This university report addresses the development of a vision system for the Robot Assisted Ultrasound(RAU) product to pass clinical trials. The research objectives focus on enhancing the sonographer's ability to verify correct patient positioning, improving the manipulator's behavior for smoother movements, and detecting patient movements after the reconstruction process.

For the first objective, a 3D point cloud of the patient was created using RealSense D415 cameras. A bounding box was defined to guide the sonographer in positioning the patient. However, the interface is guiding the Sonographer with 2D images.

The second objective aimed to improve the manipulator's smoothness during movements. Initially, it was thought that reconstruction quality was the cause of erratic behavior. However, it was determined that adjusting the function responsible for directing the robot's movements resolved the issue. Future software development may involve parametrically describing the mesh to enhance computational efficiency and accuracy.

The third objective involved detecting patient movements after reconstruction. An Iterative Closest Point (ICP) algorithm was implemented to estimate the transformation between initial and current point clouds. While the method can detect movements, further improvement is needed for accuracy by testing different hyperparameters. The appropriate response of the system upon detecting movement is yet to be determined and requires future consideration.

In conclusion, the research objectives addressed important aspects of the vision system. The sonographer guidance and manipulator movement improvements will be utilized, while further investigation is needed for mesh description and patient movement detection.