The design and initial calibration of an optical tracking system using the Microsoft Kinect

Robust motion correction in medical imaging requires accurate and reliable motion tracking. Current systems use devices such as the Polaris Vicra position sensor to monitor the position and orientation of a tracking tool which is fixed to the subject. Although in principle these methods offer high positional accuracy this is lost if the tool slips. Markerless motion tracking aims to track the object directly without the use of markers or a tracking tool. To date these methods have either been unsuccessful or too expensive to have been widely implemented. The Microsoft Kinect is a low cost RGB+D (colour plus depth) video camera. We have used the Kinect to perform motion tracking of a head phantom using a CT of the head as a high resolution template. We present initial results that show the Kinect can track rigid body motion to within <;2 mm of that measured by a Polaris system. We use the PointCloudLibrary open project [1] algorithm implementations to register the CT template to the Kinect frames.