經由一連串連續動作影像的分析來判定物體移動的速度，再藉此速度來求得物體與偵測器間的距離是本研究的目標。為了達此目標，在本論文的研究中，我們發展一個結合了光流偵測 (optical flow estimation) 與深度計算 (range determination) 的被動式距離檢知即時系統。在光流偵測中，我們採用兩個不同性質的方法來估計影像中的光流值；一是Horn-Schunck方法；二是 Lucas-Kanade方法。在Horn-Schunck方法中，我們加入多重解析度的處理以加快光流計算的收斂速度，並得以使本系統更加符合即時的要求。藉由已知的光流資訊，配合有少許動作限制的透視投影公式，我們就能夠計算出目標物與偵測器間的距離，而此項距離資訊將可應用於自動飛行導航系統和軍事戰術演習上。 In the applications of autonomous flight navigation and tactical simulation, we need the range information acquired from the motion cues. In this study, we develop optical flow estimation methods as well as a range determination method to achieve the purpose. Two methods for estimating optical flow from image sequences are compared; one is the Horn-Schunck method and the other is the Lucas-Kanade method. Based on the motion data of optical flow and the principle of perspective transformation, the distance from the targets to the sensor is determined. In the optical flow estimation, the multiresolution strategy is utilized to speed up the process to make our system being suitable for real-time applications.
