基于特征点的整数小波目标跟踪方法的研究

基于提升方法的整数小波变换的诸多优点,以Harr整数小波变换为例,提出了动态目标跟踪算法。对标准图像进行整数Harr小波变换,并将提升项取整,对提升项的参数用一定数量训练图像进行学习。选取在训练参数平方和为最小意义上的、使整数小波变换后的图像高频部分具有较大值的点作为特征点。对包含有目标的参考图像进行整数小波变换,选择高频分量具有较大值的点,利用训练过的提升项参数使目标和基准图像配准。由于算法采用整数小波变换,使提取具有较好的鲁棒性,从而实现目标发生旋转、平移及尺度变化等的跟踪过程。仿真实验表明,该方法能对动态运动目标进行跟踪。

Research of a target tracking method based on features of integer wavelet transform

Based on lifting scheme having many advantages for integer wavelet transform, an algorithm of target tracking using Harr wavelet transform was presented. The integer Harr transform was performed to the image. Free parameters in the lifting term were learned by using training images. Determining these free parameters is to minimize their squared sum and compute them only for the points where high frequency components are large. Low and high frequency components of reference image were computed using Harr integer wavelet transform. The positions where the high frequency components are large were chosen. At these positions, by combining the already determined parameters, the image could be registered. Tracking target was realized resisting the changes of orientation, scale and translation. Simulating experiment illustrated that the approach can track moving target.