Registration for long-term multi-temporal remote sensing images under spatial constraints from historical data

In long-term multi-temporal images, the objects are very different and the feature matching is very difficult. However, in short-term multi-temporal images, the differences of objects are relatively insignificant. With the small difference, the correct matching points and the right affine transformation can be easily obtained. According to long-term multi-temporal image matching, in the range of time span we insert some historical image to reduce intervals in temporal difference. Through a series of affine transformation between neighboring temporal, the rough spatial relations between the two images can be gradually determined. Under the rough spatial constraints, the accuracy of the points matching can be significantly improved.