Automatic Aggregation of Text and Sign on traffic panels using spatial extensions to BOVW

Traffic panel detection and recognition is use to support road maintenance and to help drivers. It is use to detect traffic panels and recognize the information present on street-level images. The images of traffic panel are taken by high resolution digital cameras or smartphones.it recognizes text and symbol accurately. To recognize text, system extracts local descriptors after applying green and white color segmentation. Then, images are classified using Naïve Bayes and represented as a “bag of visual words”. In images if a traffic panel has been detected then Text detection and recognition method is applied on it to automatically store the information contain on the panels. We propose the system which uses spatial extension to BOVW such as sliding window, branch and bound. To recognize text exactly, we compute the prior probabilities of all the words using unigram language model .The language model completely based on a dynamic dictionary. Various algorithm use which is based on SIFT descriptors to recognize single characters and also on HMMs to recognize whole words. Keyword: Bag of visual words (BOVW), HMMs, SIFT Descriptors.