Geometric quality of UAV's consumer grade camera: Lessons learnt.

UAVs (Unmanned Aerial Vehicles) gain wider acceptance nowadays for mapping related activities and inspections of engineering structures. This flying platform is usually equipped with consumer grades digital cameras. Such off-the shelf cameras have long been known to suffer unstable lens elements and electronics. Regardless of camera types being mounted on the UAV board, these instabilities are major impediment to achieve a high accuracy measurement in mapping project. For these cameras, a position of the lens and its elements are located and aligned arbitrarily in a predefined location relative to the CCD (Charged Coupled Device) sensor plane. In facts, deviations of the projected lens axis onto the CCD plane from the center of the sensor can be considerable amount of the extent of the sensor. These departures can be exacerbated by the UAV's engine vibrations. Therefore, this paper evaluates geometric perturbations of the camera's lens distortion model during the UAV's flight missions by assessing bundle adjustment based camera calibration by varying physical calibration parameters. This study use a method so called self-calibrating bundle adjustment to model four groups' source of errors, namely radial distortion, tangential and affinity distortion, as well as principal point and principal distance stability. Analyses is performed in three sequences: prior to, during, and post flight missions. It can be revealed that the camera focal length and location of the lens axis onto the CCD plane severely perturbed during the flights and peaked to about 1mm of discrepancies. In contrast, lens distortions models are relatively stable during the mission. [ABSTRACT FROM AUTHOR]