基于传感器参数和目标轮廓中心的自动配准算法研究

Automatic registration algorithm based on sensor parameters and target's contour centroid

通过对光电成像型反舰导弹的成像过程分析,提出了一种自动配准算法。其基本思路是将图像变换模型分解,逐步简化。可见光和红外图像配准时的变换为仿射变换。首先,利用传感器参数的调整,消除图像间的比例变化,将仿射变换简化为刚体变换;然后,依赖图像信息,用形态学边缘检测的方法求取目标的轮廓中心,以此为控制点消除图像间的平移变化,实现图像的完全配准;最后,通过观察海天线是否重合以及利用均方根误差原则,对算法的配准效果做详细评估。仿真实验表明,该算法准确、快速,配准精度满足目标识别的要求,可以较好地解决异类传感器弱小目标图像配准的难题。

Based on the analysis of the imaging process of optoelectronic imaging anti-ship missile, an automatic registration algorithm is proposed, and its basic idea is to decompose the transform model and simplify it step by step. Originally, the distortion between infrared and visible images is affine. First, by adjusting sensor parameters, the scaling change between images is eliminated and the affine transform is simplified into rigid transform. Then depending on image information, the centroid of the target's contour is computed by morphological edge detection and chosen as control point, which is used to eliminate the translational change between images and achieve the complete alignment. Finally, the registration effect is assessed by judging whether the sea-sky-lines of the two registered images are superposed and using the rule of root mean square error. The simulation experiments convince that the algorithm is accurate, fast, and can meet the precision requirement for target recognition, providing a good way for solving the difficult registration problem of small target images with different sensors.