Abstract: With a spatial resolution of up to 1m, SAR data has significantly improved applicability of land remote sensing project. Feature point locating accuracy was analysis based on synthetic aperture radar imaging mechanism and GCP selection principles were proposed. Suitable SAR geometric correction models were suggested through analogy analysis. In this work, TERRASAR、COSMO SkyMed and RADARSAT-2 data were used for deep study. Tests showed that 10-15 GCP is required to obtain good result, and about 3-5 m RMS for 1m spotlight mode image and about 5~8 m for 3m strip mode image which respectively meet the requirements of 1: 10,000 and 1: 25,000 land survey monitoring geometric accuracy.