基于不变矩特征的红外成像制导跟踪算法研究

随着红外成像制导技术的发展,它对制导跟踪算法的精度和实时性的要求越来越高。制导系统已被要求在20 ms的时间内输出跟踪结果。某些跟踪算法虽然效果较好,但达不到实时输出。经典的制导算法-相关跟踪算法主要基于像素灰度特征进行跟踪,无法解决跟踪过程中由于目标的旋转、膨胀以及仿射变换带来的跟踪点漂移问题。为了解决经典相关跟踪算法无法解决图像旋转时的跟踪稳定性问题以及目标图像急剧膨胀时的跟踪点漂移问题,研究了图像不变矩的特征。利用图像不变矩的旋转-伸缩-平移不变性,选择合适的不变矩特征用于跟踪。通过用归一化的乘法相关函数以及基于相关系数值的模板更新策略设计跟踪算法,解决了复杂背景及强噪声条件下不能对尺寸和形状发生变化的目标进行稳定跟踪的问题。

Research on Tracking Algorithm for Infrared Imaging Guided Missile System Based on Invariable Moment Features

With the development of infrared imaging guided systems, higher requirements are put forward to the precision and real-time performance of guided tracking algorithms. Currently, guided systems have been required to output tracking results within 20 ms. Although some tracking algorithms have good tracking effectiveness, they can not output their tracking results in real time. Since the classical correlating tracking algorithm is mainly based on pixel gray-level features, the tracking point drift problem caused by the rotation, expansion and translation of the target in the process of tracking can not be solved. To solve the problem that the classical correlation tracking algorithm can not operate stably when the image is rotating and the tracking point drift problem present when the target image is expanding quickly, the invariable moment features of images are studied. The features of rotation-scale-translation invariance of images are used to choose the suitable invariant moment features for tracking. The tracking algorithm is designed by using the multiplication correlating function and the template update function based on correlation coefficient. The problem that the infrared target of which the size and shape are changed in complex and yawp environment can not be tracked stably is solved.