Design of a detection and alert system for animal road crossings (elephants) in Electric Vehicles using Computer Vision.
Abstract
Detection and alert systems in electric vehicles inform the driver of impending collisions with other objects along the host vehicle path and similarly electric vehicles as powered by batteries for propulsion thus clean transportation as compared to internal combustion engines. Since Uganda’s road network traverses most game parks/reserves, tourism subjects the country's road users to risks associated with animal road crossings. In this project, we developed a low cost and efficient detection and alert system based on a YOLOv4 object detector and a camera for forward vehicle-animal (elephant) collision detection. We trained the object detector on an African elephant dataset and an elephant sign post open dataset. We then implemented a distance, speed and direction estimation algorithm in python based on the triangular similarity using a single reference image with known distances to determine the distances of the new objects detected. Our results for the object detector were based on the @mAP scores of 93.56% and 93.4% for the elephant dataset and elephant signpost dataset respectively evidencing good system performance and accuracy. We had a detection range of 20m from the camera giving accurate distance measurements for first moving frames and also demonstrated implementation of our detection and alert system with a single sensor and processor (Raspberry pi 4 module) making it relatively cheaper and compatible with electric vehicle designs in Uganda.