The use of a 3D sensor (Kinect) for robot motion compensation : The applicability in relation to medical applications

Abstract

The use of robotic systems for remote ultrasound diagnostics has emerged over the last years. This thesis looks into the possibility of integrating the Kinect sensor from Microsoft into a semi-autonomous robotic system for ultrasound diagnostics, with the intention to give the robotic system visual feedback to compensate for patient motion.

In the first part of this thesis, a series of tests have been performed to explore the Kinect's sensor capabilities, with focus on accuracy, precision and frequency response. The results show that the Kinect gives the best measurements at close distance and that the accuracy and precision decreases fast as the distance increases. At $1m$ distance the Kinect could measure the depth between two surfaces with an average accuracy of $96\%$ and standard deviation of $4\%$. When measuring a moving target there was an average delay of $0.04s$ between actual position and measured position.

Secondly the Kinect was interfaced into the robot control system, where depth measurements from the Kinect have been used to directly control the position of the robot tool. The system could successfully track and follow the motion of a moving surface in front of the Kinect. The system could easily follow motion similar to respiratory motion with a total system delay of less than $0.2s$.