Application of Convolutional Neural Network in Image Processing

Abstract

With the continuous development of computer technology, the wave of machine learning has become popular in various fields. Convolutional neural networks are based on the development of artificial neural networks and are widely used. How to optimize the algorithms of convolutional neural networks in image processing has become a research hotspot. There are many optimization algorithms for convolutional neural networks, but the results have not been ideal. In order to overcome the over-fitting phenomenon of convolutional neural network in image processing, and improve the accuracy of image classification and description conformity in complex scenes, this paper makes improvements based on the VGG model.First of all, this study chooses the VGG16 convolutional neural network model as the basic model, and introduces the Batch Normalization method to standardize the data. Then the Relu activation function with faster convergence speed is used to activate the neurons. Then use the improved dropout regularization method to optimize the model parameters. Finally, the improved model is tested by simulation experiments.The experimental results show that the error rate of the algorithm is as low as about 7% when iterates 300 times.Keywords: Convolutional Neural Network, Image Processing, Deep Learning, Image Classification, Dropout Regularization Method.