Online Change Detection

Online change detection (OCD) focuses on rapidly identifying shifts in data streams, a crucial task across diverse fields like security and surveillance. Current research emphasizes developing robust and efficient algorithms, including those based on statistical modeling of deep features, non-parametric methods like QuantTree, and techniques leveraging maximum mean discrepancy or density ratio estimation, often with a focus on minimizing false alarms and achieving provable optimality. These advancements improve the reliability and speed of change detection in various applications, ranging from video surveillance and robotic mapping to anomaly detection in complex systems. The development of asynchronous and computationally efficient algorithms is a key trend, enabling real-time processing of high-volume data streams.