Robust real-time object tracking under varying illumination condition

An object tracking is simply a problem of finding the different positions of the object in each frame of a video. A novel method is proposed here for extracting target objects from real time video and then tracking them under varying illumination condition. It operates on video imagery taken from multiple video cameras. Each camera is connected to an infrared sensor and is distributed around the site to get the complete picture of the suspicious activities of intruders. By using an array of infrared sensors, each video camera is made event-driven. Thus a camera starts recording only after being triggered with the output of nearby infrared sensor. Thus it becomes a compressed sensing system. Our proposed method uses homomorphic filtering to overcome varying illumination condition. The illumination component of an image has slow spatial variations. Homomorphic filter is employed to reject very low frequency components that usually represent illumination variation.