The detection system for stored-grain live insects was introduced based on visible-near infrared computer vision, and the software realization of the main parts in the system was given. The near infrared images were sloped and distorted by the image registration based on projection transformation with markers. The region-growing method for identifying the live insects in the near infrared image was proposed based on the area ratio of two thresholds for connecting regions. The live insects in the visible image were located accurately by the information fusion of the multi-source images. The live insects in the visible image were located accurately by the information fusion of the multi-source images. Twenty-one global morphological features and seven local morphological features of the live insects were extracted, and the feature space was optimized to seven dimensions. The insects were classified by the classifier based on simulated annealing algorithm and support vector machine. The results showed that the recognition accuracy was 94.8% for the fifteen species of the live insects.