Mounting Parameters Calibration of GPS/INS-Assisted Photogrammetric Systems

There has been a significant advancement in the remote sensing and mapping technology in the last few decades. Due to its decreasing cost and increasing resolution, digital cameras are rapidly replacing the need for the conventional large format analogue cameras. Also, with the advent of GPS/INS, direct sensor orientation became possible, providing the means for an almost control-free mapping environment. Although several advantages are offered by the direct sensor orientation, precaution should be taken when dealing with multi-sensor systems. In GPS/INS-assisted photogrammetric systems, besides the camera calibration, the geometric relationship between the sensors (mounting parameters) must be known as well. More specifically, the lever-arm offset between the sensors, as well as the misalignment (boresight angles) between the IMU body frame and the photogrammetric camera should be determined. The offsets are usually measured using traditional surveying techniques, while approximate values for the boresight angles are known from the mechanical alignment. Since these initial mounting parameters might be biased, they should be refined through an in-flight calibration procedure. Also, some of the camera IOP, such as the principal point coordinates and the principal distance, might experience changes under operational conditions and need to be refined in the in-flight calibration as well. The objective of this paper is to introduce a methodology for the in-flight photogrammetric system calibration, as it relates to control and flight configuration requirements. The paper starts with a brief discussion of the concept of GPS/INS-assisted photogrammetric triangulation procedure. Then, the concept of the proposed rigorous analysis to devise the optimum flight and control configuration for reliable estimation of the system parameters is presented. The validity of the presented analysis is demonstrated through experimental results using simulated and real datasets.