초분광 반사광 영상을 이용한 배추 종자(Brassica campestris L) 비파괴 품질 측정기술 개발

Cabbage (Brassica campestris L) is an important crop for Asian countries, and especially so for Korea, Japan and China. In order to achieve uniform and high-yield rates of cabbage product, seed lot quality needs to be controlled. Non-destructive evaluation of seed viability is an important technique for investigating seed quality. Hyperspectral imaging technique, which combines the features of imaging and spectroscopy, has been considered one of the most powerful nondestructive evaluation methods allowing comprehensive analysis of the physical and biochemical characteristics of materials. In this study, the feasibility of hyperspectral reflectance imaging technique was investigated for the evaluation of seed viability. For the investigation of viable and non-viable seeds, some viable seeds were artificially aged. Hyperspectral reflectance technique was used to discriminate aged cabbage seeds from normal seeds. The PLSDA and simple image threshold methods were applied to investigate the feasibility of distinguishing the aged seeds from the normal seeds. The discrimination accuracy was 96.7% for the calibration set and 96.9% for the test set. The resultant images from the PLS-DA method showed high classification performance in distinguishing the nonviable from the viable seeds, which is an impossible task by naked eye and by conventional color cameras. Hyperspectral reflectance imaging has good potential for discriminating nonviable cabbage seeds from massive amounts of viable cabbage seeds.