A review of aquaculture: From single modality analysis to multimodality fusion

Abstract

Efficient management and accurate monitoring are crucial for the sustainable development of the aquaculture industry. Traditionally, monitoring methods have relied on single-modality approaches (e.g., physical sensors, vision, and audio). However, these methods are limited by environmental interference and inability to comprehensively capture the complex characteristics of aquatic organisms, leading to data bias, low identification accuracy, and poor model portability across different settings. In contrast, multimodal fusion technologies have emerged as a promising solution for intelligent aquaculture due to their strong environmental adaptability, information complementarity, and high generalization ability. Despite this potential, there is a lack of comprehensive literature reviewing the transition from single-modal to multimodal systems in aquaculture. This paper addresses this gap by presenting a systematic review of both single-modal and multimodal fusion technologies in aquaculture over the past two decades. We analyze the strengths and limitations of each approach, focusing on four key areas: water quality monitoring, feeding behavior analysis, disease prediction, and biomass estimation. Through this comprehensive analysis, we provide theoretical and practical insights into the application of multimodal fusion technology in aquaculture, highlighting its potential to enhance efficiency and sustainability while overcoming current limitations.