中文摘要：

准确获取母羊临产前行为活动方式及其规律对预测母羊分娩时间、判断其在分娩时是否需要提供人工助产具有重要的意义。目前,在大规模集约化养殖环境下大多依靠饲养员对母羊产前行为进行连续观察,耗时耗力,且主观性较大。为此,设计了一种基于可穿戴数据检测装置的母羊产前行为特征实时监测系统。该装置以单片机STM8为主控芯片,采用MPU6050三轴加速度传感器获取母羊产前行为数据,通过Wifi无线传输模块USR-C2 1 5将数据传输至上位机。利用K均值聚类算法对母羊产前3种行为（趴卧、行走、站立）进行特征分析与识别。试验表明：系统可以快速、准确的获取母羊产前行为的加速度信息,并对母羊的3种产前行为能够正确的分类识别,识别率为82.69%,能够满足母羊产前行为识别的要求。

英文摘要：

Acquiring the pattern and regularity of ewe＇ s behavior prior to delivery in real time has a great significance in predicting the delivery time of ewes and determining whether it needs to provide an artificial midwifery at birth. Presently,the ewe＇s behavior prior to delivery has been continuously observed by breeder under large-scale intensive farming envi- ronment, which not only time-consuming and labor-intensive but subjective. So a monitoring system of ewes＂ behavior prior to delivery based on a wearable data detection device was designed in this paper. In the designed device, STM8 the chip microcomputer was used as the core control unit and the 3-axis acceleration sensor MPU6050 was employed to de- tect the data of ewes＂ behavior prior to delivery, then the data was transmitted to the PC through the WiFi wireless trans- mission module USR-C215. The K-means clustering algorithm was adopted for analyzing and identifying the ewe＇ s three kinds of behaviors including lying, walking as well as standing. The tests showed that the system can acquire and transfer the ewe＇ s activity information in real time, and can correctly classify the behaviors prior to delivery. The accuracy can reach 82.69% , satisfied with the requirements of behavior identification for the ewes prior to delivery.