Web Application Development for Product Label Reading using Machine Learning Techniques

This paper analyse and develop the web application using camera based assistive text reading of product labels from objects to help visually challenged people. Camera serves as main source of input. To detect the objects, the user will move the object in front of camera and this moving object will be detected by Background Subtraction (BGS) Method. Text region will be automatically localized as Region of Interest (ROI). Text is extracted from ROI by combining both rule based and learning based method. A novel rule based text localization algorithm is employed by identifying geometric features like pixel value, color intensity, character size etc. and also features like Gradient magnitude, gradient width and stroke width are learned using SVM classifier and a model is built to differentiate text and non-text region.