15. Analysis and evaluation of the geometric accuracy of orthorectification of Remote Sensing imagery without Ground Control Points based on Rigorous and Ground Control Point Models

Abstract

In the processing of high and very-high-resolution optical remote sensing imagery, geometric correction (orthorectification) is a critical step to eliminate distortions caused by the sensor, terrain, and imaging geometry, ensuring that the image is accurately aligned with the map coordinate system. Traditionally, this process requires the use of Ground Control Points (GCPs) to achieve high positional accuracy. However, in many situations, especially those requiring automated, large-volume data processing and near real-time response in emergency scenarios, orthorectification without GCPs methods based on intrinsic sensor models has become an important approach in modern processing systems. Two primary categories of models are commonly used: (i) rigorous sensor models, which rely on satellite orbit parameters and sensor calibration data; and (ii) the Rational Polynomial Coefficient (RPC) model, which is provided with most commercial satellite imagery. The objective of this paper is to analyze and evaluate the geometric accuracy of high- and ultra-high-resolution remote sensing images orthorectified without GCPs using both rigorous physical models and pure RPC-based approaches. The study results show that the orthorectification without GCPs using a rigorous model achieves point positional errors of approximately 3–5 meters in flat terrain and 15 - 20 meters in mountainous areas. In contrast, using the RPC model results in point positional errors of approximately 37 - 39 meters.