YOLO Multi-Camera Object Detection and Distance Estimation
Bojan Štrbac, Marko Gostović, Željko Lukač, Dragan Samardžija
Abstract
Due to challenging requirements in the competitive market and significant software expansion in automotive industry, there is a need and opportunity to develop and implement new algorithms and solutions, which can further enhance performances, features and quality of this fast growing and constantly changing industry. The aim of this paper is to present the possibility of using cameras instead of LIDARs for distance estimation. The proposed solution is based on the YOLO deep neural network and principles of stereoscopy. This solution uses two slightly moved cameras which obtain two pictures, which goes through algorithm for stereoscopy-based measurement. And estimate distance to detected objects.