고해상도 위성 데이터 기하보정 정확도 향상을 위한 영상 분할 기반 RPCs 보정

Typically, the geometric correction of high-resolution satellite images is performed using RPCs (Rational Polynomial Coefficients). For high-resolution satellite data obtained from stable platforms such as large satellites, simple bias compensation of RPCs is often sufficient to achieve significant improvements in geometric accuracy. However, when the attitude during data acquisition is unstable, simple bias compensation across the entire image may not be sufficient to ensure accuracy. This issue is likely to become more pronounced as we move toward medium- and small-sized satellites. Therefore, this study proposes dividing the image into segments and performing RPCs bias compensation for each segment to address more complex image errors that occur across the entire image. Considering the characteristics of pushbroom images, the segments were divided along the line direction, and connectivity constraints were applied to ensure continuity between adjacent image segments. Experiments were conducted using data from Kompsat-3 satellite, and the results confirmed that this method enables more stable geometric correction compared to traditional RPCs bias compensation applied to the entire image.