摘要
引入三轴加速度传感器对山羊的行为特征进行监测,按照三种不同的记录间隔时间(1 s、2 s、3 s)分批次获取山羊的三轴加速度数据。同时,利用全程的视频监控结合动作发生的时间分析山羊的3种典型日常行为。试验结果显示:1 s和2 s的记录间隔时间对山羊行为模型的分类精度分别为87.75%和86.73%,无统计学差异(P>0.05);而3 s的记录间隔时间相比较前两种记录间隔显著性差异明显(P<0.05),且分类精度明显下降为74.32%。总体来说,在不影响分类模型精度的前提下,记录间隔时间可延长至2 s,这样既可以减少数据样本量与数据处理的压力,同时也可以减少人为干预对动物正常行为的应激影响。
Tri-axial accelerometer could be used to monitor goats daily behavior, the devices collect the three- dimensional acceleration data based on pre-defined recording intervals (1 s, 2 s, 3 s). Meanwhile, the video-re- corded observations also be used to identify the goats typical daily behavior. The experiment results showed that dif- ferent recording intervals of 1 s and 2 s corresponding classification accuracy of 87.75% and 86.73% respectively in goats activity model, but the influences were not remarkable ( P 〉 0.05 ) ; However, classification accuracy in the 3 s recording interval was lower compared to the former two recording intervals, and the classification accuracy reduced to 74.32% , the influences were remarkable (P 〈 0.05 ). Overall, the time of recording interval can be extended to 2 s under the premise that didn't affect the accuracy of classification model, in this way, it can not only to reduce the pressure of data sizes and data processing, but also reduces the stress influence by the human inter- vention on animals normal behaviors.
出处
《科学技术与工程》
北大核心
2015年第2期254-258,共5页
Science Technology and Engineering
基金
国家高技术研究发展计划(863计划)(2012AA101905)资助