Detection of navel oranges canker based on hyperspectral imaging technology

Feature bands principal component analysis method and band ratio algorithm were proposed to fast differentiate citrus canker from normal fruit skin and other surface defects based on hyperspectral imaging technology. Navel oranges with cankerous, normal and other nine kinds of common defects were studied. First, region of interests (ROIs) spectral features of various defected peel areas were analyzed and combined with principal component analysis method to determined five optimal bands (i.e. 630, 685, 720, 810 and 875 nm). Next, principal component analysis was again performed based on feature wavelengths and the fifth principal component (PC-5) was used to classify and identify canker lesions on navel oranges. The overall classification success rate was 80% regardless of the presence of other confounding defects. In order to improve classification success rate, feature bands principal component analysis method and band ratio algorithm were combined to detect canker on the surface of navel oranges with an accuracy of 95.4%. The study results show that the hyperspectral imaging technology can be used to effectively classify and identify navel oranges with canker lesions.