配備頻寬平順技術之RTP/RTSP即時互動式多媒體串流監控系統設計與實作

Abstract

多媒體串流（Streaming）技術的目的，是使消費者在下載多媒體檔案的同時，可以先觀賞已收到的部份，而不需要等到完整下載後，才開始觀賞。目前最被廣泛接受的串流通訊協定是Real Time Protocol(RTP)，它是針對即時串流的特殊需求所設計的。此協定也可搭配Real Time Streaming Protocol (RTSP)使用。RTSP提供Server與Client之間的雙向溝通，Client可以透過RTSP下指令給Server作像是「播放」、「暫停」、「停止」等動作。市面上有許多軟體廠已有解決方案，此些Servers都符合RTP及RTSP的規範，但只提供單純的多媒體檔案串流。此些Servers無法支援如即時串流無線攝影機所或USB攝影機所拍攝到的影像，或將變動位元速率 (Variable-bit-rate) 的多媒體檔案，做規劃平整 (Smoothing) 的動作，來達到較平順的傳送等。 本論文針對此需求，提出一套符合RTP及RTSP且可以很輕易被擴充功能的Client/Server軟體系統平台 。在本系統裡，Server端可以向各種攝影機擷取影像及聲音，即時的壓縮到MPEG-4的格式，儲存或馬上串流到多個Clients。Client可向Server要求串流攝影機的即時影像和聲音，或已儲存的多媒體檔案等。Client如果要求的是已儲存的檔案，Server會先針對Video的統計特性，做頻寬平整的動作，達到較佳的串流效果及網路資源的利用。Client若為監控中心，因為此系統提供簡單的軟體輸出/輸入界面，所以可在系統上輕鬆的加裝影像處理程式，達到自動化監控的效果。本論文將仔細地介紹此系統的架構、設計想法及使用方法，尤其會特別闡述如何使Server可以在非常準確的時間，將每張frame傳送出去。 本文所提出的目標及期望達到的功能都已達到和實現。然而，在觀察Server的傳送情況後發現，每張frame被傳送出去的時間準確度並不理想，且此誤差在多個clients連線時將更明顯。本文在最後將提出可改善此系統的方向及未來可努力的目標。

Streaming technology allows people to enjoy the multimedia contents while still downloading. Up to date, the most widely used and accepted streaming protocol is Real Time Protocol (RTP), which is specially designed for the needs of real-time streaming, and can be used in conjunction with Real Time Streaming Protocol (RTSP). RTSP provides bidirectional communication between the server and the client. Several software companies have come out with their streaming solutions that comply with RTP and RTSP. However, they merely fulfill the needs of streaming stored media files. They can neither stream real-time captured video/audio acquired from various types of cameras, nor transmit VBR multimedia contents in a smoother manner. In this work, an RTP/RTSP-compliant client/server streaming system that is flexible enough to be added with new functions is implemented. In particular, the server can acquire audiovisual data from different kinds of cameras, real-time encoding them into MPEG-4 format, and store or stream the resulted bitstreams to multiple clients. The client can ask the server to stream live-captured or stored media contents. If the client requests the stored video, the server can run the smoothing algorithm to smoothen the VBR traffic, so a better streaming experience and more efficient utilization of network resources can be obtained. Since the client software is designed to provide a clean and easy-to-use software input/output interface, image processing functions can be added to it effortlessly to allow automotive monitoring of unattended areas. Certainly, the architectures, design ideas, and the usages of the client and server systems will be explained in details. How the server can stream data following tight timing constraints will be addressed specifically. Apparently, the proposed goals are achieved and expected functions are realized. However, it is observed at the server side that the inter-departure times between frames are not equal to the expected values. This discrepancy becomes more obvious when more clients are present. Finally, possible ways to enhance the proposed system and the directions of the future work will be suggested.