Evaluation of Control Points’ Distribution on Distortions and Geometric Transformations for Aerial Images Rectification

Abstract

Geometric distortions are inevitable in aerial images. A raw uncalibrated aerial image acquired from a non-metric digital camera which carried by an aircraft normally has lens and perspective distortions and could not be used directly without undergoing image rectification. Ground control points (GCPs) are important features used in non-parametric approach for aerial image rectification. Although the importance of GCPs in rectifying remote sensing images has been highlighted, not many recent studies research on the quality selection of GCP, effectiveness of GCPs’ distribution and sufficient quantity of GCPs. A simulation test is conducted using grid images to examine the effect of different distribution patterns of control points on distortions and geometric transformations. The rectification results are measured by using the total root mean square error (RMSE). It shows that lower order global transformation has limitation in rectifying images with complex distortions. It also demonstrates that centre distribution gives the lowest total RMSE and its total RMSE is extremely low. However, the distance analysis of control points which reflects the distortion rate before rectification shows that control points distributed at the centre of the image is actually much less distorted than control points that are placed at border and corner. Hence, uniform distribution provides better distribution of control points with the consideration of overall deformation rates at the entire image for rectification.