Image segmentation method of adherent corn ears based on improved watershed

Aiming at the problem of corn ears adhering in the process of corn test, the images of adherent corn ears were taken in the whole plot as our research object, and image segmentation method of adherent corn ears based on improved watershed was studied. We mainly proposed an improved watershed segmentation method based on limit corrosion and convex hull detection for common adherent corn ears, such as weak adherent corn ears, ring adherent corn ears, strong adherent corn ears and mixed adherent corn ears. The method can not only divide a variety of common sorts of the adherent corn ears, but also quantitatively calculate the lengths and diameters of the corn ears after the effective correction of the image distortion and the two-dimensional calibration of the image. Experimental results showed that the improved watershed segmentation method could achieve automatic segmentation for common adherent corn ears. The average accuracy rate of the ear length is 97.79%, and the average accuracy rate of the ear diameter is 96.70%. The accuracy rate of the whole system has reached 97.24%. This study provided an important technical basis for the measurement of characteristic parameters of adherent corn ears, and also an important solution to the segmentation problem of class cylinder.