Efficient Implementation of GF-CNN for Advanced Traffic Sign Recognition in Transportation Systems

Traffic violations cause accidents nowadays. Traffic accidents sometimes result from disobedience of traffic signs owing to low sight, inadequate light, and lack of expertise. ADAS and ITS are the only solutions to prevent accidents. ADAS detects traffic signs and ITS recognizes images for various transport applications. ADAS and ITS can detect minor traffic signs and provide driving information to guarantee road user safety. A key function of ADAS is to inform drivers and traffic signs. One research used traffic sign detection to aid vehicle navigation and driving assistance. Overfitting in the current technique reduces traffic sign detection performance. For image analysis, this study employs Gabor filter, Grey Level Co-occurrence Matrix, and Local Optimal Oriented Pattern feature extraction approaches. Our multi-threshold Otsu approach and cuckoo search method aid in segmentation. The SVM classifier performs traffic sign detection. The hybrid feature extraction approach enhances traffic sign identification by exploring the form, color, and texture of the board. Research used an enhanced CNN to recognize traffic signs. Current hybrid feature extraction approach achieves 98.97% accuracy in classifying six traffic signs, however processing time is high for multi-operations. To categorize 10 traffic signs with 99.1% accuracy, an enhanced CNN is presented.