Research on Deep Learning-based Image Processing and Classification Techniques for Complex Networks

Image classification task is a fundamental problem in the field of computer vision. With the rapid development of the Internet and artificial intelligence technology, a large amount of image data is generated every day. In this paper, for the problem of invalid feature information generated in the process of semantic segmentation of images, and the loss of local detail information of images, the paper proposes an encoder based on DCNN, ECANet and DSA_ASPP. Based on the above encoder, an image classification algorithm based on lightweight and multi-scale attention fusion is proposed.After analyzing and comparing the commonly used image feature extraction algorithms, SIFT features are used as the nodes of the image feature network and the commonly used similarity measures are analyzed, and the correlation coefficients are used as the weights of the connected edges in the network.The average intersection and concurrency ratios reach 69.6% and 73.6%, respectively. Compared to the existing state-of-the-art network models, the detection performance of this paper’s method is better, which can effectively capture local detail information and reduce image semantic pixel classification errors. Finally, the performance of PreactResNet on two benchmark datasets, CUB-200-2011 and Stanford Dogs, outperforms the existing network image performance.