Development Of Near-infrared Reflectance Spectroscopy Equations Of Fatty Acid Composition, Glucosinolates & Oil Content

Near infrared reflectance spectroscopy (NIRS), has been successfully applied as anon-destructive, rapid, and cost effective plants quality analysis technique. It is verysuitable for the plants breeding choice, and it is also an important technique for theplants breeding choice.The national oil-crop improvement centre in Hunan have beenbought a FOSS NIRsystem 5000 with original equations for 5 yeas; the equationshave never been modified and maintained, and also no new equations were developedfor this instrument during those period. So the results which were produced by theinstrument were questioned by the users.A total of about 670 samples of Brassica napus cultivars were collected fromdifferent ways. The Fatty acid composition of 557 samples were determined by thenational standard analysis method, 255 samples were analyzed for it's oil content and90 for glucosinolate. For instead the formers, new equations were developed from thesamples which were analyzed.The R squared (RSQ) values of oleic acid and erucic acid were 0.98(ideal valueshould be approaching 1). The 90% of total samples have a absolute error limit lessthan 4.25, and less than 5.83 for erucic. It has demonstrated an obvious advance to theformer equations, which has a absolute error less than 16.90 and 17.61 in 90%samples for oleic acid and erucic acid.A separated equation were developed for the low content of erucic acid, it showsa good ability to determine the sample's erucic acid content which has a actual valuelower than 4%. The average error between actual values and predicted values is lowerthan 0.5.The values of RSQ and 1-VR for the oil content equation are 0.97 and 0.96.The90% of total samples have an absolute error limit less than 1.26, and the relative erroris less than 3.18%.The values of RSQ and 1-VR for the glucosinolate content equation are 0.99 and0.97.The 90% of total samples have a absolute error limit less than 7.62, and therelative error is less than 21.78%.Two more conclusions were obtained from the development of equations. First,it is a good idea that calibrate separately for the extreme concentration ranges of oleicacid or erucic acid, but not good for the medial ranges. When we develop a separateequation for low erucic acid content samples, the samples which erucic acid content isabout 10% should be included, and the samples which have no erucic should belimited. Otherwise, the all values which predicted by the equation will be too lower.Second, the out range predict ability of the oil content equation is very will, thesamples, which actual value 3% over the max value of the calibration samples ,willpredicted very will with a average absolute error what will limited in 1.1. The averageabsolute error will limited in 2.5 when a sample was predicted by the oleic equation,even the sample's actual value is 5% over the max value of calibration samples. Butwhen the sample's actual value is lower than the limit value of the calibration samples,the result will be bad. Even though, we can know where the actual value tend to.