Single-photon sensors combined with eye-safe, high repetition rate, pulsed laser sources enable a solution for many applications when conventional imaging with a multi-pixel sensor is not possible or otherwise expensive such as three-dimensional ranging. The temporal response of single-photon LIDAR results in a surface-to-surface resolution at the millimetre scale. However the requirement of high resolution reconstructions in real time can result in very sparse low photon counts. Here we investigate fusing data from a single photon sensor with other low cost image (RGB/infra-red) information in a deep learning framework to predict depth at high resolution in milliseconds.

This presentation is part of Minisymposium “MS44 - 3D Image Depth/Texture/Reflectivity Tracking, Modelling and Reconstruction”
organized by: Catherine Higham (University of Glasgow) , Roderick Murray-Smith (University of Glasgow) .

Authors:

Catherine Higham (University of Glasgow)

Roderick Murray-Smith (University of Glasgow)

Keywords: