[Study on the Detection of Slight Mechanical Injuries on Apples with Hyperspectral Imaging]

Nondestructive detection is one of the hottest spots in the application of hyperspectral remote sensing. The apple is easy to produce slight mechanical injuries that affects its quality in the process of picking and transporting. The hyperspectral images of 54 “yellow banana” and “Yantai Fushi” apples with slight injuries in the visible and near-infrared (400~1 000 nm) ranges are acquired; the mean spectral curves of injury regions on apples are extracted; the endmember spectrum are extracted based on minimum noise fraction (MNF) and geometric vertex principle; and the spectral angle is calculated between spectral of injury region and endmember spectral; a model of endmember extraction spectral angle (EESA) is constructed to detect slight mechanical injuries on apples. The slight mechanical injuries on “yellow banana” and “Yantai Fushi” apples are detected by setting spectral angle threshold, and the detection accuracy is compared with MNF and principal component analysis (PCA) method. The results show that the accuracy of EESA model is the highest, and the detection accuracy rate reaches 94.44% and 90.07% respectively.