Wednesday, 21 Feb 2018
You are here:

Subscribe to our Newsletter

How many eyes has a typical person? (ex: 1)

Twitter Feed

EMail Print


Karthik Vaiapury (QMUL) visiting Prof. Atta Badii (UoR)
30.06.2013 - 14.07.2013
Internal Visits


Karthik Vaiapury visited University of Reading Research lab as part of the collaborative work on the calibration mini-project. During this visit, Karthik worked with researchers from UoR on integration of new approach using walking people to calibrate surveillance cameras. Traditional camera calibration methods (e.g. Tsai’s method or Zhang’s plane-based method) often require the scene is well organized so that some specifically calibration grid can be seen and can be easily manipulated. In reality calibration grids are not always present. This makes camera self-calibration algorithms particularly attractive. The proposed method estimates the camera projection matrix instead of estimation each of the physically camera parameters (e.g. focal length, rotation, translation) separately. This way, the relationships between the geometric property of a walking human and the camera parameters are made more explicitly, and more clear to see. The proposed approach also demonstrates that there is no need to pre-detect three orthogonal vanishing points. Rather, they can be detected automatically by exploiting the special assumption about the shape of the image pixels. Therefore, the proposed method can reduce one required user input to pre-estimate the three vanishing points, which is two orthogonal lines marked on the ground as two axes in 3D scene.