使用專家系統求得具有次圖素精確性的二維定位資訊之研究

Abstract

本論文採用次圖素演算法為工具，利用視訊設備來做物件的二維定位。我們利用數位影像來偵測待測物的位置，並將它移到特定的位置上，期望達到次圖素的精確度。整個系統包含兩個部分：第一部份是先對影像做前處理，接著建立參考座標和參考點，再利用次圖素的方法來求得輸入圖樣中特徵點的位置。當特徵點的位置求得之後，可以用來建立一個關於待測物位置和旋轉角的資料庫。由於移動平台相互之間會有干擾等誤差，我們在第二部分更提出一套專家系統，基於模糊控制的方法，使二維定位的誤差減少，獲得次圖素的精確度。由實驗的結果可知，本系統運用在CCD攝影機模組的定位上，亦即將攝影機鏡頭固定的情況下，把CCD感測器移到正確的位置上，能有效提供次圖素的精確度，呈現令人滿意的結果。

The content of this thesis is a study of the two-dimensional object alignment by using a vision system and a subpixel algorithm. The object position is detected by its digital image. And this object would be moved to a desired location with subpixel accuracy. The system proposed in this thesis is divided into two parts. In the first part, some image processing tech-niques are used to deal with the input images, and a reference coordinate system and reference points are established. Then the coordinates of the feature points of the input image would be found by subpixel algorithm. A database including the coordinates of feature points and rota-tion angles of every representative image is set up after finding these coordinates of the fea-ture points. Because of interference errors between linear stages, an expert system is designed to reduce the error in the second part. Then the two-dimensional alignment results would be superior with subpixel accuracy. By the experimental results, the system is applied to the alignment of the CCD camera module. The CCD sensor would be moved to the desired loca-tion when the camera lens is fixed. And it is clearly to see that the alignment of the CCD camera module is superior with subpixel accuracy by this system.