Abstract: In order to explore a rapid and nondestructive recognition method to identify kiwifruits with surface defect, the hyperspectral imaging system was used to collect the hyperspectral image of Guichang kiwifruits with no defect and surface defect. After the minimum noise fraction transform (MNF), threshold segmentation and mathematical morphology, the recognition algorithm for kiwifruits with surface defect was proposed and then applied in 60 intact kiwifruits and 60 kiwifruits with surface defect. The results showed that the noise of the hyperspectral image of kiwifruits could be effectively removed by MNF transform. And the intact kiwifruits and kiwifruits with surface defect had obvious reflectance value between 700-810 nm and 810-1 000 nm. The spectral reflectance between 0.30-0.56 at 785.98 nm and the spectral reflectance between 0.54-0.73 at 982.59 nm were both selected as the threshold conditions to distinguish the region of normal and surface defect on kiwifruits. The correct identification rates for the intact kiwifruits and kiwifruits with surface defect reached 98.3% and 95.0%, respectively, which indicated that the hyperspectral imaging technology could be effective for identifying kiwifruits with surface defect.

Key words: kiwifruit, surface defect, hyperspectral imaging, nondestructive recognition