Dynamic timber cell recognition using two-dimensional image measurement machine

Abstract

Image motion blur and defocus blur often occur when there is a relative motion between the imaging camera and the detected object. In this paper, we propose a robust timber cell recognition scheme using the low quality color timber cell images with the above-mentioned image blurs. First, a novel two-dimensional image measurement machine is devised, to obtain the object images sequentially by using a color camera. Second, the image-moment-based blur invariant features are calculated. Third, timber cell recognition is performed by using the computed Euclidean distance based on the moment invariants. We have experimentally proved that the effective use of image blur information improves the recognition accuracy of camera-captured timber cells. Moreover, the allowed maximum translation speed of the moving gallery is also discussed theoretically and experimentally. This scheme can identify the timber species by means of the cell recognition so as to judge the physical property and economic value of different timber species correctly.