University of Nottingham
Joe completed his PhD as part of the Manufacturing Metrology Team at the University of Nottingham last year. His PhD focussed on leveraging machine learning for the automisation and optimisation of optical coordinate measurements. He is now a research engineer in the field of autonomous robotics.
Object pose estimation for stereo imaging using raycasting
A method for estimating and refining the pose of an object within a measurement volume is presented for systems utilising calibrated stereo vision cameras. An initial pose is gained by masking the object from the background generating a pair of binary masks. A raycasting approach is used to estimate the object’s pose from these masks. This pose is then refined in an iterative minimisation procedure which generates rendered binary masks from the current pose prediction using the object’s computer aided design (CAD) data. A loss function is defined between the real binary masks and the predicted binary masks which is minimised by the optimisation algorithm. The output is the six degrees-of-freedom pose of the object relative to the imaging system