Robust Classification of Traffic Signs using MRMR Feature Reduction Technique

With the advent of rising global population and vehicular traffic on the road, driver safety has become a great concern for mankind. There has been an increased interest of research in area of Traffic sign recognition (TSR) which happens to be one of the critical aspects of road safety and concept of autonomous cars. The technological advancements in artificial intelligence and related IOT technologies have provided much needed momentum to research. Cars have been fitted with advanced systems which can scan the traffic signs on the road and then can either recommend the best action to the driver or can be programmed to take the action by system itself. There have been excellent efforts put by researchers in past but still today, the accuracy prediction of traffic sign detection remains a challenge for most car producers and drivers, especially under tough weather situations. In this research paper, authors have proposed an innovative approach for traffic sign image recognition by using dimensionality reduction technique using maximum relevance and minimum redundancy (MRMR) algorithm coupled with ANN for classification. The results achieved have been compared with related work done in past and have been found to be encouraging.