Active Collision Warning System for Waterway Transportation via Multi-Stream Image Processing and Deep Neural Network

Due to the fast-paced growth of transportation infrastructure and the expanding waterway industry, collisions between vessels and ships continue to occur frequently, especially in low-light conditions. To address this issue, this paper proposes an integrated system for object detection and object distance/speed estimation in waterway environments using dual-stream imaging information from two functional cameras which are switchable between RGB and infrared images for daytime and nighttime, respectively. Object detection task is performed by specifically leveraging the YOLOv8 architecture for fast and effective performance. By processing both RGB and infrared data streams, the system enhances object identification capabilities under varying lighting conditions, thereby enabling precise distance and speed estimation of recognized objects. The experimental results of a case study in a local area of Mekong Delta demonstrate the system's effectiveness in realistic scenarios, showcasing its potential for applications in collision warning systems for waterway transportation.