立體影像系統用於移動物體追蹤之研究

Abstract

在這篇論文中我們設計並製造完成一組立體影像追蹤系統。 為了預測目 標物的移動以減少追蹤的誤差,我們應用了卡曼濾波器(Kalman filter)與 模糊邏輯(fuzzy logic)來設計不同的估測器,並以實驗來測試這兩種估測 器的性能。實驗的內容是將一塊釘了釘子的板子傾斜某一個角度,並使一 顆球由斜坡頂端沿斜波自由滾下由本系統進行追蹤並記錄誤差。實驗結果 發現當傾斜角度小於30°時由模糊邏輯所構成的估測器有較佳的表現若傾 斜角度大於30°時兩種估測器的性能都急劇劣化。 We design and implement the stereo vision tracking system. In order to reduce the tracking error, we design two predictors with Kalman filter and fuzzy logic respectively to predict the movement of the target. We design an experiment to test the performance of fuzzy predictor and Kalman predictor. The setup of experiment device is taking a white flat board with some nails clinched on the surface, elevating one end of the board and making the board form a descent. Then put a little ball on the top of the descent and let the ball roll down the descent freely. The tracking system will track the ball and record the tracking error. Fromthe results of the experiment, we know that when the acclivitous angle is smaller than 30°, the tracking error of fuzzy predictor is smaller than Kalman predictor. But when the acclivitous angle is larger than 30°, the performance of the two predictors will become bad quickly.