|LEE, HOONSOO - Chungnam National University|
|JEONG, D - Chungnam National University|
|Delwiche, Stephen - Steve|
|Chao, Kuanglin - Kevin Chao|
|CHO, BYOUNG-KIWAN - Chungnam National University|
Submitted to: Journal of Food Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/24/2014
Publication Date: 10/10/2014
Citation: Lee, H., Kim, M.S., Jeong, D., Delwiche, S.R., Chao, K., Cho, B. 2014. Detection of cracks on tomatoes using hyperspectral near-infrared reflectance imaging system. Journal of Food Engineering. 14(10):18837-18850.
Interpretive Summary: The consumption of fresh produce has recently been linked to several food-borne outbreaks. Fresh tomatoes that exhibit surface cracks may present both safety and quality problems for producers and processors. Cracked fruit can be more vulnerable to entry by and establishment of pathogenic bacteria, more perishable, and less desirabile to consumers in the fresh produce market. Consequently, a rapid imaging technique to differentiate fruit not suitable for the fresh market (but suitable for processing) would be useful. Hyperspectral images were acquired on 240 intact mature tomatoes; half of these tomatoes exhibited cuticle cracks near the stem-scar area and half did not. The images included spectral data across the near-infrared (NIR) region. Images were analyzed to determine a pair of two-band ratios suitable for development of a crack-detection algorithm based on PC images and ratio images. By quantifying tomato stem-scar shapes using two morphological features (roundness and size), calculated from the PC and ratio images, the algorithm demonstrated over 94% accuracy in distinguishing tomatoes with and without cracks. This study shows that screening of whole tomato fruits for cracks could be implemented using hyperspectral near-infrared reflectance imaging systems, in conjunction with other tomato safety or quality inspection tasks. The findings will benefit food technologists, the fresh produce industry, and regulatory agencies interested in the development of automated screening and classification methods for food safety and quality.
Technical Abstract: The objective of this study was to evaluate the use of hyperspectral near-infrared (NIR) reflectance imaging techniques for detection of cuticle cracks on tomatoes. A hyperspectral near-infrared reflectance imaging system in the region of 1000-1700 nm was used to obtain hyperspectral reflectance images of a total of 240 tomatoes, half with cracks and the other half without cracks along the stem-scar regions. For classification and detection of cracks on tomatoes, hyperspectral images were subjected to principal component analysis and iterative ANOVA of two wavelength ratios to determine the most suitable ratio pair to create principal component (PC) score and ratio images, respectively, to classify tomato samples. Two morphological features, roundness and minimum-maximum distance, to quantify the stem scar shapes were calculated from the selected PC and ratio images and were used in discriminant analysis for tomatoes with and without cracks. Results demonstrated 94.4% and 91.7% accuracies for PC and ratio image based classifications, respectively, for tomatoes with and without crack defects. Hyperspectral near-infrared reflectance imaging could be potentially used for detection of crack defects on tomatoes, in addition to the traditional NIR spectroscopy-based quality assessments of the fruits.