飞行器光学成像与测量技术研究室知识库

Fast and robust small target detection is one of the key technologies in the infrared (IR) search and tracking systems. With the development of deep learning, there are many data-driven IR small target segmentation algorithms, but they have not been extensively surveyed; we believe our proposed survey is the first to systematically survey them. Focusing on IR small target segmentation tasks, we summarized 7 characteristics of IR small targets, 3 feature extraction methods, 8 design strategies, 30 segmentation networks, 8 loss functions, and 13 evaluation indexes. Then, the accuracy, robustness, and computational complexities of 18 segmentation networks on 5 public datasets were compared and analyzed. Finally, we have discussed the existing problems and future trends in the field of IR small target detection. The proposed survey is a valuable reference for both beginners adapting to current trends in IR small target detection and researchers already experienced in this field.& COPY; 2023 Elsevier Ltd. All rights reserved.