本文旨在開發一小型／微型物件辨識系統，藉由此系統求出待測零件所屬模型影像及其對應點，以便於後續之微自動化插配、組裝工件之進行。因在多重微物件下之影像，各微物件所佔之像素點數相對而言少很多，物件外形嚴重失真，不利於辨識。因此，為提高影像品質，以獲得較佳之辨識效果，本系統以六角格子影像取代直角格子影像，並建立六角格子影像之辨識系統。為使系統能辨識具多重物體之影像，影像經自動二值化處理後以元件標記的方式分割影像中之各物體，並求出各物件之輪廓外形，再依此一外形找出輪廓特徵點作為比對依據。藉由整合式遺傳演算法與類神經網路求出模型影像在待測影像中之特徵點對應解，並以姿勢群集法除去偽特徵點對應。

The purpose of this paper is to develop a mini/micro object recognition system. According to this system, we can find the position of the part in the image. It can be used as the preprocessing of the micro automotive handling and assembly. On a multiplicate mini/micro object image, the contour of an object may consist of only a limited number of pixels, so it is not easy to find out the corresponding contour feature accurately. In our method, we will replace the rectangular grid with the hexagonal grid for better angular resolution and image quality, and will develop a hexagonal-based object recognition. In order to recognize the images with multiplicate objects, we use a connective component labeling method following a thresholding image process to segment all objects in the image and to find the contours of the objects. By combining the use of the genetic algorithm and the neural network, we solve the corresponding matching pairs between the model graph and the scene graph. Then, pose clustering method is performed to delete the spurious matching pairs.