Computer digital image processing for wheat variety identification

When the existing digital image analysis system for wheat variety identification in the Department of Grain Science and Industry at Kansas State University was evaluated as the part of the study, the program was incapable of discriminating between wheat varieties grown at different locations and in different crop years. Significant interactions among crop years, locations and varieties were observed in 11 out of 20 parameters used as discriminators. Only one parameter (KR2) was able to discriminate between wheat varieties. The bugs in one or more computer programs used in processing image data caused significant variations in calculated parameters. The new digital image processing system developed for wheat variety identification as the second part of this study showed no instrumental or programming errors. About 90 percent of wheat samples used in the study were correctly identified by the developed program. Some parameters (BA1, BA2, BA3, RCENL, GAREA5, GPERI5, GLENG5) demonstrated good characteristics of discriminating variables. Further, a 20% reduction of the top lighting intensity during image acquisition caused significant reductions in within variety variations for germ parameters. The parameters employed in the new program were not affected by interactions between variety, location, or crop year. A significant reduction of within variety variation was observed when the upper part of the germ was not used for feature extraction. A semi-automated, single kernel positioning device, developed as part 3 of the study, delivered about 90% of wheat kernels to the final point in a crease down position.