方法,该方法通过颜色特征和局部特征共同定位目标,从而实现目标准确跟踪.在保证跟踪实时性的前提下,该方法改善了CamShift算法在目标跟踪过程中对背景、尺度、旋转和遮挡变化的敏感性和目标跟踪的特征单一性.实验对比结果表明,该改进方法较单独基于特征匹配的目标跟踪方法,其快速性有很大提高,较CamShift跟踪方法在目标尺度变化、旋转变化、光照变化、背景变化和遮挡变化条件下的鲁棒性更强,同时增强了两种算法的跟踪准确性.%A target tracking method based on BRISK feature matching and CamShift is proposed under the framework of CamShift target tracking.In this method,both the color features and local features are used to locate the targets for realizing accurate target tracking.Compared with traditional CamShift,this method can keep the real-time performance,while improving the sensitivity to the changes of background,dimensions,rotating and shading,and the oneness of the characteristics of the target tracking.The contrast experimental results show that:1) The tracking method proposed is faster than the method based only on the feature matching;2) Compared with CamShift,it has higher robustness to the changes of background,dimensions,rotating and shading;and 3) The precision of tracking is improved.