An algorithm of excising leafstalk while keeping its main body intact for leaf recognition

The machine recognition of leaf images is a fast and effective plant classification method. However, the existence of redundant leafstalks affects the recognition accuracy seriously and hinders the recognition of leaves on branches. Currently, most research avoid this problem and only Wang et al. proposed applying the opening operation to erase leafstalks. But as the experimental analyses show, the method is infeasible because it is difficult to determine the size of the structure element. Herein, we propose a new algorithm of excising leafstalk. Through the experimental verification, it can complete the accurate excision of the redundant leafstalk while keeping its main body intact.