在复杂背景和噪声干扰下，如何利用红外探测系统快速且准确发现特征少、强度低、尺度变化、运动状态未知的非合作小目标是一项具有挑战性的任务，备受国内外学者的关注。为了让读者全面了解该领域的研究现状，本综述将从算法原理、文献、数据集、评价指标、实验、发展方向等方面进行总结概括。首先，解释了为什么要以“红外多尺度小目标(点源和小面源)”为对象进行研究的原因并分析了红外多尺度小目标及背景的成像特性；其次，分别讨论了基于经典算法和深度学习算法的原理、设计策略和相关文献，并对比分析了这2类算法的优缺点；然后，总结了现有的红外小目标公开数据集和算法评价指标；最后，分别选取7种经典算法和15种深度学习算法进行了定性和定量的对比分析。通过对单帧红外图像多尺度小目标检测技术的全面回顾，我们对该领域下一步的研究方向给出了9条具体建议。本综述不仅可以帮助初学者快速了解该领域的研究现状和发展趋势，也可作为其他研究者的参考资料。此外，在本领域研究过程中，我们还将现有的20种经典算法和15种深度学习算法集成在人机交互系统中，相关系统的介绍发布在kourenke/GUI-system-for-infrared-small-target-detection (github.com)。

In complex and strong electromagnetic interference environments, especially when facing stealth targets with extremely low radar cross section (RCS), traditional radar detection is almost ineffective. In such scenarios, the infrared search and track (IRST) system, with its strong anti-interference capability, all-weather and all-airspace passive detection, emerges as a viable alternative to radar in target detection. Therefore, it is widely used in tasks such as reconnaissance and early warning, maritime surveillance, and precision guidance. However, the efficient and accurate use of the IRST system in identifying non-cooperative small targets with minimal features, low intensity, scale variations, and unknown motion states in complex backgrounds and amidst noise interference remains a challenging task, drawing attention from scholars globally. Up to now, the research on infrared (IR) small target detection technology is mainly aimed at long-distance weak and small (point source) targets, but when the scene and target scale change significantly, false alarms or missed detections are easy to occur. Therefore, this review focuses on the problem of IR multi-scale small target detection technology. To provide a comprehensive understanding of the current research status in this domain, this review summarizes the field from the perspectives of algorithm principles, literature, datasets, evaluation metrics, experiments, and development directions. First, the research motivation is clarified. In the practical application background, with the change of the motion state of non-cooperative small targets, the scale will also change greatly, from point-like targets to small targets with fuzzy boundaries to small targets with clear outlines, which is usually difficult to be clearly defined. Therefore, to be more in line with the practical application background, this review is comprehensively analyzed from the perspective of IR multi-scale small target detection technology. Second, the imaging characteristics of IR multi-scale small targets and backgrounds are analyzed. Among them, the targets are characterized by various types, scale changes (from point sources to small surface sources), low intensity, fuzzy boundaries, lack of texture and color information, and unknown motion status. The background is characterized by complex and variable scenes and serious noise interference. Then, the algorithm principles, related literature, advantages and disadvantages of different algorithms for single-frame IR image multi-scale small target detection techniques are summarized. In this review, we categorize the IR multi-scale small target single-frame detection techniques into 2 main categories: classical algorithms and deep learning algorithms. Among them, classical algorithms are classified into background estimation method, morphological method, directional derivative/gradient/entropy method, local contrast method, frequency-domain method, ultra-complete sparse representation method, and sparse low-rank decomposition method according to different modeling ideas. Deep learning algorithms are categorized into convolutional neural network (CNN), classical algorithm +CNN, and CNN + Transformer according to the network structure. In these network structures, to adequately extract the IR multi-scale small target feature information, design strategies such as contextual feature fusion, multi-scalar feature fusion, dense nesting, and generative adversarial networks have been introduced. To reduce the computational complexity or the limitation of data sample size, strategies such as lightweight design and weak supervision are introduced. Objectively speaking, both classical algorithms and deep learning algorithms have their advantages and disadvantages, so the appropriate algorithms should be selected according to the specific problems and needs. In addition, combining the two types of algorithms to give full play to their advantages is a current research hotspot. Finally, 10 existing public datasets and 17 evaluation metrics are organized, and 7 classical algorithms and 15 deep learning algorithms are selected for qualitative and quantitative comparative analysis, respectively. In addition, in the process of research in this field, we have integrated 20 existing classical algorithms and 15 deep learning algorithms in a human-computer interaction system, and the introduction of the relevant system is published in kourenke/GUI-system-for-infrared-small-target-detection (github.com). Through a comprehensive review of multi-scale small target detection techniques in single frame IR images, we have provided 9 specific suggestions for the next research directions in this field. This review can not only help beginners quickly understand the research status and development trends in this field, but also serve as a reference material for other researchers.