Abstract: Based on RFM (rational function model) and the polynomial model, the geometric correction precision of TH-1 satellite images was studied on different landforms, and the relationship between the number of GCPs (ground control points) and geometric correction precision was also studied. Experimental results show that TH-1 images have a magnitude distortion caused by optical projection; RFM can obviously improve TH-1 geometric correction precision with only a few GCPs; in the polynomial model, more GCPs can achieve higher correction precision, and 2 order can get the highest correction precision; without a parameter model, high order RFM is the best choice for TH-1 image geometric correction.

Key words: RFM, geometric correction, polynomial model, TH-1 satellite, accuracy