A modified technique for 3D camera calibration

In this paper a modified technique for 3D camera calibration is applied on a stereo-camera, in order to get accurate depth information of a desired object on the scene. The technique requires observing a scene or a planar pattern with known geometry at some different orientations. A curvature based corner detection technique is applied to detect both fine and coarse features at low computational cost to obtain the intrinsic and extrinsic parameters of cameras and achieve the translation between the world and image coordinates. SURF matching approach is aimed for optimizing the translation rule from the image to spatial coordinates by using a set of rotation-invariant interest points on stereo images. The result together with accuracy and calculation speed is reported.