our project opens up a world of possibilities for solving day to day problems for the visually impaired and sighted alike. The proposed system overcomes four major issues addressed by the prototype for visually impaired persons. Firstly, Identification of colors of the object present around the visually impaired people. Color variability is done through grayscale based vision algorithm that is difficult to work. So, in this paper we propose a colorvision algorithm that consists of two methods namely 1.) Anartificial color contrast as a pre-filter that aims at highlighting the target while suppressing its surroundings 2.) Statistically based fast bounded box (SFBB),that utilizes the component analysis technique to characterize the target features in color space which is obtained from a set of training data through which the color classification can be performed accurately and efficiently. Secondly, the currency along with the fake note detection is done by a component based recognition system which uses the speed up robust features (SURF) algorithm. The input image is converted to grey scale, the grey image is extracted and compared with an algorithm and then the voice output is produced through the earphone to visually impaired users. Thirdly, the optical character recognition algorithm is usedto convert the given text into voice output with the help of the e-speak engine. Finally, the detection of the obstacle in front of visually impaired users is also been achieved by using Ada-Boost model to differentiate between human andvehicles and given as voice output
Key words: Statistically based fast bounded box(SFBB), Speed up robust features(SURF), Optical character recognition(OCR).
|
