A Novel Intelligent Image Recognition Scheme based on Fully Convolutional Neural Network
N. Liu · 2023
As a deep feed-forward neural network, the convolutional neural network (CNN) model has achieved major breakthroughs in scheme of image recognition. Compared with the traditional classification methods (KNN, SVM, PSO), the fully automatic classification algorithms reply on convolutional network is separated from human identification, which is tedious, and at the same time backpropagation algorithm is used to automatically optimize the model parameters. Inspired by this, the paper proposes the novel intelligent image recognition scheme with the integrations of fully convolutional neural network. (FCNN). To solve the challenge that overall training accuracy of the general model is reduced due to the overfitting phenomenon caused by the commonality of patterns of the image features while training the models, this study introduced the AlexNet. The loss function is improved and optimized, and for preprocessing of the images, the compressive analysis model is integrated. Through the single and comparison tests, the performance of the designed algorithm is tested, and overall speaking, the recognition accuracy is better than the traditional methodologies.