本研究主要目的在於開發一套虛擬影像導航系統，以模擬自主式水下無人載具(Autonomous Underwater Vehicle , AUV)進行水下檢測任務時之最佳航線規劃。在研究中將呈現AUV透過虛擬影像檢測環境，同時 也對未知的障礙物進行避碰以確保航行安全。在路徑規劃上，則是使用多目標粒子群演算法(Multi-Objective Particle Swarm algorithm, MOPSO)結合虛擬影像之障礙物資訊，對各待選航向進行迭代搜索，並以時間與 電量消耗做為依據，以判斷出最佳航向。在影像的虛擬化過程中，將在有限的影像空間內進行偵測點分佈， 並參考立體視覺影像(Stereo Vision)技術的特點，以判斷影像中的環境深度(即影像深度的可視化)，藉此將 可清楚判斷障礙物的特徵。綜而言之，本系統除了有利於最佳化路徑的選取外，也可在進行水下檢測任務 時快速判斷待測物的特徵，進而達到追蹤檢測的目的。最後，在模擬過程中將建構不同的水下待測物進行 測試，分析結果將在影像導航與AUV之水下檢測系統之結合上提供一重要參考依據。

This study aims to develop an virtual-image navigation mode to support the optimal route plan of Autonomous Underwater Vehicle (AUV) for the inspection task. For the safety of navigation, the results present the inspection process by virtual image and obstacle-avoidance as well. In terms of path planning, the optimal route is chosen by adopting the method of Multi-Objective Particle Swarm algorithm (MOPSO) with respect to virtual images of obstacles. Thus, each feasible route is searched iteratively according to two objectives, i.e. sailing time and joules consumption. In the process of image virtualization, the detecting points will be distributed over the limited space and the environmental depths for identifying the features of obstacles will be visualized referring to the stereo vision technique. In summary, the system is not only beneficial to optimize feasible routes but also identify features of obstacles for the purpose of tracking. Eventually, several underwater objects would be constructed and tested in the simulation for combining the image navigation with the underwater inspection of AUV.