Artificial intelligence and remote sensing for marine surface pollution detection: A systematic review

This is the authors' abstract. We don't add key points for this paper.

Marine surface pollution threatens ecosystems, fisheries, coastal economies, and maritime safety. Remote sensing offers wide-area monitoring, while artificial intelligence (AI) supports automatic interpretation of large optical, synthetic aperture radar (SAR), hyperspectral, multispectral, satellite, and unmanned aerial vehicle data streams. This paper presents a PRISMA-oriented systematic review of AI-based remote-sensing approaches for detecting marine surface pollution. A structured Scopus search was executed on April 4, 2026, and 102 eligible studies were retained after screening. The synthesis shows that the retrieved corpus is dominated by oil-spill detection, SAR imagery, and convolutional deep-learning models; within this corpus, floating marine litter, non-oil pollutants, cross-regional validation, public benchmark datasets, and operational deployment constraints receive less attention. The review identifies persistent challenges related to look-alike phenomena, small or private datasets, heterogeneous metrics, and limited evaluation of real-time performance. It also highlights future directions based on multi-sensor fusion, transformer and self-supervised models, lightweight edge inference, standardized datasets, and closer integration with environmental monitoring agencies.