针对无人艇载高清多光谱影像特性,提出了集成单应性矩阵内配准与置平薄板样条外配准的多光谱影像整体配准方案:(1)傅里叶-梅林变换引导下的多光谱影像相关匹配自动生成连接点;(2)单应性矩阵最小二乘联合估计实现多波段影像相互配准(内配准);(3)构建“置平”薄板样条变换模型实现多光谱影像序列无地理坐标配准(外配准),最后对无人飞艇搭载多CCD相机低空拍摄的高清多光谱影像进行配准验证,取得了约1个像元误差的内配准精度以及较好的外配准视觉质量。

High-solution multispectral imagery is attractive to various users and can be collected with a group of Charge-Coupled Device (CCD) cameras on board low-altitude Unmanned Airship (UA). Due to imperfect system integration and inherent nature of UA, however, the registration of High-solution multispectral imageries from UA are subjected to several challenges. In this paper, a low-cost multispectral imaging system composing of four CCD cameras is described as well as its potential issues to registration operation are summarized. After that, Fast Fourier Transformation (FFT) induced multispectral imagery Cross-Correlation Matching technique is developed to automatically generate high-precision tie pints for overlapped UA imageries. Next, Homograph joint estimation based on the Least Square adjustment is described in detail to align multispectral imagery acquired at each exposal and “leveled” Thin Plate Spline (TPS) transformation is established to register multispectral imagery sequences as well. Finally, experiments of registering high-solution multispectral imageries collected from low-altitude unmanned airship are implemented with proposed approaches and valuable conclusions are conducted.