The detection of anomalies in surveillance videos is essential for maintaining public safety and security. Since surveillance systems record large amounts of video data continuously, especially in cities, transportation centers and public spaces, identifying unusual activities is a major problem. Environmental factors like changing illumination, dense crowds, and random movement patterns are contributing factors to such complexity. This paper undertakes a review of recent developments in anomaly detection methods, their merits and demerits, and their suitability for real-world applications. Special focus is given to deep-learning-based methods, which have significantly improved detection rates and processing speed. Through a review of recent trends and methods, this survey attempts to develop a deeper insight into the field and suggest avenues for the improvement of anomaly detection methods for practical applications in dynamic environments.