Objective To achieve the automatic sorting of citrus fruits. A smart sorting system combining machine vision algorithms and an intelligent camera is designed.Methods The system is composed of a detection room, an intelligent camera, a mirror, a lighting system and rollers (including conveyor belts). The intelligent camera is equipped with a microprocessor, which can capture and process images without a dedicated computer. The lighting system consists of LED light strips positioned above the fruits, operating in stroboscopic mode synchronized with the camera to ensure uniform and stable illumination. For the captured fruit images, machine vision algorithms are applied to perform image segmentation and image labeling, achieving fruit size estimation, circle detection, color index estimation and local defect recognition.Results Experimental results based on a dataset of oranges collected from a certain orchard demonstrate that the proposed system achieves high prediction accuracy in fruit size and color index estimation.Conclusion The designed online smart sorting system can efficiently complete the sorting and detection of citrus fruits. Compared with other advanced computer vision algorithms, including automated grading systems based on the Raspberry Pi platform and nondestructive quality detection systems integrating machine vision and spectroscopy, the proposed system achieves higher accuracy in local defect classification.