基于无线传感器网络的产前母猪行为监测系统

摘要

目前大规模集约化母猪养殖业主要是依靠饲养员对产前母猪行为的连续观察,通过直觉和经验从而判断母猪的分娩时间.但是该技术不仅要求大量人力,效率低下,而且往往会出现人为疏忽造成刚分娩的仔猪死亡现象.针对这一问题,设计了一种基于无线传感器网络的产前母猪行为特征监测系统.该系统通过给母猪佩戴加速度传感器采集节点,实时监测母猪的行为,并通过支持向量机分析加速度数据并建立模型,分类识别母猪的站立、躺卧、进食、筑巢等典型行为特征.试验结果表明,系统能够实时连续的记录母猪产前行为特征参数,可准确识别出母猪的四种典型行为,准确率为93.3％.%According to their intuition and experience,feeders can judge sows' delivery time based on continuous observation of sows' behavior before farrowing,which is a main feature of current large scale intensive pig breeding industry.But it is inefficient and requires lots of manpower,and piglets just born would die due to feeders' negligence.To solve this problem,the monitoring system for sows' behavior features before farrowing based on wireless sensor networks is designed.The system can monitor the behavior of sows in real-time by acceleration sensor node wore on sows,and establish the model of classification through the analysis of acceleration data with SVM to classify sows' typical behavior features including standing,lying,eating and nesting.Experimental result shows the system can real-timely and continuously record sows' behavior features,and can detect four kinds of sows' behavior accurately and the accuracy is 93.3％.