基于阴影图像本征值的室外目标跟踪

Outdoor Target Tracking Based on Intrinsic Values of Shadow Images

针对阴影导致目标跟踪中的目标丢失问题,提出了一种基于图像本征值的室外目标跟踪方法.首先,根据相机成像时阴影内外像素值的线性关系,推导出一种新的阴影图像本征值计算方法;然后,将经过阴影的目标所在区域变换到本征值生成的本征图像中;最后,利用MS(mean shift)跟踪算法在本征图像中进行目标精确定位.此方法不需要任何相机的相关参数信息,计算速度快.本文对多组含有阴影的序列中的目标进行了对比跟踪实验.与其它跟踪方法相比,本文的方法具有更高的跟踪精度,计算速度也可以达到实时的要求.实验结果说明本文方法对室外复杂光照条件下的跟踪具有更好的适用性和跟踪精度.

For the target losing problem caused by shadows during target tracking, an outdoor target tracking method is proposed based on the intrinsic value of a shadow image. A new method is proposed to calculate the intrinsic value of a shadow image based on the derivation of a linear relationship between the pixel values inside and outside the shadow areas. The sequence images are transformed into intrinsic images, which are insensitive to shadows, and then a mean shift tracker is employed to locate the object in the image. Our method doesn＇t need any of the camera＇s parameters, and has a high calculation rate. Object-tracking experiments have been done on sequence images with shadows, Compared with other tracking methods, our method has a higher tracking accuracy, and the calculation speed can achieve real-time requirements. The experimental results validate the performance of our method.