Adaptive Structured Light Imaging with Exposure and Focus Fusion for 3D Reconstruction and Autonomous Robotic Exploration.

摘要

This thesis research proposes several significant improvements to a previously developed structured light range sensor in order to enhance its robustness. By applying modern methods to classical structured light techniques, the improved sensor is capable of adapting to many different environments and generating 3D surface reconstructions of more general and unconstrained scenes. This is achieved by combining several algorithms in parallel, which permits the sensor to adapt in a reliable and autonomous manner to multiple colours, reflective characteristics and depths of field of the scene. The main motivation of this research is to ultimately mount the range sensor on a mobile platform, and perform autonomous navigation, mapping, modelling and exploration of complex environments. This thesis presents enhancements to the processing stage of the sensor, a complete overhaul of the acquisition stage, as well as a comprehensive set of results that demonstrate how the sensor adapts to the environment. Also, a complete prototype for a robotic mobile exploration system is presented and tested, validating the methods and techniques presented in this work.