This method combines several image features using Speedup Robust Features (SURF) by accumulating them into a union form of features inside each place cluster. In this method, the mobile robot captures omnidirectional visual information and converts them into topological place clusters. | Journal of Automation and Control Engineering, Vol. 1, No. 1, March 2013 Unsupervised and Online Place Recognition for Mobile Robot based on Local Features Description Tang . and G. Hamami Universiti Putra Malaysia, Malaysia Email: saihong@, B. Karasfi Universiti Putra Malaysia & Islamic Azad University, Qazvin Branch Email: karasfi@ D. Nakhaeinia University of Ottawa, Canada Email: dania@ Abstract—Place recognition approaches have been used for solving topological mapping and localization problems. These approaches are usually performed in supervised and offline mode. In this paper, a robust appearance-based unsupervised and online place recognition algorithm, which is inspired from online sequential clustering methods, is introduced. This method combines several image features using Speedup Robust Features (SURF) by accumulating them into a union form of features inside each place cluster. In this method, the mobile robot captures omnidirectional visual information and converts them into topological place clusters. Experimental results show the robustness, accuracy, and efficiency of the method as well as its ability to create topological place clusters for solving mapping and qualitative localization problems. The performance of the developed system is remarkable in term of recognition precision performance. basic problem in mobile robot navigation system. In the past, several researches have been established to build an accurate and complete metric or topological map of the environment based on the data gathered by the mobile robot [3], [5]-[8]. Mobile robots needs to move to perform their tasks, so having a robust navigation system is essential. "Where am I?" is a fundamental question in the navigation system. Answering to this question is a research interest, which is related to mapping and localization methods. The way of mapping and localized the robot inside a map depends on various .
