Abstract: The coal-dominated energy structure will persist for an extended period during China’s industrialization process, making the safe and efficient extraction of coal resources critically important. With the advancement of intelligent coal mine construction and machine vision technology, object detection-based intelligent safety management systems have been widely adopted in the coal industry. Among these, the YOLO series algorithms demonstrate significant advantages in underground intelligent detection tasks due to their high real-time performance and strong generalization capabilities. As the primary site for coal production, fully-mechanized mining face pose multiple challenges for the practical application of YOLO algorithms due to their complex, harsh environments and dynamic human-machine interaction scenarios. To systematically organize and advance the development of YOLO algorithms for target detection in fully-mechanized mining face, this paper first explains the fundamental principles of YOLO algorithms. It systematically reviews the evolution and technical architectures of over ten typical variants, including official versions from YOLOv1 to YOLOv12 and YOLO3D, identifying three key technical evolution paths: accuracy enhancement, lightweight design, and scene adaptability. Subsequently, focusing on the unique conditions of fully-mechanized mining face, it systematically summarizes image preprocessing methods, classic public datasets, and model evaluation frameworks. Centered on the four-dimensional application requirements of “person-machine-environment-management”, it thoroughly analyzes the current application status of YOLO algorithms in tasks such as monitoring unsafe human behaviors, identifying equipment status, intelligent environmental perception, and safety management assistance. It compares and analyzes the improvement strategies and field performance of different algorithm versions. Building on this foundation, we identify four major interference factors-low illumination, heavy dust, intense vibration, and dense small targets, and summarize mainstream optimization approaches including image enhancement, multimodal fusion, temporal compensation, and multi-scale feature integration. Finally, considering both the demands of intelligent coal mine development and the challenges of the complex underground environment, this paper discusses the future direction of object detection algorithms for fully-mechanized face. This work aims to provide a theoretical reference and technical support for intelligent safety monitoring and the high-quality development of intrinsically safe practices in China’s coal mines.