Wood cell recognition using geodesic active contour and principal component analysis

In this paper, we propose a robust wood cell recognition scheme using color wood cell images. First, a novel 2-D cell image collection system is devised, and the wood cell images are segmented by using a dual-threshold segmentation algorithm. Second, a geodesic active contour (GAC) is applied in the segmented binary image to extract the edge contours of multiple cells simultaneously. Third, wood cell recognition is performed based on the Mahalanobis distances calculated by using the principal component analysis (PCA) algorithm. We have experimentally proved that this scheme improves the recognition accuracy, which can efficiently discriminate the intraspecific cell's shape variation and the interspecific cell's shape variation.