Comparison of a neural network and traditional classifier for machine vision inspection of potatoes

Research output: Contribution to journalArticle

26 Citations (Scopus)

Abstract

This work addressed the relative strengths and weaknesses of the backpropagation neural network versus the Fisher discriminant function. Their performance was compared for machine vision inspection of greening, shape, andshatter bruise in two potato cultivars. The backpropagation network's number of hidden nodes were varied from zero to eight for each defect type to determine the optimal network classification size. The network was trained and tested five times at each hidden node number and defect type to minimize local minima variation. For greening, the best backpropagation network averaged 74.0% with three hidden nodes while the Fisher method performed with a 70.0% accuracy. The backpropagation method also performed better for shape discrimination with a 73.3% average accuracy at seven hidden layer nodes versus a 68.1% accuracy. The Fisher method performed better for shatter bruise detection with a 76.7% accuracy versus a 56.0% average accuracy at four hidden layer nodes for backpropagation.

Original languageEnglish (US)
Pages (from-to)319-326
Number of pages8
JournalApplied Engineering in Agriculture
Volume11
Issue number2
StatePublished - Mar 1 1995

Fingerprint

Backpropagation
Computer vision
Classifiers
Inspection
Neural networks
Defects

All Science Journal Classification (ASJC) codes

  • Engineering(all)

Cite this

@article{7ae8c9274a204796b80d019e055af091,
title = "Comparison of a neural network and traditional classifier for machine vision inspection of potatoes",
abstract = "This work addressed the relative strengths and weaknesses of the backpropagation neural network versus the Fisher discriminant function. Their performance was compared for machine vision inspection of greening, shape, andshatter bruise in two potato cultivars. The backpropagation network's number of hidden nodes were varied from zero to eight for each defect type to determine the optimal network classification size. The network was trained and tested five times at each hidden node number and defect type to minimize local minima variation. For greening, the best backpropagation network averaged 74.0{\%} with three hidden nodes while the Fisher method performed with a 70.0{\%} accuracy. The backpropagation method also performed better for shape discrimination with a 73.3{\%} average accuracy at seven hidden layer nodes versus a 68.1{\%} accuracy. The Fisher method performed better for shatter bruise detection with a 76.7{\%} accuracy versus a 56.0{\%} average accuracy at four hidden layer nodes for backpropagation.",
author = "Deck, {S. H.} and Morrow, {C. T.} and Heinemann, {Paul Heinz} and {Sommer, III}, {Henry Joseph}",
year = "1995",
month = "3",
day = "1",
language = "English (US)",
volume = "11",
pages = "319--326",
journal = "Applied Engineering in Agriculture",
issn = "0883-8542",
publisher = "American Society of Agricultural and Biological Engineers",
number = "2",

}

Comparison of a neural network and traditional classifier for machine vision inspection of potatoes. / Deck, S. H.; Morrow, C. T.; Heinemann, Paul Heinz; Sommer, III, Henry Joseph.

In: Applied Engineering in Agriculture, Vol. 11, No. 2, 01.03.1995, p. 319-326.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Comparison of a neural network and traditional classifier for machine vision inspection of potatoes

AU - Deck, S. H.

AU - Morrow, C. T.

AU - Heinemann, Paul Heinz

AU - Sommer, III, Henry Joseph

PY - 1995/3/1

Y1 - 1995/3/1

N2 - This work addressed the relative strengths and weaknesses of the backpropagation neural network versus the Fisher discriminant function. Their performance was compared for machine vision inspection of greening, shape, andshatter bruise in two potato cultivars. The backpropagation network's number of hidden nodes were varied from zero to eight for each defect type to determine the optimal network classification size. The network was trained and tested five times at each hidden node number and defect type to minimize local minima variation. For greening, the best backpropagation network averaged 74.0% with three hidden nodes while the Fisher method performed with a 70.0% accuracy. The backpropagation method also performed better for shape discrimination with a 73.3% average accuracy at seven hidden layer nodes versus a 68.1% accuracy. The Fisher method performed better for shatter bruise detection with a 76.7% accuracy versus a 56.0% average accuracy at four hidden layer nodes for backpropagation.

AB - This work addressed the relative strengths and weaknesses of the backpropagation neural network versus the Fisher discriminant function. Their performance was compared for machine vision inspection of greening, shape, andshatter bruise in two potato cultivars. The backpropagation network's number of hidden nodes were varied from zero to eight for each defect type to determine the optimal network classification size. The network was trained and tested five times at each hidden node number and defect type to minimize local minima variation. For greening, the best backpropagation network averaged 74.0% with three hidden nodes while the Fisher method performed with a 70.0% accuracy. The backpropagation method also performed better for shape discrimination with a 73.3% average accuracy at seven hidden layer nodes versus a 68.1% accuracy. The Fisher method performed better for shatter bruise detection with a 76.7% accuracy versus a 56.0% average accuracy at four hidden layer nodes for backpropagation.

UR - http://www.scopus.com/inward/record.url?scp=0000355580&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0000355580&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:0000355580

VL - 11

SP - 319

EP - 326

JO - Applied Engineering in Agriculture

JF - Applied Engineering in Agriculture

SN - 0883-8542

IS - 2

ER -