VALIDATION OF THE TOUCHING CORN SEPARATION USING IMPROVED CONVEX SEGMENTATION

Author

Ali Majeed Mahdi, Southern Illinois University CarbondaleFollow

Date of Award

5-1-2013

Degree Name

Master of Science

Department

Electrical and Computer Engineering

First Advisor

Chen, Ada

Abstract

Convex segmentation was first proposed to separate the touching corn kernels in an image. The algorithm show poor performance when the corn kernels are clustered touching each other. This problem was reduced by adding Watershed segmentation as a preprocessing step to break the clusters of touching corn into smaller groups, and then segment the small groups using the convex segmentation. This modified algorithm is called improved convex segmentation. The improved convex segmentation is tested and compared to the unmodified convex segmentation and the watershed method. The result of segmentation of the three algorithms shows that when the corn is aligned in double lines or randomly dumped, the improved convex segmentation accuracy is significantly higher than the unmodified convex segmentation and watershed. When the corn is aligned in single lines, the improved convex segmentation accuracy is significantly higher than watershed, but not higher than convex segmentation.