Based on the technical requirement in the field of stored-grain insect image recognition nowadays, aiming at the complex network structure and low recognition rate of the existing stored-grain insect image recognition algorithm, convolutional neural network is introduced to realize the image recognition of stored-grain insect. The development process of convolutional neural network is briefly introduced, its network structure is analyzed. Five kinds of stored-grain insects are selected as training samples. The process of image recognition of stored-grain insects is analyzed. The Alexnet model based on convolutional neural network is obtained by the test, which accuracy reaches to 97.62%. It shows that the image recognition of stored grain insects based on CNN has higher accuracy rate.