气象云图分级配准方法研究

Research on Nephogram Hierarchical Registration Method

气象云图序列图像由气象卫星分时拍摄而获得。在相邻两幅图像的拍摄时间间隔内,卫星传感器的位置及大气云层的形态都发生了一定的变化。因此,气象云图序列图像同时具有刚性形变和非刚性形变。针对这一特征,提出气象云图分级配准方法。首先采用基于Forstner特征点的刚性配准算法,校正由于传感器位置变化引起的旋转和平移等图像整体刚性形变;然后采用基于B样条的非刚性配准算法,校正云层局部扭曲等非刚性形变。实验结果表明,分级配准方法不但提高了图像配准的精度,使其达到亚像素级,还提高了算法的运行速度。

Serial nephograms are obtained by meteorological satellite with fixed time interval.During the time interval,not only the position of the sensor,but also the form of the atmospheric clouds has changed.As a result,the serial nephograms have both rigid and nonrigid deformation.Due to this characteristic,a hierarchical registration method is proposed.First of all,a Forstner-based rigid registration method is applied to correct the overall rigid deformation of nephogram caused by the position change of the sensor,then,the local cloud distortion is registered by the nonrigid registration method based on B-spline.The algorithm has been applied to the fully automated registration of nephograms,and the results clearly indicate that the proposed hierarchical registration method can not only achieve sub-pixel precision,but also decrease the runtime of the process.