<![CDATA[Dense-crowd video analysis is a branch of computer vision that has various important applications in public safety, emergency management, urban planning, pedestrian traffic engineering, and crowd management at large events, such as religious activities, music concerts, and sports matches. This study presents a Systematic Literature Review (SLR) of 30 scientific publications published between 2010 and 2025. The main objective of this review is to identify the latest research trends, classification of algorithms used, application domains, and the main challenges still faced in crowd video analysis. The results of this SLR show that deep learning-based approaches, such as Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and Transformer, still dominate various applications, especially in anomaly detection which aims to recognize suspicious behavior in dense crowds. This technology has significant potential for preventing dangerous events such as riots, mass panic, or accidents. In addition, trends in the integration of new technologies are also found, such as the use of hybrid algorithms that combine several approaches, federated learning for distributed model training, and the use of multimodal data and drones to improve monitoring effectiveness. However, many challenges remain, such as limited representative datasets, decreased accuracy under extreme conditions, computational limitations for real-time applications, and issues of privacy and model interpretability. Therefore, the results of this SLR are expected to make a strategic contribution to the development of more sophisticated, adaptive, and relevant crowd analytics systems.]]>
