Dissertation: Increasing Road Safety Using Image Processing and Artificial Intelligence

Vehicles on the Maltese roads increase every year, and consequently accidents are more likely to occur. To mitigate this problem, modern technology such as computer vision can help to increase the safety on our roads.

After having carried out a study on different road safety measures already available, a red light running solution has been proposed in this dissertation through the use of Red Light Cameras (RLC) that so far not yet been implemented in Malta. Many countries that installed RLC reported an immense decrease in red light running violations.

Additionally, the traditional implementations of RLC make use of either induction loops, rubber hoses filled with air or radars. A computer vision approach does not require these types of sensors to create a functional RLC system. The proposed solution has numerous advantages over these traditional methods, such as: lower costs; faster installation period; ability to monitor both online and offline; uninterrupted surveillance as well as no road modification requirements.

An Automatic Number Plate Recognition (ANRP) has been implemented using Python and the OpenCV library to detect any red light running violations. The ANRP system has been augmented through Optical Character Recognition (OCR) software that has been implemented. The OCR implementation is based upon the k-Nearest Neighbor algorithm to perform number plate recognition.