COMBINED MULTI-VIEW MATCHING ALGORITHM WITH LONG-STRIPS OF SATELLITE IMAGERY FROM DIFFERENT ORBITS

Abstract. Existing matching algorithms aim to match conjugate points among overlapping satellite scenes acquired from the same orbit and can generally achieve good matching performance. Unfortunately, no algorithm can avoid the difficulty of simultaneously processing the data sets of long-strip imagery acquired from different orbits. In this paper, the combined matching algorithm we propose introduces the LBP/C operator, which, when combined with existing feature detectors for the first time, can make possible the extraction of more stable interest points and candidates. At the same time, based on the typical characteristics of Chinese satellite imagery, we improved the filter method and achieved an effective combination of several image matching algorithms. A comparison among several kinds of matching transfer modes was presented; and to evaluate this algorithm, Chinese Mapping Satellite-I data are used as the reference data.