一個好的光學成像系統必須建立在沒有像差的基準上，而會造成像差的原因有許多種，其中最常發生的原因是無法對焦於同一光軸上的任一點，各相機系統廠為解決此項問題大多利用雷射準直儀，作為鏡片及CCD Sensor的平行對位校正，但此類型的設備較為昂貴。 目前機器視覺的技術已發展成熟，廣泛的應用在各產業界中，其中不乏使用在對位技術方面，本研究係經由影像切割將背景與檢測樣本作區分後，在進行行距間的量測判定，利用兩組CCD分別架設於機台的上側與前側，計算出樣本XYZ軸的平行度與間距是否與設計值相同，並透過類神經網路的學習過程，將判定的準確率接近至百分之百，使其數位相機的像差問題，可獲得有效的改善。 A good optical imaging system must be established benchmarks in the absence of aberration ,The reasons can cause many types of aberration ,which is the most common reason for not focusing on any point on the same optical axis ,The ODM(Own Designing & Manufacturing) factory of DSC(Digital Still Camera) to solve the problem mostly use the laser on the favorite ,as the lens and the CCD Sensor parallel alignment correction ,However ,this type of equipment is more expensive. Currently ,machine vision technology has been developed ,Widely application in all works of life and always can be find in technology of location alignment ,This paper is research how to measurement space ,after Image segmentation to distinguish between background and test samples ,using two CCD and set up at the working platforms top side and front side ,Calculate the Parallelism and spacing of sample XYZ axis is same or not with the design value ,through the learning process of Artificial neural network ,let error rate will be close to design values as low as ,and aberrations problem of digital camera could be improved.
