PhD forum: Illumination-robust foreground detection for multi-camera occupancy mapping

Foreground detection is an essential preprocessing step for many image processing applications such as object tracking, human action recognition, pose estimation and occupancy mapping. Many existing techniques only perform well under steady illumination. Some approaches have been introduced to detect foreground under varying or sudden changes in illumination but the problem remains challenging. In this paper, we introduce a new texture-based foreground detection method which is robust to illumination change. Our method detects foreground by finding the correlation between the current frame and a background model. A region with low correlation is detected as foreground. We compare the performance of our proposed technique with other techniques from literature (edge-based, ViBe and Gaussian mixture model) as a preprocessing step of the multi-camera occupancy mapping system. The evaluation demonstrates that our technique outperforms the other methods in term of object loss.