CIGR VI 2019

講演情報

Oral Session

Postharvest/Food Technology and Process Engineering

[4-1600-C] Postharvest/Food Technology and Process Engineering (4)

2019年9月4日(水) 16:00 〜 18:15 Room C (3rd room)

Chair:Kornkanok Aryusuk(King Mongkut's University of Technology Thonburi, Thailand), Itaru Sotome(University of Tokyo, Japan)

17:15 〜 17:30

[4-1600-C-06] Segmentation of common scab lesion on intact potatoes using single near-infrared image

*Dimas Firmanda Al Riza1,2, Kazuya Yamamoto3, Kazunori Ninomiya3, Tetsuhito Suzuki1, Yuichi Ogawa1, Naoshi Kondo1 (1. Laboratory of Biosensing Engineering, Graduate School of Agriculture, Kyoto University, Kitashirakawa 6068267, Kyoto(Japan), 2. Department of Agricultural Engineering, Faculty of Agricultural Technology, University of Brawijaya, Jl. Veteran 65145, Malang(Indonesia), 3. Product Planning Department, Shibuya Seiki Co., Ltd. 2200, Minamiyoshida, Matsuyama, Ehime, 791-8042(Japan))

キーワード:image processing, common scab, external defects , near infrared, segmentation

Segmentation and identification of potato common scab lesion on a single near infrared image is the objective of this research. A more simple machine vision system is desired for sorting system application. Thus, a minimum number of the image is preferable compared to multispectral images for speediness if it can give good results. In this research, we proposed an imaging system using an InGaAs camera and a bandpass filter at 1600 nm which found to provide a good contrast of common scab lesion to normal potato skin. An image correction method was employed to deal with uneven light distribution due to the various shape of potatoes. Image segmentation has been successfully carried out with a Dice Sorensen coefficient of 0.72. Results also show that external defects, such as common scab and some mechanical damage types, and soil deposits appear brighter in the near infrared region, against the normal skin background. Thus, further image analysis of area shape and textural features has been provided to discriminate common scab and other classes. The results could be used as consideration for a real sorting application.