Low Cost AI Based Detection of Floating Objects Using Stereo Cameras and Radar

Abstract In this paper, we present a low-cost situational awareness system designed to assist visually impaired sailor in navigating constrained maritime environments. The system integrates stereo vision, consisting of two black-and-white cameras and a central color camera, along with radar, to detect and classify obstacles at sea. This setup is a cost-effective alternative to traditional Light Detection and Ranging (LiDAR) systems, a You Only Look Once (YOLO) v7-based model, for real-time obstacle classification. Designed specifically for operation under fair weather conditions, the system captures and processes visual data to identify various types of floating objects, enhancing navigation safety. The stereo vision provides obstacle classification and depth information, while the radar complements the detection range with velocity, heading and azimuth. After calibration and synchronization, a projection based method fuses data between camera and radar. The system’s affordability and versatility make it a promising solution for the future development of accessible maritime navigation technologies. Keywords Situational Awareness; Stereo Vision; FMCW radar; marine object classification; visually impaired navigation