Abstract: This paper presents a robust human tracking algorithm. As human detection and tracking are important to intelligent surveillance, it is essential to ensure that the system can adapt to any environmental changes and track the objects without interruption. This paper proposes a human tracking algorithm robust against sudden changes of illumination. Sudden illumination changes are automatically detected. The camera’s
exposure time is adjusted to compensate the changes. A mean shift algorithm is developed to continuously track the same object before and after the sudden change so as to achieve reliable tracking. Experiments verify effectiveness of the proposed algorithm.