[[alternative]]Neural Network Implement On Position Learning Control With Robot Vision

碩士工業教育與技術學系[[abstract]]本研究主要目地是要將機械視覺與機械手臂控制結合一起，應用類神經網路模式，能夠學習兩者的關聯。在訓練樣本的取得，使用電腦模擬，使用計算機圖學模擬空間中的3D物件經空間座標轉換至CCD取得影像的過程，考慮到五軸控制軸的特性，以及CCD成像的性質，經過與實體CCD攝影機所拍攝出來的影像，使模擬與真實影像一致。 經過模擬的影像的推導，作為類神經網路的訓練，因類神經網路具有學習的特性，在正確的訓練下，得到類神經網路，在回想階段，能夠快速達到事先設定的影像位置。 本研究著重以類神經網路解決所面對的問題，在實作上的深入討論，是以理論應用於實際的問題上。[[abstract]]The research aims to combine the mechanical vision with manipulator control. Using neural network could learn the relations between them. The neural network training samples can be obtained by using computer simulation. In the process of using the computer graphics to simulate the 3D object being transformed to the image of CCD, the research considered the range of five controllable axes and the property of the formation of CCD image. The result of comparing the real image with the simulation is almost the same. Through the inference of image simulation, we can get the samples of neural network training .The neural network has the property of learning.With proper training, we can get a neural network. In the step of recalling, the neural network will know how to move to the place of CCD image which we set in advance by the image of current. The research is to solve the problem by neural network