使用PTZ攝影機之近環場式視訊監控保全系統

A Near Panoramic Video Surveillance System Using the PTZ Camera

張厥煒(Chueh-Wei Chang) ； 許政元(Zheng-Yuan Xu)

《臺北科技大學學報》 42卷1期 (2009/06) Pp. 67-81

https://doi.org/10.29768/JNTUT.200906.0005

摘要

近年來社會犯罪案件頻繁，銀行、超商及住家等搶案不斷地發生，為能減低及嚇阻犯罪案件的增加，監控系統的存在，就彰顯出其極其重要性。而拜科技進步所賜，監視器材不但比以往精密，價格更為便宜，也讓監控保全方面的系統，能得以大量的開發。然而，絕大部份的監控系統，仍然是以固定靜止的方式來進行特定視角的異常偵測。由於固定式的監測系統容易存在著監測死角的問題，如果要執行更為廣泛的監控範圍，則必須加裝更多的監視器，成本需求必然增加。
為能夠解決監測死角之問題，並且有效降低系統裝置成本，我們則是提出了一套結合PTZ攝影機(Pan/Tilt/Zoom Camera)的智慧型監控保全系統。透過PTZ攝影機來拍攝所需監測的環境局部區域影像，並且藉由使用SIFT 特徵點選取與RANSAC影像接合技術(Mosaicing)，來將其接合成新的環場影像(Panorama Image)。接著是針對環場背景影像，來進行相關的特徵資訊分析。而為了能夠分辨出監測環境中，是否有異常物體的存在，我們則是進行即時影像與環場背景影像之間的比對，透過特徵點比對的結果來判斷PTZ攝影機的監控位置，以及前景與背景影像之間的差別。一旦比對的結果是有異常物體的存在時，系統則可立刻針對異常物體來進行物件分割、辨識、追蹤等其它相關後續應用。

關鍵字

監控系統Surveillance System ； PTZ攝影機 ； SIFT特徵點偵測 ； 環場影像Panorama

並列摘要

The social crime is getting worse in recent years. The robber of the banks, convent stores and families, are happening frequently. In order to reduce and prevent the increase of the crime, the existence of the surveillance system shows how import it is. With the help of the science and technologies, the functions of surveillance systems become more accurate and cheaper. Till now, most of the surveillance systems can only detect the abnormal events using a still camera and still has the problem about the blind corners. To solve this problem, more cameras must be installed and will spend more money.
In this paper, we propose an intelligent surveillance system by using the PTZ (Pan/Tilt/Zoom, PTZ) camera. In our surveillance system, we need to stitch a panorama background from a sequence of continuous images. A panorama background can be generated by the SIFT (Scale Invariant Feature Transform) feature point detector and the RANSAC (RANdom SAmple Consensus) correspondence matching algorithm. A feature point information database is created from the feature points in the panorama background. By the process of the similarity measure between two sets of the feature vectors from current image feature points and background feature points, we can find the location where the current camera is. Once the current image location has been detected, we can easily check whether there has a different object in the surveillance area or not by a background subtraction algorithm. Our system can then segment the abnormal object(s) from the background, and follow a tracking or recognition procedure.

並列關鍵字

Surveillance System ； PTZ camera ； SIFT Feature Point Detector ； Panorama