以高斯混合模型為基礎並使用陰影濾除之動態背景影像模型建立

Abstract

在本論文中，我們提出一個能隨環境改變，穩健且適應性的背景濾除系統，此系統主要分成三個部分，首先，採用高斯混合模型(GMM)的方法去建立動態的背景模型，並利用期望值最大演算法(EM Algorithm)去估計模型的參數，接下來結合背景模型中顏色與梯度的統計資訊，階層式地做前景與背景的判定，並由判定的結果與連續影像之間變化的程度，更新並重建背景的模型，讓模型具有記錄所有發生過狀況的能力。因為環境的改變或是前景物體的移動，容易產生光影的變化，會降低背景濾除的正確性，因此，我們的系統在結合短期模型的機制下，提出適用於GMM方法的陰影濾除演算法，最後，針對室內環境中可能發生的各種狀況，進行實驗驗證及討論。

In this thesis, a robust and adaptive Background Subtraction algorithm is proposed and implemented. This system mainly consists of three stages. In the first stage, we build a dynamic background image model based on Gaussian Mixture Model method and estimate the model parameters using Expectation Maximization Algorithm. Then, a hierarchical method with color and gradient statistical information to separate background and foreground is proposed. The parameters of background model are updated according to the results of separation and variation of sequence images so all events happened are recorded in the background model. The changes of environment or moving object may cause brightness variations in background scene. These effects will deteriorate the performance of system. Therefore, a shadow removal algorithm combining long-term model with short-term model is proposed. Lastly, the methods are experimented in various indoor conditions. Discussion and comparison on the results are given.