The images of tea leaf in Qingming period were taken by digital camera for extracting tea color information. Firstly, the color information of G and G-B component of tender leaf and old leaf based on RGB color model was analyzed. The color difference between tender leaf and old leaf was able to distinguish the tender leaf from background. Then the original threshold from color analysis was set and the segmentation thresholds for G and G-B component were calculated by improved Ostu (the algorithm of threshold automatically extracted according to the maximum deviation). Finally, the recognition algorithm of tea tender leaf was proposed. Experimental results showed that the algorithm is useful for eliminating the impact of light and rapid in identifying the tender leaf from background. The accuracy rate is more than 92% when the distance between camera and tea bush is about 10βcm. The research results were useful to provide recognition technical supports for an intelligent picking machine development.