基于多尺度Fourier-Mellin变换的末制导目标跟踪

针对光电成像末制导阶段目标尺寸迅速增大且可能伴有旋转的稳定跟踪问题,采用了一种基于多尺度Fourier-Mellin变换的目标跟踪方法.根据高斯尺度空间理论和Fourier-Mellin变换,构造了基于多尺度Fourier-Mellin的图像变换参数估计策略.利用平均绝对差分匹配准则对模板和待匹配图像进行图像匹配,当匹配误差过大时,基于多尺度Fourier-Mellin变换进行模板参数估计,求出模板与当前目标之间的尺度和旋转变换参数,并利用双线性内插调整模板,求得目标匹配位置,对模板进行刷新.仿真结果表明,该算法能够适应末制导阶段目标尺寸的急剧变化,实现对目标的稳定跟踪,其跟踪精度和稳定性优于传统方法.

Terminal guidance target tracking based on multi-scale Fourier-Mellin transform

To achieve robust terminal guidance tracking of a ground target whose size and pose change rapidly, a new approach based on Multi-Scale Fourier-Mellin Transform (MSFMT) is used. MSFMT based on Gaussian scale-space and Fourier-Mellin Transform is proposed to evaluate the image transform parameters. Image matching between the template and the query image is performed based on Mean Absolute Difference (MAD). When matching error is too high to track the target robustly, the template scale and rotation parameter evaluation based on MSFMT is performed. The template is adjusted based on bilinear interpolation according to the template parameters. MAD match is performed again to get the accurate target location. The template is updated when the template size has changed. Experimental results show that the proposed method is robust to rapid changes of target size, and provides robust tracking during terminal guidance.