부산대학교 도서관

As the physical sizes of many electrical devices and their components shrink to sometimes only a few microns, expanding dextrousness and obtaining rapidity ofa precise micro-manipulation system become one of the major issues in the industry. Especially for an assembly process that handles complicated mechanical alignments of many tiny objects in the three-dimensional space, a proper identification of the alignment errors and a fast decision of manipulation plan should be guaranteed in order to keep the manufacturing costs in an affordable range. In most of the current precise manufacturing processes, identification of the misalignment is normally carried out by machine vision systems. And the limited data process speed of the vision system usually hampers enhancement of overall system performance. In the present work, a modified binary region median filter combined with a labeling based segmentation scheme for the enhancement of a machine vision system is introduced. The suggested algorithm shows not only shows improved data processing speed but provides a better accuracy on the identification of a target object.