Feature-based image fusion scheme for satellite recognition

Despite the variety of technologies and algorithms studied, satellite recognition is not fully researched in the uncontrolled space environments. In this paper, a low complexity and efficient satellite recognition scheme by fusing infrared and visible image features for recognition was brought forward. Invariant moments are taken to represent the characteristics of satellites´ pictures. Unlike optimal image feature fusion by classic intelligent computing algorithms, a low computation and efficient fusion rules are developed to improve the performance of recognition. Due to the compute power of space-based computer, a new fusion method by associating combined blur and affine moments invariant (CBAI) with Zernike moments is introduced. The experiments results with Semi-physical simulation images indicate that the recognition consistently demonstrated better performance than others solely based on either infrared or visible image.