Text information extraction and analysis for autonomous vehicle

Scene analysis is an integral part of autonomous vehicle. Information such as traffic sign boards information, Toll information, Vehicle Number Plate information etc., from the scene is helpful in analyzing the current location. Vision based techniques are critical for a given scene analysis and happens to be an important area of research, today. This paper present localizing and recognizing text data from a given image. The focus is on traffic scene images and the proposed system works well across different colored images, images with perspective errors (less than 90°) and also for different fonts of text. The emphasis here is on extracting relevant information from the given scene image so as to enhance the performance of the autonomous vehicle. This method has been tested on standard ICDAR 2015 word recognition dataset and has been demonstrated with encouraging results.