摘要
隨著資訊科技的進步，使得電腦硬體的效能提升、成本下降，整個運算環境比過去擁有更高的執行效率與便利性，而傳統的人工作業流程已逐漸被電腦自動化流程取代，因為電腦自動化作業的關係在處理流程中能夠節省更多的時間與人力資源，因此讓現今的產業能獲得更大的效益。本文中提出一套即時生物追蹤系統，目的是為能應用在生物行為研究上，藉此輔助研究人員而達到自動化作業及減少時間、人力與成本，並且降低觀測上之人為誤差。 本系統使用JAVA為開發工具，利用攝影機擷取連續影像。藉由背景相減方式、區域成長演算法、連通成份標示法等，系統能夠從連續的影像變化中自動擷取出物件資訊，利用這些影像數據來分析個別不同物件之重心位置、行進路線並將其記錄下來，達到物件追蹤，並且利用CIE2000色差公式將重疊的物件分離，增加系統準確性。除此之外本系統亦包含物件旋轉角度量測、物件長度、頭部位置偵測等功能，使得系統成為多元化的生物行為量化平台。
並列摘要
Because of the advances in computer technologies, applying information technologies for automation could save a lot of manpower. Therefore, lots of traditional manual processes were gradually replaced by the automated processes. In this thesis, we would introduce a real-time bio-tracking system for studying animal behaviors. This system not only could save manpower but also could eliminate the error made by the observers. This system is developed in JAVA. Image sequences were captured by IP cameras. Interested objects were segmented automatically from each image. The location of the barycenter of each object was computed and recorded. These locations would form the locus of each object. In order to solve the problem of the overlapped objects, CIE2000 color difference formula was applied to tell different objects. In addition to recording the loci, information like angular velocities of objects, the lengths of the objects, position of the head in each object were recorded for further analysis.