To be able to forewarn rice borers and control the spraying pesticide dosage, to realize the nondestructive detection of rice borers’ damage. A feature band detection method based on principal component analysis and an optimal band detection method based on iterative threshold is proposed, the characteristic band and the optimal band of rice stem borers detection are determined, and the images of single band and the combined band are extracted to segment wormholes, to realize the accurate nondestructive detection of rice borers. Firstly, the reflectance information of 120 samples obtained by hyperspectral analysis determined that the spectral region was 450~1 000 nm. Band detection method based on principal component analysis characteristics, principal component analysis in the hyperspectral image, in which the first five principal components determine the third principal component images as the best image comparison, and then according to the third principal component in the image, the contribution rate of each band features to select wavelength (668.8 and 750 nm). Finally, global threshold segmentation and image masking are combined to distinguish the wormhole region. Moreover, utilization based on iterative threshold detection method, the optimal band in the visible band 450~750 nm range and near-infrared band 750~1 000 nm range application to pick the best single band, mixing distance by single band combination, a combination of single band and band to iterative threshold segmentation. Among them, 753.2 nm single band has the best segmentation effect, and 753.5 nm single band is determined as the optimal band. And then extract the band images using a wormhole extraction method based on iterative threshold and morphological processing. Finally, we can distinguish the rice stalk foraminifera region to realize the existence of rice stems infested with borers. The results showed that the detection rates of 60 pest-rice stalks and 60 normal rice stalks were 95.8% and 93.3% respectively, at 668.8 and 750 nm bands by using the principal component analysis-based characteristic band detection method. The optimal band detection method based on the iterative threshold has a detection rate of 96.7% at 753.5 nm band. This indicates that the optimal band detection method based on the iterative threshold is more accurate for the detection of rice borer and also indicates that the acquired characteristic band and optimal band provide theoretical reference for the future multi-spectral imaging technology of rice borers’ damage.