Rottenness is a prevalent and devastating disease that threats citrus fruit. Automatic detection of rottenness can enhance the competitiveness and profitability of the citrus industry. In this study, hyper-spectral image technology was used nondestructively to detect citrus rottenness. Spectral curve in defects peel region of interest was analyzed and combined with principal component analysis to extract the four best bands. Principal component was used based on four best bands: 615 nm and 680 nm, 710 nm and 725 nm peaks combination respectively and ultimately selected component (PC-2) as image classification and recognition obtained from the 615 nm and 680 nm principal component analysis and identification rate was 100% with a simple threshold segmentation. These results showed that using hyper-spectral as a kind of detection methods could be used for the evaluation of citrus rotteness recognition.