基於無線感測網路資料相關特性之類視訊壓縮方法

Abstract

無線感測網路通常含有大量的節點，這些節點長時間的量測與記錄環境中的資料，並將感測的資料傳送至資料伺服器(Data server)儲存。資料伺服器儲存這些感測資料前，需加入時間標籤與偵測環境相關參數以辨別蒐集資料的節點位置與時間。這些龐大的資料若不經處理直接儲存在伺服器中，除佔去大量的儲存空間、不利資料查詢，更可能嚴重降低資料伺服器的效能。我們觀察某些環境中的感測資料，發現位置相近的不同節點感測到的資料數值相差較小，具有空間上的關係；而同一節點連續兩筆偵測到的資料數值差異不大，具有時間的相關性。因此，我們利用感測資料的相關性，欲開發基於無線感測網路資料相關特性之類視訊無失真壓縮方法，稱為VLLC。於VLLC中，感測資料經無失真的重新排列與數值轉換後，成為以時間排列、連續的格式化影像。接著，我們將這些連續的格式化影像以無失真影像壓縮的標準，H.264，將其壓縮。排列轉換後的影像中，存在無線感測資料的時間相關性與空間相關性，故可有效地被壓縮成較小的檔案。VLLC同時提供壓縮與解壓縮的平行化方法，可依系統需求與硬體配置作彈性的應用。而VLLC除了可架構於既有的資料庫上，提供外部資料壓縮，亦可延伸其功能，成為一獨立的資料庫系統。實驗結果可分為兩部分，第一部分實驗為影像大小影響壓縮效能的比較；第二部分實驗中，我們將VLLC視為一簡化的資料庫，並與一廣泛使用的開放式資料庫，MySQL，作資料建立、資料壓縮與資料存取等效能的比較。由實驗結果可得知，VLLC存取大量資料的效能表現可與MySQL相當，但VLLC建立資料的時間，較MySQL節省約94%；而資料壓縮後，VLLC較MySQL省下約79%的儲存空間。

Wireless Sensor Networks (WSNs) consist of group sensor nodes which are placed in an area to monitor the changes of environment. Numerous of sensory data in WSNs are usually gathered in data server for many purposes. Those sensory data should be well organized and compressed, or they will occupy a vast amount of storage, and lead to the decrease in server’s performance which occurs in critically ill. In this thesis, we propose a video-like lossless compression method, VLLC, which aims to adopt the spatial-temporal correlation of sensory data in WSNs to enhance the performance of data compression. In VLLC, raw data will be transformed and arranged as the formatted images without loss according to the spatial correlation. Several continuous images with temporal correlation are maintained as sequential frames in video, and then be compressed by lossless video compression method. A flexible data retrieve method is also described in VLLC, of which parallel processing can meet the requirements of data server. VLLC can be an external compression method to existing frameworks, or its capability can be extended to be an individual database. The trade-off between space saving and retrieve time is discussed with real-world data. A well-known database, MySQL, is compared in our experiments. The experiments show that our method saves approximate to 94% of construction time than MySQL, and the storage cost of our video-like lossless compression database is 79% less than MySQL after normalizing the value. Although this method can save lots of storage and construct time for sensory data, the performance of its data retrieve can be achieved as well as the performance of MySQL.