Abstract:Accurate identification of tea sprouts is a prerequisite to intelligent tea picking. As the image segmentation of tea sprouts under natural environment is greatly affected by weather, light and other factors, we proposed a segmentation method for tea sprouts based on SLIC super-pixel. Fourteen color components (R, G, B, H, S, V, Y, Cb, Cr, super-red, super-green, Cg, r-b, g-b) were extracted. The SLIC super-pixel segmentation algorithm was used to obtain the super-pixel blocks, and the average abscissa, average ordinate, average super-red, average Cg and average g-b were extracted from each super pixel block as the segmentation basis. Threshold segmentation, small target removal, filling and“logic and”operations was used to get the color images of tea. Experiments showed that the average accuracy of tea sprouts segmentation in different regions and environments based on SLIC super-pixel reached 75.6%, which was 16.6% higher than that of traditional G-B threshold segmentation. This method can not only counter the effect of light and other factors on tea images, but also effectively segment tea sprouts with robustness.