INVESTIGATIONS INTO GEOMETRIC RECTIFICATION OF REMOTE SENSING DATA

Abstract

Remotely sensed images recorded by sensors onboard satellites and aircraft contain geometric distortions. The quantitative use of remote sensing images in many applications requires that the geometric distortions inherent in these images be corrected, or rectified, to a desired map projection. Geometric rectification plays an important role in applications like land cover mapping, image mosaicing, change detection analysis (multitemporal studies) etc. The conventional procedure of geometric rectification relies on the use of Ground Control Points (GCPs) located in the image and the corresponding map in order to empirically determine a mathematical coordinate transformation to correct the geometry. However the accuracy of geometric rectification through conventional procedure depends upon many factors such as quality and quantity of GCPs and their spatial distribution, the order of polynomial transformation, topographic variation, type of remote sensing data, quality of reference data, and resampling method. It is thus necessary to assess the effect of each factor on geometric rectification, to achieve better results. In the present study, some of these factors have been investigated to evaluate their effect on geometric rectification of IRS 1B LISS II remote sensing data acquired for plain and hilly areas both. The geometric rectification with uniformly distributed GCPs has provided significantly better results than those produced by locating the GCPs either on the edges or clustered in an area. There has not been a significant improvement in rectification accuracy as the order of polynomial is increased from first order to second and third order. In fact the accuracy decreased as the order has been increased. A general observation is that as the number of GCPs has been increased, the accuracy of geometric rectification has also increased. On analyzing the effect of resampling method, it has been found out that digital numbers are altered more with.cubic convolution and bilinear interpolation than with nearest neighborhood. Finally it has also been shown that requirement of number of GCPs is more for rectification of an image of hilly areas than of plain areas in order to achieve better geometric rectification. It can thus be concluded that these factors need due consideration for obtaining proper geometric rectification of remote sensing images in order to derive meaningful results.