一种新的相关跟踪方法研究

相关跟踪过程中，序列图象中的实时图和参考图之间必然存在灰度差异以及一定程度的几何形变和对目标的局部遮挡，而传统相关匹配算法中，每一对像素点对匹配结果的贡献是均等的，这样就使得算法的性能很容易受到图像中噪声、局部遮挡等因素的影响。为此，从另外的角度提出了一种新的图象相似度度量方法，即将目标图象中与模板图象相接近的点的个数作为相似性度量来进行匹配，这种方法得到的相关曲面更尖锐，匹配置信度更高。又因为跟踪过程，实时图必然存在这样或那样的变化，对模板合理地进行更新是相关跟踪的关键，在上述的图象相似度度量方法的基础上，另外又提出了一种非常有效的模板修正方案，从而使得跟踪算法对环境的适应能力和稳定性得到大大提高，实验结果证明了该方法的优越性。

A New Correlation Tracking Method

The presence of the difference of the gray value and the distortion between the object image and the reference image is inevitable, traditional matching method will tend to be degraded by the noise and the occlusion because that every pixel in the image will do equal contribution to the matching result. From another viewpoint, this paper proposed a new image matching method based on a novel distance measurement. It counts not the distance of the gray level of the two images, but the number of the pixels in the object image which are close to the pixels in the reference image. Then applied the measurement method to the process of the image matching. The correlation curve got in this way is much more sharp. So the image matching method has high match precision. And in the procces of the tracking, there must be many change in the object image, then the resonable updation of the reference image will be the key of the tracking problem.On the basis of the similarity measurement, a suitable template updating strategy is presented. The stability of the algorithm has been improved greatly in this way. The experiments showed that this method was very promising.