Nighttime Ship Detection Using VIIRS DNB Data: An AutoML Approach

Nighttime ship detection plays a vital role in understanding oceanic patterns and human activities in marine environments. As an observational approach in ocean science, it enables researchers to monitor vessel distribution patterns, analyze maritime traffic flows, and collect valuable data about human interactions with marine ecosystems. While the VIIRS Day-Night Band (DNB) sensor enables nighttime vessel detection from space, conventional detection methods primarily rely on threshold-based techniques, which show limitations in handling complex environmental factors such as cloud coverage and varying atmospheric conditions. To overcome these challenges, this study presents an automated ship detection approach that combines VIIRS DNB imagery with AutoML techniques. Our AutoML framework automatically optimizes model parameters and features to adapt to various environmental conditions, providing more robust detection capabilities compared to traditional threshold-based methods. The methodology incorporates AIS data for model training and validation to enhance detection accuracy. Our experimental results demonstrate improved detection performance across diverse maritime environments and weather conditions, effectively addressing the limitations of conventional threshold-based approaches. This research contributes to advancing pattern recognition in oceanic observations by providing an automated approach for identifying vessel activities in nighttime satellite imagery.