Real-Time Traffic Sign Recognition using Image Processing and CNN

The processing recognition of driving signs in autonomous vehicles, known as Traffic Sign Recognition (TSR),has considerably advanced in recent years. This work utilizes deep learning techniques, specifically aConvolutional Neural Networks (CNN) classifier, for robust recognition of traffic signs. The dataset used for thisanalysis consists of thousands of images featuring different categories of traffic signs including speed limit,warning, and do not enter signs. The model is subjected to multiple preprocessing steps such as image resizing, normalization, and augmentation to improve generalization capabilities and reduce overfitting during training. The accuracy achieved after training on the CNN model was 98.02% on the test set, showing effectiveness in recognizing traffic signs with respect to different lighting conditions, occlusion, and background noise. Besides this, the model was put through a comparative analysis with other machine learning approaches such as Support Vector Machines (SVM), Random Forest, and Gradient boosting; confirming that CNN is a better solution due to its image spatial and hierarchical feature extraction capabilities. The project has also implemented videobased traffic sign detection and recognition within OpenCV, increasing the practicality of the model. This research presents a robust, scalable, and efficient solution for traffic sign recognition, contributing significantly to the advancement of autonomous vehicle technologies.