基于无线传感器网络支持向量机奶牛行为特征识别

Cow Behavioral Features Recognition Using Binary Decision Tree Support Vector Machines Based on Wireless Sensor Network

鉴于我国现有奶牛发情鉴定主要依赖于人工观测、效率低和误判率高的现状,通过给奶牛佩戴嵌入加速度传感器的无线传感节点,实时监测奶牛的行为,并通过基于二叉决策树支持向量机时间序列模型分析加速度数据,逐层分类奶牛的静止、慢走、快跑和爬跨行为类型。实验结果表明,该算法对于奶牛轻微和剧烈的运动特征分类准确度达95.5%,研究结果为初步判断奶牛是否发情提供了有效的依据。

Nowadays,identification of cattle estrus conditions mainly relies on manual observation.This method is inefficient and has high error rate.A monitoring system based on wireless sensor node installed on a cow ＇ s neck was developed in this research.Movement acceleration of a single cow can be captured through an accelerometer on the node.The collected acceleration data are categorized into standing,walking,running and mounting behaviors using binary decision tree SVMs models.Experimental results indicated that the method can effectively distinguish behaviors such as slight and severe movement with the accuracy rate approached 95.5%.The results provided profound basis and efficient methodology for the cattle estrus identification modeling.