Mapping the environment at range: implications for camera calibration

Abstract

In the last period Advanced Driving Assistance System are growing exponential and offering new features to drivers in the direction of achieving Autonomous Driving. Active safety is achieved with multiple sensors but still one of the most important ones is the camera sensors within the architecture. Creation of depth maps which describes the surrounding environment is a common practice in the field. In order to create an accurate map the correctness of the camera calibration is an important factor. In this paper we analyze the impact of camera calibration on the environment mapping reconstruction, by exploring the effect on the reconstruction error from short range to long range. Focusing on the latter, we show that the mean reprojection error may not be a sufficient indicator of the calibration accuracy and we propose better measurements and the associated calibration methods.