Minimum hardware SfM/SLAM for sparse data point mapping of retail stores

Sammanfattning: This report aims to compare methods that can be used to efficiently and accurately map digital price labels position in 3D space using a mobile robotic platform equipped with camera vision in a retail environment. The robotic platform built during the project uses four wide angle cameras to cover the full circular view in the horizontal plane. Installation of external systems for e.g. navigation can be very costly and time demanding. The accuracy, repeatability and time consumption have been evaluated for a set of methods and combinations thereof with and without external hardware. The result of the tests shows a proof-of-concept with good results but also emphasizes the need for a robust system. The mapping is done with high accuracy even when the external hardware for navigation is completely removed. As a part of the project a novel line following algorithm has been developed which shows results far better than any published results found and yet capable of running in real time utilizing limited hardware.