Evaluation of the Lens Distortion Errors in 2-D Image to 3-D World Coordinate Projections

Mapping of information from a camera coordinate system into the real world coordinate system is an important task for many applications. Information in camera coordinates can be the position and the dimension (width and height) of a detected object. Objects of interest used in this paper are detected pedestrians from a surveillance system for a safety application task. To combine and share information between different users, e.g. a surveillance system and driver assistance systems (DASs), a very accurate mapping is required between the surveillance system and the world coordinate system. The world coordinate system is a common place for data fusion between different camera systems. This paper is focused on the error of the lens distortion of the surveillance system and the influence to the mapping into the real world coordinate system. The analysis will include real measurements of detected pedestrians and the relation of distorted and undistorted mapping of information from the image.