Automated Traffic Sign Detection Using HSV Filtering and ROI based SVM Classification

Traffic sign detection and classification play key roles in pre-signal processing and feature extraction of autonomous driving and Advanced driver assistance systems (ADAS). A methodology of detecting and classifying traffic sign using HSV color space filtering, ROI detection and Support Vector Machine (SVM) classification is presented in this paper. To improve visibility of traffic signs, input images are preprocessed and denoised in the proposed approach. An HSV filter is then applied to remove color attributes that do not belong to highway traffic signs, and the result increases segmentation accuracy substantially. The detected ROIs are then bounded boxes, focusing the classification to most probable traffic sign regions. High dimensional feature spaces and diverse sign classes are handled with minimal errors with an SVM classifier trained on a labelled traffic sign dataset. The method is shown to have high accuracy in real world scenarios, and as a consequence is suitable for use in real time application such as autonomous driving and driver assistance. The system can process images under different conditions and also successfully deal with rare challenges such as low light conditions and generalizes the framework to multiple classes of traffic signs. This work for robust and scalable traffic sign recognition systems which are essential for the advancement of road safety and automated navigation.