利用時空域分析與背景相減法作視訊移動物偵測

No Thumbnail Available

Date

2009

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

利用電腦視覺方式做移動物偵測時,所遭遇到最大的問題就是動態背景雜訊以及前景本身因移動而產生的雜訊,尤其在使用背景相減法作前景擷取時,這兩種雜訊更為明顯。因此,本論文提出結合前景物時域、空間域以及色彩資訊等方式來改善偵測的正確性。本方法可分為主要三個部分:(1) 利用時序統計長方圖的方式建立可隨時間更新的背景。(2) 再以angle-module方法將三維色彩資訊轉換為二維的色相變化與色彩強度資訊,利用自適應的背景相減法擷取動態前景物,運用前景與背景色彩資訊的差異性來將前景物雜訊去除(陰影、小變化雜訊)。(3) 最後結合影像時間與空間資訊的概念,來去除動態背景雜訊(例如搖曳的樹枝、雨天..等)。 實驗結果顯示,本研究的系統在室內或室外環境下都有九成以上的偵測正確率。對陰影、動態背景雜訊、以及攝影機輕微搖晃等容易造成誤判的條件下,系統也能夠有著不錯的偵測準確率。
The critical issues of motion detection based on computer vision are the noises in the dynamic background and the noises from objects’ moving in the foreground. These two noises are more obvious, especially at using background subtraction method. In this study, A method that combined with temporal-spatial and color information is used to improve the detection accuracy. The method can be divided into three sections: (1) The time-varying updated background is built by temporal statistic histogram; (2) Three dimension color information is transferred into two dimension color phase and color intensity by angle-module rule. Next, moving objects in the foreground are extracted by adaptive background subtraction, and the noises (shadows and small change) are removed according to variations of color information in the background and foreground; (3) Dynamic background noises (ex: branches movements and rain interferences) are removed by the concept combined with temporal and spatial information of video sequences. As the results present, our accuracy of the detection is upper than ninety percentage in the outside and inside environments. The system also has good performance when the false detection is caused by shadows, dynamic background noises, and camera shakings.

Description

Keywords

時空域分析, 陰影偵測, 移動物偵測, 背景重建, 背景相減, angle-module 色彩座標轉換, Temporal-spatial analysis, shadow detection, dynamic object detect, background rebuilding, background subtraction, angle-module rule

Citation

Collections