In video surveillance systems, a common method for real-time detection of moving objects involves background subtraction. The background subtraction detects objects by differentiating background pixels from foreground pixels and thresholding the difference. This method contains difficult parts. Because, in real world, the background image varies by gradual and sudden illumination change, moving trees and so on. To handle these problems, we use principal component analysis (PCA) for background images and projection onto eigen space. We analyze the experimental results of real-time operation. Finally, we show the robustness of this method.
video surveillance systems
principal component analysis
PCA
eigen space