基于深度学习的野生动物识别  

Wild life Recognition Based on Deep Learning

作　　者：黄志静 邵慕义 张庭瑞 沈嘉轶 Huang Zhijing;Shao Muyi;Zhang Tingrui;Shen Jiayi(Computer School,Beijing Information Science and Technology University,Beijing,100101)

机构地区：[1]北京信息科技大学计算机学院,北京100101

出　　处：《电子测试》2022年第22期69-71,10,共4页Electronic Test

基　　金：北京信息科技大学大学生创新创业训练计划项目-计算机学院(5112210832)支持。

摘　　要：为了更好地保护野生动物以及动物基因库的种类,保障生物链的完整性。运用深度学习技术对野生动物的图像进行识别,并且为了降低噪声信息的干扰及提高野生动物图像识别的准确率,提出了基于深度残差收缩网络的野生动物识别模型。目的是可以更好地帮助社会对野生动物进行监管和保护。该模型在深度残差网络的基础上融入注意力机制和软阈值函数,从而降低噪声信息的干扰,提高图像识别的准确率。将深度残差收缩网络与深度残差网络模型对相同野生动物数据集进行训练作对比,同时对部分野生动物图像进行了测试。实验结果表明,深度残差收缩网络提高野生动物图像识别准确率。In order to better protect wild animals and the types of animal gene banks, the integrity of the biological chain is guaranteed. Deep learning technology is used to identify wild animal images,and in order to reduce the interference of noise information and improve the accuracy of wildlife image recognition, a wildlife recognition model based on deep residual shrinkage network is proposed. The aim is to better help society regulate and protect wildlife. Based on the deep residual network, the model incorporates the attention mechanism and soft threshold function, so as to reduce the interference of noise information and improve the accuracy of image recognition. The deep residual shrinkage network and the deep residual network model were trained on the same wildlife dataset, and some wildlife images were tested. Experimental results show that the deep residual shrinkage network improves the accuracy of wildlife image recognition.

关 键 词：深度残差收缩网络 图像识别 野生动物 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]