Cereal Chem. 73 (1):136-142 |
Analytical Techniques and Instrumentation
Image Texture Analysis for Discrimination of Mill Fractions of Hard and Soft Wheat.
I. Y. Zayas (1,2) and J. L. Steele (1). (1) Electronics Engineer and Research Leader, U.S. Grain Marketing Research Laboratory, USDA, ARS, Manhattan, KS 66502. (2) Corresponding author. Accepted September 15, 1995. This article is in the public domain and not copyrightable. It may be freely reprinted with customary crediting of the source. American Association of Cereal Chemists, Inc., 1996.
Digital image texture analysis was utilized to identify mill fractions from different mill streams and to assess wheat hardness differences. The study was conducted using a soft red winter wheat (Terra SR-87) and a hard red winter wheat (Thunderbird). Black and white images were acquired in a 256 × 256 pixel format to examine samples of coarse and fine mill fractions. Sixteen 64 × 64 pixel subimages per image were evaluated using texture analysis. Software was developed to calculate the image textural features used to develop the mill stream and hardness classification models. Several models based on image textural features were computed for different sets of subimages belonging to wheat of different hardness or mill stream. Recognition of hard wheat vs. soft wheat was achieved with 100% correct recognition rate for each mill fraction when a three-feature model was used for pairwise analysis. Different mill fractions of the same wheat, coarse vs. fine, were similarly discriminated with 100% accuracy for each pairwise comparison. All four mill fractions were successfully recognized with 100% correct recognition rate when a three feature model was used for four class analysis. The wheat class and mill fraction discrimination was achieved with <3 g (approximately 0.2 g/subimage) of material.