Tuesday, November 14, 2006 - 1:30 PM

Image Analysis of Kernel Characteristics in Wheat and Rice.

Charles Erickson1, Harold Bockelman1, and Gilbert Sigua2. (1) USDA-ARS, NSGC, 1691 S 2700 W, Aberdeen, ID 83210, (2) USDA-ARS, USDA-ARS-STARS, 22271 Chinsegut Hill Rd., Brooksville, AR 34601

Several kernel characteristics are utilized as descriptors to phenotypically describe accessions in the National Small Grains Collection (NSGC).  In the past, these descriptors have been highly subjective (e.g. wheat kernel color and rice bran/hull color) or have been very time consuming and tedious to measure (e.g. thousand kernel weight).  This has been a slow, incomplete, and problematic process with tens of thousands of accessions in the collection still to be fully described.  NSGC and the Dale Bumpers National Rice Research Center are now using GrainCheck™ 2312 (Foss Tecator) video analyzer to measure kernel characteristics in wheat and rice.  This system is specifically designed for purity analysis of grain.  It is fully computerized and is conveniently operated using menu based commands and over a hundred samples can be evaluated in a normal work day.  Besides the purity analysis, the system images individual kernels and measures 13 kernel characters, generating a statistical analysis for each character.  The sample also is weighed and the number of seed counted, generating a kernel weight value.  Bran and hull color in rice and kernel color in wheat are measured by converting the RGB and intensity values from the analyzer into the appropriate descriptor color codes.  Kernel weight both in wheat and rice, and kernel length, width, and length/width ratio in rice, along with the color codes are recorded on the Germplasm Resources Information Network (GRIN).  This system provides a precise and fairly rapid method to phenotype the accessions in the NSGC.  In addition to the data from the video analyzer, voucher images of 47,225 wheat, 7955 rice, 6697 barley, 4061 oat, and 2137 Aegilops accessions are recorded on GRIN.  These images, taken with a flat-bed scanner, serve the same purpose as herbarium samples and give a visual voucher of the descriptor data.