An automated panorama image construction for dimension inspection

Abstract

In hard disk drive industrials, one of the processes in quality control section is to measure the dimensions
of the hard disk drive components. To measure the dimension using high resolution visualization, we utilize a
microscope to acquire images of hard disk drive components. The component that we are focusing on is a bore
hole of pivot arm. Since dimension measurement is needed to be very accurate and the field of view of the
microscope is not wide enough to capture the whole part of the bore hole, we capture several parts of the bore
hole and reconstruct the whole part using a panorama image construction system. In particular, in this paper,
we propose to use a block matching and texture operator with majority voting filter for the translation image
stitching. The result of the panorama image construction system shows that an average root mean square error
is 1.374 pixels to reconstruct the whole parts from an image sequence containing 40 images. The reconstructed
bore hole image is visually correct as well.