結合廣角攝影機與旋轉式攝影機之視訊監控系統

近年來，視訊監控系統在生活中扮演越來越重要的角色。傳統視訊監控的方法是在多個重要地點分別裝設攝影機，並透過人力監看各攝影機畫面的狀態。本篇論文旨在設計一個可由單人簡易裝設完成之全自動的視訊監控系統，結合廣角攝影機與旋轉式攝影機之特性：廣角攝影機之大範圍監控與旋轉式攝影機之高解析度畫面，以達到同時俱有即時性、高偵測率與目標物體高解析度影像的目標。我們透過結合下列數種方法組成此系統：(1)攝影機系統校正：先利用二維的平面校正法求得攝影機之內在參數，再利用空間中已知相對位置的物體求得各攝影機間的外在參數；(2)移動物體偵測：利用混合式高斯分佈，即時更新背景畫面之統計模式，並偵測與背景不同的物體位置；(3)移動物體追蹤：利用mean-shift演算法追蹤畫面上的移動物體；(4)人臉偵測：採用階層式的boosting演算法，快速且準確地偵測出畫面中的人臉。在系統運作的過程中，我們將偵測及追蹤兩種方法加以結合，互相配合以控制旋轉式攝影機的運動，使旋轉式攝影機不斷地對準被追蹤物體。本系統的所有安裝及校正步驟皆可由單人完成。而在單一目標物的情形下，可準確持續地追蹤物體。

關鍵字

攝影機校正；視訊監控系統；影像分析；物件偵測；追蹤

並列摘要

Visual monitoring systems will play a more and more important role in our daily life. Traditional Close Circuit Television System (CCTV) sets cameras in many key positions, and need a guard to keep an eye on the videos of those cameras. In this work, we developed a fully automatic visual monitoring system that can be set up with a simple method. Utilize the properties of the wide-angle cameras and of the pan-tilt camera (PT-camera), i.e., large view field of wide-angle cameras and high-resolution images of PT-cameras, we can design a real-time visual monitoring system having high detection rate and clear target images. Important methods used in this system include the following: (1) Camera calibration: First, we estimate the intrinsic camera parameters using a camera calibration method that require simply placing a 2D calibration plane at different positions. Then, we use a cross rig having calibration markers, each marker with known 3D position, to estimate the extrinsic parameters between cameras. (2) Moving object detection: For the wide-angle cameras, Gaussian mixture is used to update the background model in real-time, which is then used to detect objects different from the background. (3) Object tracking: For the PT-camera, we use the mean-shift algorithm to keep track of the target. (4) Face detection: A cascade of boosted classifiers, working with Haar-like features, is used to obtain fast and accurate face detection. In the experiments, we combine the detection and tracking methods together to reliably fixate the PT-camera on the object. One single person can easily set up the whole system and calibrate all cameras. Our experiments showed that, when there is only one moving object in the monitoring area, our system could track the object successfully.

並列關鍵字

Image Analysis ； Visual Monitoring System ； Tracking ； Object Detection ； Camera Calibration

參考文獻

[Arun 87] K. S. Arun, T. S. Huang, and S. D. Blostein, “Least-Squares Fitting of Two 3-D Point Sets,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 9, No. 5, Sep. 1987.

[Besl 92] Paul J. Besl, “A Method for Registration of 3D Shapes,” IEEE Transactions on Pattern Analysis and Machine Intelligence, VOL. 14, No. 2, Feb, 1992.

[Brow 71] D. C. Brown. “Close-range camera calibration,” Photogrammetric Engineering, 37(8), pp.855–866, 1971

[Chen 99] Chu-Song Chen, Chi-Kuo Yu, and Yi-Ping Hung, “New Calibration-free Approach for Augmented Reality Based on Parameterized Cuboid Structure,” ICCV, pp.30-37, 1999.

[Chen 01] Hwann-Tzong Chen, and Tyng-Luh Liu, “Trust-Region Methods for Real-Time Tracking,” ICCV, pp. 717-722, 2001

國際替代計量

結合廣角攝影機與旋轉式攝影機之視訊監控系統

全文下載

主題瀏覽