Automatic ore screening is one of the key links to improve the efficiency of mineral resources processing. Aiming at the complex and changeable illumination in the ore mining site, it is difficult to detect the ore in the conveyor belt automatically in real time, and so on, a real-time 3D vision screening method based on infrared structure light was proposed. In order to solve the interference of solar radiation on the surface structure light imaging of ore under complex illumination, the infrared structure light near 800 nm was used as the active light source to obtain stable structural light images of ore surface under various illumination conditions. The method of neighborhood cumulative difference feature analysis was proposed. The center of structured light strip was extracted by locating the boundary point of strip quickly, and then 3D coordinate data of ore could be obtained in real time. The results show that the proposed method can better adapt to the changes of the shape and direction of the line structure light and the stability of the method. The time for screening of light images of each ore structure is 13.2 ms, which meets the 3D real-time screening requirements of the ore on site.