It is implemented as one of the leading algorithms in deep neural network, serving images processing purpose. CNN are inspired by biological visual nerves, in the visual producing procedure of animals, a single cortical neuron is responsible for only a small area of the field of view, this area will be stimulated like nerve impulses into the next layer of neurons. Different sensory areas of the optic nerve are different, covering the entire field of view through the superposition of all neuronal sensory areas of the fundus. Convolutional neural network neurons in the same layer process only a certain area of the image, the next layer of neurons can receive signals on a layer of neurons, so as to obtain a wider range of image information, more and more image information is obtained by superimposing multi-layer convolution. Compared with the traditional image classification algorithm, CNN requires less human intervention, so as to avoid the dependence of previous knowledge and the noise from human bias and omissions in the manual feature design. CNN consist of input and output layers, as well as multiple hidden layers. The hidden layer of CNN generally includes a convolutional layer, a Relu layer (activation function layer), a pooled layer, a fully connected layer, and a normalized layer. Compared with other deep learning algorithms, convolutional neural networks need to adjust fewer parameters, while having a greater performance, which makes it a commonly used deep learning structure [25].