TEXT DETECTION FROM TRAFFIC REGULATORY SIGN BOARDS

Natural images contain useful information in the form of text. This can be easily identified by humans and the importance of this text can be decided by them through their experience. Text detection is useful in applications like identifying license plate numbers, reading the text in the sign board and recognition of hand written text etc. In this paper, first text detection is done on the images of the traffic signboard and suitable edge detection algorithm is explored for finding out the text from image. MSER (Maximal Stable Extremal Region) and canny edge detection are used for edge detection. The merits and demerits are analyzed. Secondly, comparisons of text classification methods are done. They are Text classification - geometric property and stroke width, Random Forest. Experimental results obtained shows that the proposed approach with canny edge detection and classification of character using machine learning algorithm efficiently detects the text in images.