Sign Understanding in Support of Autonomous Navigation (SUSAN)

"Mobile robots currently do not have the ability to detect and read arbitrary signs. Existing work on road sign recognition is template based, limiting the detectable signs to a specific set. This is a major hindrance to mobile robot usability, since theycannot be tasked using directions that are intuitive to humans. It also limits their ability to report their position relative to intuitive landmarks. We propose to develop a system for Sign Understanding in Support of Autonomous Navigation (SUSAN). PhaseI will focus on the bottom-up computer vision problem of reliably detecting arbitrary signs; Phase II will focus on understanding these signs. In Phase I, we will design and implement a sign discovery algorithm using sign detection based on colorsaturation and text detection based on texture discrimination. False alarm rejection will be based on shape analysis and evidence accumulation from tracking potential signs across multiple video frames. We will also integrate commercial optical characterrecognition software into SUSAN and evaluate the system using video captured for this project. Finally, we will design a sign understanding algorithm that classifies signs according to potential utility to the robot's navigation system and determinessemantic content of useful signs. The completed SUSAN system will have significant market potential for developers, both military and civilian, building mobile robots that need autonomous or semi-au