基于颜色传递的双波段目标跟踪算法研究

Dual-band target tracking based on color transfer technique

针对复杂背景下Camshift算法容易丢失目标的情况,提出一种基于YCBCR空间将红外与可见光融合图像与彩色参考图像进行颜色传递后,采用Camshift进行目标跟踪的算法。该算法在颜色传递时充分利用双波段图像信息,得到的目标对比度高并且颜色空间较其周围背景突出,增强了目标的颜色概率图,提高了Camshift算法效率。实验表明,通过对可见光图像、经颜色传递后的红外图像以及传统颜色传递方法得到的图像采用相同跟踪算法进行定性分析,在该算法得到的图像中,跟踪窗口中心相对于目标质心仅有3个像素的误差,跟踪精度远远优于对比实验图像的跟踪结果,并且算法的跟踪时间为每帧20.6ms,达到了实时性的要求。

Since Camshift algorithm is easy to lose tracked object in complex background, this paper puts forward a tracking method based on Camshift algorithm, before which a color transfer technique is used between color reference image and fused image in YCBCR space. During color transfer, the proposed method used information of dual-band image effectively, and the target not only had the excellence of high contrast, but also enhanced the drawing of color probability distribution, the proposed method improved the efficiency of Camshift. The comparison of the visual image, color transferred infrared image and the traditional color transferred image was made using the same tracking method, a group of experimental results demonstrate that the distance between the center of tracking window and target centroid is just 3 pixels in the pictures based on proposed method, tracking precision outperforms the results of contrast experiment pictures. At the same time, the tracking time of the proposed method is 20.6 ms, which meets the real time requirement.