Abstract: In the current energy landscape, coal serves as a pivotal supporting energy source in China’s industrialization process, and the energy structure dominated by coal will remain stable for a long period. Therefore, ensuring coal mine safety production is of paramount importance, and the intelligent construction of coal mines has become the core pathway to achieve to achieve this goal and guarantee efficient and safe operations. In this process, video analysis and recognition technology for coal mine safety production plays an indispensable role, and its in-depth development is a key technical pillar for ensuring the high-quality development of the coal industry.However, the special underground environmental conditions pose severe challenges. Due to the uneven illumination from artificial light sources, images captured by video surveillance generally exhibit poor conditions such as low illumination, poor contrast, severe noise interference, and missing details. These issues seriously interfere with the precise execution of subsequent key tasks such as target detection and semantic segmentation, thus posing a huge threat to real-time monitoring and effective early warning of coal mine safety production, and becoming critical challenges to be urgently addressed in the intelligent construction of coal mines. In view of this, a comprehensive and in-depth review of image enhancement methods in low-light environments of mines is of extremely important strategic significance.This study first systematically combs traditional low-light image enhancement methods, including histogram equalization, gamma correction, wavelet transform, Retinex decomposition, and fusion-based methods. These methods are carefully classified according to their principles, implementation processes, and application effects, while the advantages and limitations of each method are deeply analyzed, providing important experience and theoretical basis for follow-up research. Furthermore, it focuses on deep learning-based low-light image enhancement methods for mines, which are accurately divided into supervised and unsupervised categories according to the learning paradigm. For typical algorithms in each category, their innovations, advantages, and disadvantages are elaborated in detail, offering a comprehensive and in-depth technical analysis for field research.In addition, this study comprehensively summarizes common datasets and evaluation metrics for low-light image enhancement, clarifies the characteristics and application scopes of different datasets,summarizes the construction of mine low-light datasets,and conducts experimental comparative analyses of traditional methods and deep learning-based methods based on this. Finally, by closely integrating the urgent needs of coal mine intelligent development with industry trends, it analyzes the current dilemmas and challenges faced by low-light image enhancement in mines from multiple dimensions, and proposes reasonable prospects for future development directions. The goal is to provide comprehensive guidance for the continuous innovation and wide application of low-light image enhancement technology in coal mines, and help the coal industry achieve safe, efficient, and sustainable development in the intelligent era.