Neural network parameter effects on object classification and wavelength selection

Floyd E. Dowell

doi:10.1117/12.173579

1 August 1994 Neural network parameter effects on object classification and wavelength selection

Floyd E. Dowell

Author Affiliations +

Optical Engineering, Vol. 33, Issue 8, (August 1994). https://doi.org/10.1117/12.173579

Abstract

A back propagative neural network (NN) was used to select visible spectrum (400 to 700 nm) wavelengths and classify damaged and undamaged peanut kernels. Results showed kernel classifications were best, network errors were minimized, and speed of convergence was greatest when the NN was set up with 20 or more hidden nodes, a momentum of 0.45 or less, and using about 1,000,000 learning events. Reflectance data in the 620 to 700-nm range were most influential in classifying kernels followed by relative reflectance in the 400 to 480-nm range. The learning rate did not affect NN performance, but higher learning rates converged more quickly. The most accurate classification performance occurred when the NN had 40 hidden nodes and a momentum of 0.45. These settings resulted in correct classification of 87.8% of all kernels. When compared to statistical means of classifying kernels using data from specific wavelengths or data from a colorimeter, the NN correctly classified about 5% and 13% more kernels, respectively.

Citation Download Citation

Floyd E. Dowell "Neural network parameter effects on object classification and wavelength selection," Optical Engineering 33(8), (1 August 1994). https://doi.org/10.1117/12.173579

Published: 1 August 1994

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available