This paper describes a system which locates and grasps doughnut shaped parts from a pile.The system uses photometric stereo and binocular stereo as vision input tools. 本論文之主要研究目標是建立移動式機器人雙眼視覺系統，使機器人具有三維空間的距離感測能力。
Based on characters of human vision, deblurring and contrast enhancement methods are presented for true color images. 研究了受到煙霧惡化的近景彩色圖像的處理原理和方法，探索真彩色圖像基於人眼視覺特徵的消模糊和對比度增強方法。
Experiment results showed that hypercomplex SVD method is much better than traditional methods and it is more consistent with human vision judgment. 實驗結果證明該法評估結果優於傳統彩色圖像質量評估法的評估結果，與人眼視覺評估結果更相符。
An approach based on disparity space to a binocular visual odometry system is presented. 提出了一種基於視差空間的雙目視覺里程計演算法。
