Radar image-based object detection and tracking for autonomous surface vehicles

This study presents a modular, lightweight framework for radar image-based object detection and tracking specifically designed for autonomous surface vehicles (ASVs). The detection module employs HSV (Hue, Saturation, Value) color-space segmentation to identify navigational targets, particularly yellow-encoded objects, and the own ship indicator, typically rendered in white on commercial marine radar screens. To enhance stability under varying sea conditions, the system applies Gaussian smoothing, morphological operations, and area-based filtering during preprocessing. A dynamic multi-target tracking module assigns persistent object identifiers (ID) using Euclidean distance-based association scheme enhanced with velocity and direction-aware ID recycling. This approach mitigates identity fragmentation caused by occlusions or abrupt maneuvers. Furthermore, the framework estimates each target’s relative velocity, enabling inference of behaviors such as approaching or retreating from the own ship. The proposed detection method, evaluated on 7,200 radar frames, achieved a 98.7% true positive rate, improving by 4.4% over the area-based baseline and reducing the false positive rate by 59%. This demonstrates robust target discrimination in noisy and cluttered conditions. The proposed tracking system further reduced ID switches by 75% and decreased total generated identity count by 58% compared to positional-only baseline, indicating significant gains in tracking continuity and identity stability. The average processing time of 240 ± 17 milliseconds per frame confirms the framework’s suitability for real-time embedded deployment in dynamic maritime environments.