 This study proposes a denoising pre-processing algorithm, CMD-NSR, to improve the accuracy and robustness of nitrogen content prediction models in citrus leaves using hyperspectral technology. The results show that the CMD-NSR algorithm effectively reduces the loss of effective information and can be used as a reference for accurate fertilization of citrus trees. This article was authored by Chang Luan Gao, Ting Tang, Wei Bin Wu, and others.