A Survey of Multimodal Data Fusion: Applications and Adverse Conditions

Multimodal data fusion is a research field that combines information from various sources, each with its own modalities. One of the prominent challenges in data fusion is addressing adverse conditions that arise when dealing with real-world implementations. This article presents a review of data fusion, conducting a comparative analysis through a literature search across various application domains that tackle the challenges of data heterogeneity and adverse conditions. The focus of this paper is centered on establishing a robust foundation to enable the development of future research in this field. This review reveals that almost half of the analyzed documents describe adverse conditions, but only a minority conduct analyses on how their techniques address noise or benefit from considering these conditions. The number of modalities used in the research is generally low, with most studies employing static data in 1D dimensions. Finally, it is crucial for researchers to continue working on an interdisciplinary vocabulary and consider applications in real-world environments with adverse conditions to advance in this field. It is also important to explore applications with higher-dimensional data and more modalities, which could provide valuable insights for addressing specific challenges in data fusion.