運用於智慧型交通系統中之強健型背景抽取法

Robust Background Subtraction Algorithm in Intelligence Traffic System

影像式智慧型交通系統必需具備在各式道路及天候下均能運作之能力，因此，快速的背景抽取與精確成功的前景切割，將是影響該系統是否順利運行的關鍵因素。本篇提出一個運用機率模式之強健型背景相減法，本方法可以迅速且有效率的自連續彩色影像中抽取背景，而其中背景抽取的概念是藉由統計影像中每一個像素色彩變化及出現的機率，設計一個門檻值來做為背抽取的依據，在初始背景完成抽取後，除籍由影像相減對移動物體正確且即時切割外，並同時更新初始背景影像，而所提之方法可以達到即時且高效率之背景抽取及移動物件切割，目前藉由各種道路環境的實測結果，驗證了所提之方法的強健性、準確性與即時性。

A video-based Intelligence Traffic System (ITS) must be capable of continuous operation under various road conditions. Moreover, background subtraction is a very important part of ITS applications for successful segmentation of objects from video sequences. Accuracy and computational time of the initial background extraction are crucial in any background subtraction method. This paper proposes the probability-based background extraction algorithm to segment objects from surveillance videos. With the proposed algorithm, the initial background can be extracted accurately and quickly by calculating the color probabilities of each pixel to decide the background pixel color. After the initial background extraction, the intrusive objects can be segmented correctly and immediately. Meanwhile, the color background images can be updated in real time to overcome any variation in illumination conditions. Experimental results for various environmental sequences and a quantitative evaluation are provided to demonstrate the robustness, accuracy, effectiveness, and economy of computation time of the proposed algorithm.