Image guided surgery(IGS) is useful in surgeries which are hard to achieve clear view such as minimally invasive surgery. Optical tracker is generally used in image guided surgeries for its high accuracy. Augmented reality is applied to display CT(Computer Tomography) or MRI(Magnetic Resonance Imaging) images matched with the patient in real time using optical tracker. This supports the surgeons to easily imagine important parts of the patient that are not in view. However, optical tracker has a problem when occlusion occurs during surgery. When the tracking target gets occlude by something or someone during surgery, all functions to support the surgeon stops because the optical tracker stops tracking. Therefore when such problem occurs, the surgeon has to transfer to traditional surgery with out the support of image guided surgery. Also, optical tracker usually has low measurement frequency due to complicated image processing. If the measurement frequency is lower than 25Hz, the display of augmented reality has vibration and therefore the performance of navigation is low.
The purpose of this study is to overcome the occlusion problem and low measurement frequency of optical tracker using sensor fusion of optical tracker and IMU(Inertial Measurement Unit) based on Kalman Filter. In this thesis, the calibration method and attitude estimation of the proposed system are introduced. IMU is independent of occlusion and has high measurement frequency, but in the other hand it has drift error due to DC bias. Using sensor fusion of IMU and optical tracker, the occlusion and low measurement frequency problem can be solved.
Commercial optical tracker and IMU are used to make the hardware, and experiments are done to verify the performance of the proposed system. The experiments show that the divergence of measurement during occlusion is lower than 0.0237deg/sec. Even if occlusion occurs for 5 seconds, the error of measurement will be lower than 0.119deg which is negligible. Therefore, the proposed system can be used if the occlusion does not maintain over 5 seconds.