利用照度不變分量及多支攝影機連接區域應用於室內監視系統之研究

Abstract

在本篇論文中，我們發展出一個結合照度不變局部分量(反射分量)及影像鑲嵌技術的室內監視系統。系統可藉由同態濾波來取得背景的模型以避免光線變化的影響，而背景相減的過程可被用來將影像中的像素分類成背景與前景的點，再利用連接特性，可取得背景與前景的群集。在影像中的每一個像素的臨界值可以去選擇以適應監測環境光線的變化。此外，影像鑲嵌技術可以被用來連結所有攝影機的監測範圍以避免死角或使用單一攝影機而產生一些無法監測的區域。

In this thesis, an indoor vision surveillance system which uses illumination invariant local components (reflection component) and image mosaic technique is proposed. The background model will be obtained by using the homomorphic filtering in order to prevent the effects of light-varying, and the background subtraction procedure can be applied to classify the image pixels into the background pixels or foreground pixels. Then, the connectivity information can be used to the background-foreground classification process. Threshold for every pixel in the image can be selected to adapt the change of lighting. In addition, the image mosaic technique can be used to cover the full scene of the vision surveillance system for monitor the activity in order to prevent the corners or some non-visible areas of one camera view.