Abstract: Accurate cell identification is based on accurate sampling of cell images．This paper put forward that using the concept of image moments samples cell image and extracts mathematical features of wood cells based on rotational and translation invariance．We first grayed colored images and splited cell images with dynamic threshold segmentation method preliminarily,then implemented segmentation process for adherent cells with adherent cell segmentation method based on image thinning,finally extracted sample cells with the concept of image moments and completed the sampling work．The experiment proves that this algorithm has a good stability and robustness.

Key words: Image moments,Adherent,Segmentation,Sampling