前景偵測技術應用於背景重建之研究

A Study of Background Reconstruction using Foreground Detection Technique

劉啟平(Chi-Ping Liu)

指導教授 ： 吳明川

國立臺北科技大學/機電學院/機電整合研究所/碩士(2010年)

摘要

在電腦視覺領域中，背景影像(Background Image) 經常作為判斷事件的參考依據，例如物件追蹤、監控保全、遺留物偵測…等。而在監控保全方面，因為場景會隨著時間而產生小部分的異動，需要時常更新背景影像。另外在複雜前景的移動干擾下(例如在人群來來往往的拍攝場合)，如何取得完整的街景或風景，也是一個困難的問題。因此
本研究分為前景偵測及場景重建兩部分。前景偵測使用改良式背景相減法，改善固定背景相減法(Static Background Subtraction)需要事先取得參考背景影像的問題，並配合色彩特徵分析判斷出前景的區域。場景重建使用核密度估測(Kernel Density Estimation)建立出前景周圍區域以及背景候選影像的機率密度函數，找出適合的影像取代前景區域，完成背景重建。

關鍵字

背景重建 ； 固定背景相減法 ； 前景偵測 ； 核密度估測

並列摘要

In the Computer Vision filed, the background image was an important reference to judge event, like Object tracking, Surveillance Security, Abandoned. Objects Detection et.al.. Scene was always changed and system needed to update the background images in Surveillance Security. It was difficult to get clear scene under disturbance of moving objects. This paper integrates foreground detection technique and background reconstruction technique into a new method which builds background images automatically.
There are two parts in this thesis, foreground detection and background reconstruction. Foreground detection used advance background subtraction and color feature analysis to get foreground area. It improved static background subtraction defects which needed to get background image first. Background reconstruction used Kernel Density Estimation (KDE) to build probability of surround area of foreground and background candidate images. It found suitable image to replace foreground.

並列關鍵字

Background Reconstruction ； Static Background Subtraction ； Foreground Detection ； Kernel Density Estimation

參考文獻

[2] R. Cutler and L. S. Davis, “Robust real-time periodic motion detection, analysis, and applications,” IEEE Trans. Pattern Anal. Machine Intell., vol. 22, Aug. 2000, pp. 781-796.

[4] Y. Ran and Q. Zheng, “Multi Moving People Detection from Binocular Sequences,” Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing, vol. 3, Apr. 2003, pp. 37-40.

[5] S. Huwer and H. Niemann, “Adaptive Change Detection for Real-time Surveillance Applications,” Proceedings of IEEE third International Workshop on Visual Surveillance, July 2000, pp. 37-46.

[6] R. Cuchiara, C. Grana, M. Piccardi, and A. Prati, “Detecting objects, shadows and ghosts in video streams by exploiting colour and motion information,” in Proc. Int. Conf. Image Anal. Process. Palermo, Italy, Sep. 2001, pp. 360-365.

[7] H.J. Elias, O.U. Carlos and S. Jesus, “Detected motion classification with a double-background and a neighborhood-based difference,” Pattern Recognition Letters, vol. 24, no. 12, Aug. 2003, pp. 2079-2092.