Detection of rust in citrus by hyperspectral imaging technology and band ratio algorithm

Hyperspectral imaging technology was attempted to detect rust in citrus in this study, and band ratio algorithm was proposed to overcome the adverse effects of uneven reflectance intensity due to curvature of spherical objects. First, Sheffield Index was used to determine two optimal bands (i.e. 625 nm and 717 nm), and the first ratio image was obtained by ratio transformation between them. Next, the optimal band with 625 nm and its neighbor band with 621 nm were performed to ratio transformation, and the second ratio image was obtained to build the mask. Then, the background noise of the first ratio image was removed by the mask. Finally, rust features on the surface of citrus were extracted by threshold segmentation and morphological image processing. The experimental results show that the rust in citrus can be detected with an accuracy of 92% by hyperspectral imaging technology and band ratio algorithm. This work demonstrates that band ratio algorithm was able to effectively reduce the adverse effects of uneven reflectance intensity, maximize the differences between bands, and improve the performance in detection of rust in citrus by hyperspetral imaging technology.