Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: Quality Based Inspection and Sorting of Specialty Crops Using Imaging and Physical Methods

Location: Healthy Processed Foods Research

Title: Detection of Mango Infested with Fruit Fly Eggs and Larvae by Infrared Imaging and Discriminant Analysis

Authors
item Haff, Ronald
item Saranwong, Sirinnapa
item Thanapase, Warunee
item Toyofuku, Natsuko
item Kasemsunran, Sumaporn
item Kawano, Sumio

Submitted to: United States-Japan Cooperative Program in Natural Resources
Publication Type: Proceedings
Publication Acceptance Date: July 7, 2010
Publication Date: August 27, 2010
Citation: Haff, R.P., Saranwong, S., Thanapase, W., Toyofuku, N., Kasemsunran, S., Kawano, S. 2010. Detection of Mango Infested with Fruit Fly Eggs and Larvae by Infrared Imaging and Discriminant Analysis. United States-Japan Cooperative Program in Natural Resources.

Interpretive Summary: Fruit fly infestation causes significant loss of perishable products around the world and is an economic threat to growers, processors, and exporters. A rapid, economical, and non-destructive technique for detection of fruit fly infestation is reported based on NIR imaging and computer classification algorithms. Five mango fruit were divided into 16 square regions into which fruit fly eggs were introduced. Five further mango were used as control samples. NIR images of each fruit were acquired 48 hours after introduction of the eggs. Near infrared spectra were extracted from the images in squares of 3 pixels by 3 pixels at the site of infestation for each region. An equal number of spectra were extracted from the images of control fruit. After spectral pretreatment, a computer program was used to classify the spectra as either infested or non-infested. Half of the samples in each class were used to train the algorithm and the other half were used for validation. The best classification results were 0.9% (4/432 spectra) false negative (infested samples classified as control) and 5.7 % (33/576 spectra) false positive (control samples classified as infested). This work clearly demonstrates the potential for hyperspectral imaging to detect fruit fly infestation.

Technical Abstract: Fruit fly infestation causes significant loss of perishable products around the world and is an economic threat to growers, processors, and exporters. A rapid, economical, and non-destructive technique for detection of fruit fly infestation is reported based on hyperspectral imaging and discriminant analysis. Five mango fruit were divided into 16 square regions into which fruit fly eggs were introduced. Five further mango were used as control samples. Hyperspectral images of each fruit were acquired 48 hours after introduction of the eggs. Near infrared spectra were extracted from the hyperspectral images in squares of 3 pixels by 3 pixels at the site of infestation for each region. An equal number of spectra were extracted from the images of control fruit. After spectral pretreatment, discriminant analysis was used to classify the spectra as either infested or non-infested. Half of the samples in each class were used to train the algorithm and the other half were used for validation. The best classification results were 0.9% (4/432 spectra) false negative (infested samples classified as control) and 5.7 % (33/576 spectra) false positive (control samples classified as infested). This work clearly demonstrates the potential for hyperspectral imaging to detect fruit fly infestation.

Last Modified: 10/30/2014