Evaluation of 3D camera technology for patient positioning within the field of radiation therapy

Abstract

The thesis describes shortly the Time-of-Flight technology and also the experiments, algorithms and results for the registration of point clouds acquired by a Time-of- Flight camera. The aim is to explore possible applications of the ToF camera for patient positioning in the field of radiotherapy. A distance function between two point clouds has been defined for the registration of point clouds in different frames. Afterwards the distance of the reference frame with a linear transformed point cloud is minimized. For this purpose the UOBYQA (unconstrained optimization by quadratic approximation) algorithm has been used. The following results could be obtained: a single pixel of the camera itself had a standard deviation of 0.5mm; the translation of a spherical object could be determined with an average error of 0.5mm. The algorithm for the registration of arbitrary surfaces with six degrees of freedom are not completely finished and will be improved in further work. The cameras are sensitive to temperature fluctuations, which has been tried to be avoided by using two cameras at the same time. For this purpose the cameras have been calibrated and their coordinate system transformed from one to the other. Finally a motion surveillance application is presented, which creates a real time video stream with a visualization of the distance from the actual frame to the reference frame.