基于改进卷积神经网络的羊行为识别

Sheep behavior recognition based on improved convolutional neural network

本文根据羊不同行为的特征,提出一种基于改进卷积神经网络的羊行为识别方法。构建卷积核尺寸全部为3×3的卷积神经网络(Convolutional Neural Networks,CNN);使用缩放指数线性单元(scaled exponential linear units,SeLU)为激活函数,使网络具有自归一化功能;以最大池化(max pooling)为下采样;在全连接层中采用丢弃(Alpha dropout)操作提高网络泛化能力,使用余弦退火动态学习率进行动态微调;进一步使用softmax分类器作为网络输出,最终构建出羊行为识别网络模型。实验结果表明:本文方法对羊进食行为识别准确率达到90.30%,站立行为识别准确率达到94.16%。坐卧行为识别准确率能达到91.90%。该模型能够实现羊不同行为的监测,且有较高的准确性,有助于提高畜牧管理效率和养殖智能化水平。

According to the characteristics of different behaviors of the sheep,a sheep behavior recognition method based on improved convolutional neural network is proposed in this paper.Firstly,the paper constructs convolutional neural network(CNN)with convolutional kernel size of 3×3,and using scaled exponential linear units(SeLU)as the activation function,the network has the function of self normalization.Then,max pooling is used as the down sampling.After that,in the full connection layer,the Alpha dropout operation is used to improve the generalization ability of the network,and the cosine annealing dynamic learning rate is used for dynamic fine-tuning.Finally,the softmax classifier is used as the network output,and the sheep behavior recognition network model is constructed.The experimental results show that the accuracy of this method for sheep eating behavior recognition is 90.30%,and the accuracy of standing behavior recognition is 94.16%.The recognition accuracy of sitting and lying behaviors reaches 91.90%.The model could monitor different behaviors of the sheep with high accuracy,which is helpful to improve the efficiency of animal husbandry management and the level of intelligent breeding.