一种基于单演相位特征的动态目标跟踪方法

A Method of Tracking Active Object based on Monogenic Phase

颜色作为一个有效的视觉特征,被广泛用于目标跟踪中。但在光照变化环境下,颜色会发生变化,从而导致基于颜色的跟踪算法性能不稳定。单演相位作为包含图像重要信息的特征,具有在光照变化环境下性能稳定的特点。为了提高在光照变换环境下的跟踪稳定性,提出一种基于单演相位的动态目标跟踪方法,首先运用高斯模型更新背景并得到各动态目标,再计算出各动态目标的单演相位直方图,最后和待跟踪目标单演相位直方图模板进行匹配得到Bhattacharyya系数,比较各Bhattacharyya系数确定并跟踪目标。实验结果证明该方法在光照变化环境下具有很好的跟踪鲁棒性。

Color,an available visual character,is widely used to track video object.But in the environment of variant lighting,the color is easy to change,and it makes the tracking method based on color not stable.Monogenic phase as a character including important informa tion of an image,has the property of brightness invariant.In order to improve the stability of tracking in the environment of variant light ing,a new method of tracking active video object is proposed in this paper based on monogenic phase.In the proposed method,the Gauss ian model is implied to update the background and then the active objects can be detected.After that,the phase histogram of every active object is calculated to make a comparison with the monogenic phase histogram of the object being tracked and thus the Bhattacharyya dis tance is computed to confirm which object is the best match to the video object being tracked.The experimental results demonstrate that the proposed method can achieve robust tracking performances in the environment of variant lighting.