High precision geometric correction based on DEM data and GCP outlier detection for MODIS data

Abstract

Terrain Height Error is one of the most important factors which influences the accuracy of direct system correction and has an obvious effect along the scanning direction. In this paper, taking Moderate-resolution Imaging Spectroradiometer (MODIS) data received from Earth Observing System (EOS) satellites as an example, and focusing on the high precision geolocation techniques of the moderate resolution optical satellite data, a high precision algorithm based on Digital Elevation Model (DEM) data and Ground Control Point (GCP) outlier detection is proposed to correct this kind of influences of Terrain Height Error. For more efficient and intelligent processing in practical applications, a group of GCPs should be used in the further geometric correction. But the correction result using GCPs is not so stabile, because it is often influenced by the data conditions. In this paper, a new idea based on local distortion similarity is proposed for the GCP outlier detection too. In this method, the distortion is described by a 2-D Gaussian distribution in the local area, and a decision rule is made to find out the abnormal GCPs according to the variance of the Gaussian distribution. Some experiment results using MODIS data show that the proposed algorithm based on DEM data and GCP outlier detection is feasible and valid. The final average location error of the new algorithm is about 0.13 pixels which is sufficient for many application requirements.