In today’s world, majority of the accidents are occurring because of not obeying the traffic rules and signs. Majority of the cases is due to not following the traffic signs. Because of the low light or not having adequate knowledge of the signs. A solution for this problem is to detect the sign and its usage automatically. It can be achieved with the help of image processing techniques. Many researchers have proposed different algorithms for traffic sign detection. Maximum, it can classify six traffic signs. Hybrid feature extraction deep neural network is one of the existing techniques and it can able to classify six traffic signs with high accuracy of 98.97%. But, it requires high processing time to perform multiple operations. Hence, in this an augmented convolutional neural network is proposed to classify ten traffic signs. The proposed method uses Gabor filter and convolution neural network to extract features. Then, extracted features are trained and tested using convolution neural network. Due to the single stage processing, the proposed can reduce its computation time for feature extraction and able to classify ten classes of German dataset. The proposed method will be implementing using MATLAB in windows 10 environment. Its performance will be evaluating using accuracy against the existing techniques.

Traffic sign detection, Gabor, Augmentation, CNN, Ten classes.