The detection system for stored-grain live insects was developed based on visible-near infrared computer vision. The system included an automatic sieving subsystem of insects, an automatic transporting subsystem of insects, an illumination box and an image acquisition subsystem. The insects could be rapidly and efficiently separated from the grain sample, and removed dust automatically in the automatic sieving subsystem of insects. The automatic transporting subsystem of insects could accurately receive the sieve material, and transport the collection box to the image vision acquisition part for image acquisition. The even illumination chamber could provide even visible and near-infrared diffuse light for the sieve material in the collection box. The image acquisition subsystem could simultaneously acquire the visual image and the near infrared image of the sieve material. The sieving accuracy of the detection system was 96.06% for the nine species of the most destructive live insects. The experiment showed that the system was practical and feasible.