DETERMINATION OF OPTIMAL GROUND SAMPLING DISTANCE FOR MATCHING GCP CHIPS AND SATELLITE IMAGES
Keywords: CAS500 Series, GCP Chip Matching, Image Matching, Geometric Correction, Sensor Model, Kompsat-3A, RapidEye
Abstract. Many satellites for high resolution image acquisition are being developed very actively worldwide. In Korea, the Government has started a new space program, CAS (Compact Advanced Satellites), to promote image utilization and national space industry. For the successful utilization of satellite images, the geolocation accuracy of them is very important. A research has been performed for automated GCP (Ground Control Point) generation and precise sensor modelling by matching small image chips centered at GCPs against satellite images. In this study, we examined whether GCP chips at 25cm GSD created for CAS-1 images can be used for CAS-4 images at 5m GSD and, if so, whether there is an optimal GSD to improve the matching between GCP chips and CAS-4 images. Due to data availability, experiments were carried out using Kompsat-3A images at GSD of 55cm, as a replacement to CAS-1 image, and RapidEye images at GSD of 5m, as a replacement to CAS-4 images. Original satellite images were upscaled to make test images with various GSDs. GCP chips were downscaled to make their GSD identical to test images. We performed matching with the downscaled GCP chips and the test images. Experiments showed that it was possible to reuse GCP chips at 25cm GSD for satellite images at 5m GSD. When RapidEye images were upscaled to the range of 1.67m GSD (one third of the original GSD) to 1.25m GSD (one fourth), we achieved improved geolocation accuracy then matching GCP chips at the original images (5m GSD).