Massive spectral data analysis for plant breeding using parSketch-PLSDA method: Discrimination of sunflower genotypes

In precision agriculture and plant breeding, the amount of data tends to increase. This massive is becoming more complex, leading difficulties in managing analysing it. Optical instruments such as NIR Spectroscopy or hyperspectral imaging are gradually expanding directly field, increasing spectral database. Using these tools allows access non-destructive rapid measurements classify new varieties according breeding objectives. Processing this challenging. a context genotype discrimination, we propose apply method called parSketch-PLSDA analyse data. ParSketch-PLSDA combination an indexing strategy (parSketch) reference (PLSDA) for predicting classes from multivariate For purpose, database was formed by collecting 1,300,000 spectra generated images leaves four different sunflower genotypes. compared PLSDA. Both methods use same set calibration test. The prediction model obtained PLSDA has classification error close 23% on average across all outperforms greatly improving qualities 10%. Indeed, built with ability take into account non-linearities among sets. These results encouraging allow us anticipate future bottleneck related generation large phenotyping.