Abstract: Underwater acoustic target localization and recognition are core functions of sonar equipment and pivotal research directions in underwater acoustic signal processing. For a long time, oceanographic physical models have been central to target localization and recognition. Target localization technology has evolved with the goal of enhancing robustness, progressing from matched field processing and matched mode processing to mode-based signal processing, and further advancing to techniques utilizing ocean acoustic field characteristics. Target recognition technology, on the other hand, has focused on extracting features of underwater targets that exhibit strong separability, generalizability, and environmental adaptability. In recent years, artificial intelligence, particularly deep learning, have rapidly advanced empowered by big data, providing new technical approaches to underwater acoustic target localization and recognition. Deep learning effectively addresses environmental uncertainties and the challenges of weak target localization, overcoming the limitations of traditional human expertise and shows significant potential in tackling underwater target recognition under practical complex scenarios. Notably, localization and recognition technologies are inherently interrelated. Precise localization information can provide spatial constraints for target recognition, while reliable recognition results can, in turn, refine localization accuracy, creating a natural synergistic relationship. Therefore, future research should focus on integrating localization and recognition technologies to mutually enhance each other, while promoting the complementary and collaborative development of oceanographic physical models and deep learning models. This will ultimately lead to a comprehensive improvement in the performance of sonar systems under practical application conditions.