Due to the genetic diversity in agricultural products, especially in the case of cereals, it is difficult to identify the plantʹs genetic basis. Also, collecting comprehensive information from different genotypes of plants during their growth and development is of great interest to the experts in the field of plant genetics. Therefore, the development of a non-destructive method that can quickly and easily identify the plant genotype has received much attention. In this research, the ability of visible and infrared spectroscopy (Vis/NIR) technology to detect different wheat genotypes was eva‎luated. Five different wheat genotypes including two bread wheat genotypes, one durum wheat genotype (T. durum), and two wheat ancestors were selected for experiments. Bread wheat genotypes included common wheat (Roshan) and Synthetic sample 159 from the hexaploid species (T. aestivum), and ancestors wheat genotypes included diploids (T. monococcum) and tetraploids (T. turanicum). A PDA spectrometer equipped with a CCD detector with a resolution of 2 nm in the range of 400-1100 nm was used to acquire the leaf spectra of wheat samples. The spectrometer was equipped with bifurcated fiber optics, a sample holder, and a halogen light source. The wheat genotypes were cultivated in three main stages of wheat plant growth including leaf opening (approximately 7 open leaves) (first stage), tillering (second stage), and flowering (third stage). For each wheat genotype at each growth stage, 50 plants were randomly selected. One leaf was chosen from each plant and three interactance spectra were taken from the selected leaf. Then, the average spectrum was calculated and used as the spectrum of that genotype in the next steps. Thus, for each wheat genotype at each growing stage, 50 average and independent spectra were obtained. To analyze the obtained spectra, the beginning and end of the spectra were removed due to noise. Different pretreatment methods were then applied to the spectra. Soft and independent modeling of class analogy (SIMCA) and artificial neural networks (ANN) were used to identify different genotypes. In each of these two methods, in the first step, different wheat genotypes were discriminated without considering the growth stages. Then, the ability of each method to detect growth stages and discrimination of genotypes at each growth stage was eva‎luated. For discrimination of different wheat genotypes without considering the growth stage, the best performance of SIMCA method was obtained by first derivative pre-processing, which resulted in a total accuracy of 89.45%. While the best ANN model was able to classify different wheat genotypes without considering the growth stages with a total accuracy of 98.49%. In different growth stages, the best SIMCA and ANN models resulted in total accuracy of 95.48% and 100%, respectively. In this case, SIMCA method was able to distinguish different wheat genotypes in the first, second, and third stages of growth with total accuracy of 93.88, 98, and 94.06%, respectively. The ANN method was also able to classify different wheat genotypes at all three stages of growth with 100% accuracy. In general, the results of this study revealed that Vis/NIR spectroscopy technique combined with the nonlinear ANN method is a powerful technique in detecting different wheat genotypes during growth, detecting different growth stages, and also discriminating wheat genotypes at each growth stage.

احمد ميره اي، عباس همت

محمد مهدي مجيدي

اسماعيل مهريار، محمدرضا سبز عليان