Computational Intelligence for Analysis of Traffic Data

This article presents a system developed for the collection and analysis of traffic data obtained from traffic camera videos using computational intelligence. The proposed system is developed using the modern object detection library Detectron2. A pipeline-type architecture is used for frame processing, where each step is an independent, configurable functional module, loosely coupled to the others. The validation of the proposed system is performed on real scenarios in Montevideo, Uruguay, under different conditions (daylight, nightlight, and different video qualities). Results demonstrate the effectiveness of the system in the considered scenarios.