基于分形几何边界提取的图像跟踪方法

为了满足现代武器装备对图像跟踪系统的实时性和跟踪可靠性的双重要求，针对武器装备所跟踪目标的特点，将基于分形几何的边界提取方法用于图像跟踪系统，提出了一种基于运动区域和边界图像的目标跟踪方案。文中的边界提取方法不仅能提取出目标的轮廓，还能较好地保留其内部细节，这样通过模板匹配得到的最佳匹配位置更加可靠。同时，基于运动区域的边界提取，尤其是对于远景的目标跟踪大大减少了计算量。此外，在利用边界图像进行匹配的基础上，采用了自适应模板更新策略，目的是尽可能地克服形变、光照等影响，使跟踪过程准确可靠。仿真实验表明，该方法运算量小，能满足武器装备实时性要求。

Image tracking method based on fractal geometry edge extraction

In order to meet real time and tracking reliability requirements for image tracing system installed in modern weapon system, based on the target signatures being tracked by the system, fractal geometry based edge extraction method was adopted in an image tracking system, and a target tracking solution based on moving area and edge image was put forward. The edge extraction method not only extracts the outline of the target, but also keeps the inner details, thus the optimum matching position found by template matching becomes more reliable. Meanwhile, edge extraction is only processed in moving area to reduce computation greatly, especially for long-range tracking. In addition, based on the matching with the edge image, the adaptive template modification is applied in edge image matching to minimize the effect of distortion and illumination variation etc. and make the tracking process more accurate and reliable. Simulation results demonstrate that this method demands much less computation and could meet the real-time requirements of the weapon system.