A frame centre matching approach to registration for change detection with fine spatial resolution multi-temporal imagery

摘要

A frame centre (FC) matching approach to the acquisition and geometric processing of multi-temporal datasets derived from airborne digital frame imagery was tested to determine if registration errors could be reduced relative to more ad hoc acquisition and registration approaches. The root mean square error (RMSE) of test points from image sets derived with the FC approach was at least 50% less than when frames for the second date of imagery were not centred on frames from the first date. The registration RMSE for FC images approached 2 pixels (2 m) for the central portion of image frames that is normally used to generate an image mosaic. The FC approach yielded visual and quantitative improvements in multi-temporal difference images that are commonly utilized for change detection analyses.