Calibrate kinect to use in computer vision, simplified and precise

Abstract

Visual sensors, active or passive, play an important role in computer vision and in visual sensors, calibration is of utmost importance. Kinect as a new developed sensor for use as a Natural User Interface is being utilized in different fields especially CV. This integrated system beside other sensors, contains two visual sensors of active and passive that demands a process of calibration. Among different methods of calibration, image-based calibration for data-fusion purposes, has lowest computational cost and can be quite simple and precise. In this study, 2 different methods, consisting of a physical interior distortion model and an eight parameters registration equation have been proposed. Besides computed parameters and their precision, a table of distortion values is introduced that can be used in registration level. Finally to evaluate chosen proposed method, a simple registration of processed data is utilized and results are discussed.