Fusing Local Shallow Features and Global Deep Features to Identify Beaks

• DOI: 10.3390/ani13182891
• Publication Date: 2023-09-12T08:38:48Z
• Authors: Qi He; Qianqian Zhao; Danfeng Zhao; Bilin Liu; Moxian Chu
• Keywords: Local Binary Patterns; Feature (linguistics); Identification; Binary classification

Cephalopods are an essential component of marine ecosystems, which great significance for the development resources, ecological balance, and human food supply. At same time, preservation cephalopod resources promotion sustainable utilization also require attention. Many studies on classification cephalopods focus analysis their beaks. In this study, we propose a feature fusion-based method identification beaks, uses convolutional neural network (CNN) model as its basic architecture multi-class support vector machine (SVM) classification. First, two local shallow features extracted, namely histogram orientation gradient (HOG) binary pattern (LBP), classified using SVM. Second, multiple CNN models were used end-to-end learning to identify performance was compared. Finally, global deep beaks extracted from Resnet50 model, fused with features, The experimental results demonstrate that fusion can effectively fuse recognize improve accuracy. Among them, HOG+Resnet50 has highest accuracy in recognizing upper lower 91.88% 93.63%, respectively. Therefore, new approach facilitated