Feasibility of conventional and Roundup Ready® soybeans discrimination by different near infrared reflectance technologies

► The feasibility of NIRS was tested to discriminate GM glyphosate tolerant soybean seeds. ► Three NIRS technologies and two classification algorithms were compared (ANN and LWR). ► Overall accuracies were around 75% when validating with new seeds from new samples. ► Validation with new seeds from samples in the training set gave a maximum 94% accuracy. ► The previous best accuracy was achieved by the light tube LWR model.