基于对数极坐标变换的景像匹配算法研究

针对成像制导过程中对实时性和准确性的要求,对存在旋转和尺度变化的航拍照片的匹配进行研究,提出了一种基于对数极坐标变换的匹配和跟踪算法.匹配时先对基准图和实时图按多分辨率金字塔法进行对数极坐标变换,利用对数极坐标的旋转和尺度不变性进行匹配.匹配后,用相对运动时目标的对数极坐标变换图像梯度下降最小相异评估方法对目标进行跟踪,通过对航拍照片进行匹配测试,证明了这种方法的有效性.

Research on Scene Registration Algorithm Based on Log-polar Transform

In order to meet the real-time and precise guiding demand in missile end game, a regis- tration and tracking algorithm based on log-polar by researching the rotated and scaled airborne imagery was brought forward. Log-polar transformation of reference and target images with multiresolution pyramid was carried out. The registration algorithm was carried out by the rotate and scale invariance property of log-polar coordinate in scene registration process. By using log-polar image gradient descent for minimizing a dissimilarity measure, mission was tracked. The registration of the airborne imagery shows that the method is efficient.