Done by Niloufar Mallam Hassam, BSc (Hons) Computer Science Yr 3 (Completed August 2021)
Traffic sign recognition (involving detection and classification) is considered fundamental for certain applications such as semi-autonomous / autonomous self-driving cars or driving alert systems. The system applies image processing techniques on a traffic sign image extracted from the images of road scene. Then, it should classify that sign into one of the correct categories & identify its meaning.
Traffic signs are simply detected by analyzing color information, notably red and blue, contained on the images. The detected signs are then classified according to their shape characteristics, as triangular, squared and circular shapes. Combining color and shape information, traffic signs are classified into one of the following main classes: danger, information, obligation or prohibition etc. Recognition process will identify the specific sign within its category. In order to make it happen, a machine learning technique can be applied, being trained with 1000s images beforehand. The algorithm analyzes the recurring patterns in the presented training images and learns to categorize new images.
The goal of this project is to classify all traffic signs images in Mauritius. We should train the model so it can decode traffic signs from natural images (road scenes) using a Mauritian Traffic Sign Dataset (to be constructed). This data should be first preprocessed in order to maximize the model performance. After selecting a suitable model architecture, fine tuning and training, the model will be tested on new images of traffic signs. Because we deal with images classification, a Convolutional Neural Network (machine learning) is a common choice for this type of problems. Detecting and classification of Traffic sign with high accuracy is the expected outcome of this project.