Accurate and fast identification of transgenic soybean plants by boosting methods with a handheld miniature spectrometer

Rapid and economical classification of transgenic soybean and non-transgenic soybean is highly important for food processing and handling. This paper developed an efficient and low-cost identification method for different categories of soybeans on the basis of a handheld miniature near-infrared spectrometer. The dataset consists of transgenic modified and non-transgenic soybeans from soybean breeders, and different pretreatment methods and classifiers are used to establish models. The identification model with the best performance is selected for the boosting models. After the data are compared by different pretreatment methods and classifiers, SG+SNV is the best, and the performance of the model constructed by the gradient lifting tree is optimized. The accuracy is 98.03 % and the F1 score is 96.74 %. The results show that the near-infrared spectrum can be used to collect the all-band spectrum of soybean, and the model can be used to classify the soybean category accurately, and quickly via a handheld miniature spectrometer.