White corn identification

White corn identification application purpose is to separate white corn kernels from yellow corn kernel.

The knowledge base consists of four color classes:

• Background
• Yellow kernel
• White germ
• White kernel

Discrimination between the white germ (present on each type of kernel) and white kernel is very difficult to make visually in the data base.

The spheroids containing 80% of the normal distributions are presented in the RGB space in the following graph:

Note the yellow kernel (yellow spheroid) and the blue background (blue spheroid), and the two white distributions (shades of gray).

Purclas was used to create linear classifiers. The following images were taken under different illumination conditions. The classified images are presented with the original images.

• Room illumination

  

• Ring lights (8x lenses;one ring on)

  

• Ring light (8x lenses;two rings on)

  

  The white kernels in the images are represented by a majority of blue pixels coding for the white kernel color class. These examples showed that white corn kernels can be separated from other kernels by color classification.

  The system set-up is described in the following figure:

  

  ---

  
  Up: Color Classification Examples Previous: Seed Corn Inspection

  ---