基于三轴加速度传感器的山羊行为特征分类与识别

Classification and Recognition of Goats' Daily Behavior Based on Three-dimensional Acceleration Sensor

为分类并识别山羊的典型日常行为特征,以半封闭圈养的波尔山羊为研究对象,利用三轴加速度传感器对山羊的三轴加速度数据进行采集,利用K-means聚类算法对采集的数据集进行反复叠加训练得到较为稳定的聚类中心,同时利用全程的视频监控结合动作发生的时间识别并验证山羊的4种典型日常行为。结果表明,将三轴加速度传感器部署在山羊羊角处基本无明显应激反应,并且此传感器可以连续不间断地记录山羊的行为特征参数,对山羊的躺卧、站立或慢走、采食、跨跳等典型日常行为识别的准确率达87.76%,为山羊福利及山羊疾病预测模型的建立奠定了基础。

To classify and identify the typical daily behaviors and the characteristics of the semi-closed breeding circle Boer goats, using three-dimensional acceleration sensor for behavior acceleration data collection, and K-means algorithm as the experiment main clustering algorithm, the study extracted the clustering center by repeated superposition and training, and identified the goats＇ typical daily behavior using the video monitor installation and the time of the action. The experiment results showed no stress reaction to the installment of three-dimensional acceleration sensor in the goats＇ horn, and the sensor recorded continuously the goats＇ behavior parameters, recognition accuracy to the typical daily behavior （including lying, standing or walking, eating, and jumping） reached 87.76%, setting up basis for establishment of the goat diseases prediction model.