Analysis of deep learning in automatic target recognition: evolution and emerging trends

Automatic Target Recognition (ATR) stands at the forefront of defense, surveillance, and remote sensing, where efficient and accurate target identification under complex conditions is paramount. Recent breakthroughs in deep learning have propelled ATR research, expanding its capabilities in handling diverse data modalities, from optical images to radar signals. Despite this momentum, the sheer volume and breadth of scholarly work can obscure emerging trends and highlight a need for systematic assessment. This study addresses that gap by conducting a bibliometric analysis of publications from 2013 to 2024, drawing on data extracted from recognized scholarly databases. Two complementary tools—Biblioshiny and VOSviewer—were employed to elucidate both quantitative metrics (publication growth, citation impact, and source diversity) and conceptual interconnections (keyword clusters, cocitation networks). Findings reveal a rapidly growing body of research, with annual publication rates exceeding 50% and an average of over 16 citations per document. VOSviewer-generated clusters emphasize a convergence around deep learning architectures, radar imaging techniques, and advanced strategies such as adversarial networks and self-supervised learning. These insights reinforce the view that ATR has entered a dynamic new phase, characterized by robust cross-disciplinary collaborations and an expanding array of real-world applications. By mapping the scholarly landscape, this paper not only provides a navigational tool for researchers seeking to identify influential work and novel lines of inquiry, but also underscores the vital role bibliometric methods play in shaping and advancing future ATR endeavors.