Indian Traffic Sign Detection and Classification Using Neural Networks

This paper presents an automatic Indian Road Traffic Sign Detection and Classification system based on Multiple Neural Networks. Road safety being an indispensable factor for driving and smooth traffic movement has been studied since long time and traffic signs provide the driver with necessary information and warnings. Being very different in color and shape when compared to natural environment these signs can be detected easily by humans.

The system proposed in this paper uses four stages.1)Image procurement and preprocessing images captured by the camera which might be blurred or corrupted due to environmental disturbances, we try to remove these by deblurring and enhancing quality of the image.2)Color segmentation based on RGB ,YCbCr and NTSC color space to detect green blue and red color respectively then morphological operations to remove any unwanted noises that might be detected.3) Blob Detection using Binarization and Otsu Thresholding to obtain the region of interest and shape classification.4) Classification using Multiple Neural Networks to decide the type of sign. From the results we can conclude that when the neural network is trained over a standard database, the recognition of region of interest has high accuracy and the proposed methodology works with real time images invariant to rotation, illumination and in many situations even with partially distorted and occluded images.

The proposed method is validated on a standard data set of Indian Traffic Signs. As far as our knowledge is concerned no work has been done on the total data set considered by us although few works have been done on restricted data set of the one we considered and the data sets for other countries were also considered.