A hyperspectral imaging system was developed for detecting various common defects on citrus. Citrus with end rot, insect dot damage, rot damage, thrip scars and normal citrus were chosen. Hyperspectral images of citrus samples and principal component analysis (PCA) were used to confirm two best wavelengths (615nm and 680nm). PC2 of PCA was selected to classify the images. Finally, the detection algorithm combined PCA and band ratio was developed and achieved an accuracy of 94%. The results showed the feasibility of the proposed method.