支持向量机在舍饲肉牛反刍行为分析中的应用

发布时间：2018-09-03 08:32

【摘要】：为进一步加强对反刍类家畜的规范化饲养,合理利用资源,提高生产品质,研究其日常行为规律便成了关键因素。舍饲肉牛的反刍行为是反映肉牛健康状态的重要指标之一。为研究肉牛的反刍行为规律,在辽宁未来牧业随机抽取50头处于育肥期的西门塔尔健康肉牛作为试验对象,进行为期7d的日常活动观察,并记录其行为。将50头肉牛分为试验组和对照组,每组25头。在试验组肉牛的额头绑上九轴蓝牙动作传感器,用于采集肉牛在x,y,z轴上的加速度,角速度和角度共9组特征向量数据。通过小波降噪对此特征向量进行预处理,选择标准归一化的方法降低数据间相互的纲量和影响。抽取特征向量中的70%作为训练集投入SVM(support vector machine,支持向量机)进行二分类训练得到反刍模型,随后将其余30%作为测试集带入反刍模型进行分类预测,以判断是否存在反刍行为。最后,利用对照组的肉牛与试验组相比对,判断试验组的肉牛是否出现异常行为。试验结果表明:所采用试验设备工作正常,信号传输稳定;舍饲肉牛在佩戴传感器节点过程中无不良反应。SVM可以通过蓝牙动作传感器发送得到的特征向量高效且准确的判断肉牛的反刍行为与非反刍行为。分类结果达到97.7728%(其中,反刍识别率97.659%,非反刍识别率97.667%),满足对反刍行为分类的目的 ,可有效的识别反刍行为。
[Abstract]:In order to further strengthen the standardized feeding of ruminant livestock, rationally utilize resources and improve the quality of production, the study of their daily behavior has become a key factor. Fifty beef cattle were divided into experimental group and control group, 25 in each group. Nine-axis Bluetooth motion sensors were attached to the forehead of beef cattle in the experimental group to collect the acceleration, angular velocity and angle of beef cattle on x, y, Z axes. Eigenvector data. wavelet denoising is used to preprocess the eigenvector, and standard normalization method is selected to reduce the dimension and influence between the data. 70% of the eigenvectors are extracted as training set and put into SVM (support vector machine) for binary training to get the rumination model. The remaining 30% is used as test. Finally, the beef cattle in the control group were compared with the experimental group to determine whether abnormal behavior occurred in the experimental group. The results showed that the experimental equipment was working normally and the signal transmission was stable. SVM can effectively and accurately judge ruminant behavior and non-ruminant behavior of beef cattle by feature vectors sent by Bluetooth motion sensor. The classification result reaches 97.7728% (of which, the recognition rate of ruminant is 97.659%, the recognition rate of non-ruminant is 97.667%). SVM can meet the purpose of classification of ruminant behavior and can effectively identify ruminant behavior.
【作者单位】：沈阳农业大学信息与电气工程学院;国网大连供电公司;
【基金】：辽宁省科技厅自然科学基金项目(2015020760)
【分类号】：S823

【相似文献】