適應性的人物偵測與追蹤技術

為因應龐大的視訊資料處理，本論文提出一種自動化與適應性人物偵測與追蹤技術。傳統的人物偵測與追蹤技術通常會面臨一個問題：人物被遮蔽導致追蹤方法失效。為有效解決這個問題，本論文提出二階段人形偵測技術，分別處理遮蔽性人物與非遮蔽性人物的情況。最後，為有效追蹤此兩種人形，以Particle filter配合Bhattacharyya distance來做為此兩種追蹤模式的切換基準。由實驗得知，所提出的方法能正確的偵測出人形，並有效的進行追蹤。

關鍵字

物件偵測；高斯混合模型；陰影去除；微粒濾波器；物件追蹤

並列摘要

An integrated technique of automatic and adaptive human detection and tracking is proposed for large amount of video data. The traditional detection and tracking method usually suffers from the following problem: The occluded human is difficult to be tracked. In order to solve this problem, a two-stage human detection scheme is proposed to deal with the cases of whether the human is occluded. Finally, the Particle filter and the Bhattacharyya distance are applied for tracking the two kinds of human shapes. From the experimental results, the proposed method can detect and track human shapes correctly and effectively.

並列關鍵字

Object Detection ； Gaussian Mixture Model ； Shadow Removal ； Particle Filter ； Object Tracking

國際替代計量

適應性的人物偵測與追蹤技術

全文下載

主題瀏覽