Abstract: In order to recognize container code precisely and effectively, a recognition method of container code based on combined features is proposed. After preprocessing the segmented binary image of single character, the character features including hole, concaveconvex, leap, stroke, etc. are extracted. Then the treelike classifier is used to classify and recognize the container code, and combine the validation rules to verify the results. The method doesn’t need complex thinning procedure on the character image, improves the computational speed and avoids the character distortion. Experiment results show that the average recognition rate can reach 93%, and the recognition speed can reach 0.065 second per container code（including 11 characters）.

Key words: container code recognition, image preprocessing, feature extraction, tree-like classifier