MODEL OF QUANTITY AND QUALITY OF PHOTO CONTROL POINTS FOR THAICHOTE SATELLITE IMAGERY

Abstract

The research was to study a model of monitoring the quantity and quality of image control points for Thaichote high-resolution satellites using Hold Out Validation (HOV) and Leave One Out Cross Validation (LOOCV) methods for screening control points and picture quality by estimating the sensor's external orientation as a constant for estimating the sensor's position ( ) and the sensor's X-axis of inclination ( ), the sensor's Z-axis of inclination correction ( ) is a polynomial squares equation. This requires 8 unknown coefficients, at least 4 ground-based photo control points (GCPs) are needed to improve the sensor model. The sensor model improvement results were applied to correct the ortho image and place the GCPs in the image boundary parallel to the orbital path, the sensor model improvement result was an RMSE of correction equal to 0.21 m and ortho image accuracy in accordance with NSSDA standards. It was found that the ortho images using 13 GCPs and 59 independent checkpoints from the HOV method had a resolution of 1.52 m (2.98 image points) and ortho images using 8 GCPs and 44 checkpoints. The LOOCV method has an accuracy of 0.84 m (1.63 image points), indicating that the LOOCV method results in better ortho-produced quality than the HOV method.