Detection of Public Information Sign in Airport Terminal Based on Multi-scales Spatio-temporal Vision Information

A new method based on multi-scales spatio-temporal vision information is presented to detect public information signs for the airport terminal service robot (ATSR) localization. Firstly, public information signs are searched roughly in big space scale by using the connected components labeling algorithm. Secondly, the signs position and size in visual field are adjusted properly based on spatio-temporal relativity. Finally, the signs are detected according to the extracted SIFT (Scale Invariant Feature Transform) features. Experiments by this new method on ATSR show that public information signs are detected rapidly and precisely.