Autonomous Mobile Robot Navigation on Identifying Road Signs using ANN

Abstract

Road sign identification and classification is important for an autonomous navigating robotic vehicle. Many novel techniques were proposed in the past years to overcome these issues. The main objective of this research work is to implement it on a mobile robot and obtain a real time accuracy w.r.t Road sign classification. A Chinese road sign dataset is used for this purpose. Descriptors of this dataset is extracted from corner detectors like SIFT, SURF, ORB and fed into supervised learning techniques like SVM, L-R and ANN. Quality Metric Parameters like Recall, Precision and F1-measure, were used to determine the best method. LQR controller is used for the robot navigation. Road signs may be present in a clean or cluttered environment, to detect these signs in such unpredictable environments and feed it to classifier, Maximally Stable Extremal Regions (MSER) is used. On recognizing the road sign, mobile robot will navigate autonomously. On multiple experiments, the solution offered in this research work is very robust and accurate for real time applications.