Near-infrared spectroscopy (NIRS) is valuable in pesticide residue detection for fruits and vegetables due to its nondestructive, rapid, and efficient nature. This review summarizes current research status in this field, compares the applicability of different spectral bands and imaging techniques, and analyzes challenges in single and mixed pesticide residue detection alongside enhancement strategies. Additionally, approaches to spectral preprocessing and feature wavelength selection are examined, with performance comparisons between chemometrics and deep learning models. Furthermore, the role of calibration transfer and transfer learning is discussed in improving model generalization across instruments and crop varieties. With continuous advancements in spectral instrumentation and intelligent algorithms, NIRS is expected to transition from laboratory analysis to on-site monitoring, offering a green and efficient solution for agricultural product safety supervision.