【摘要】 综述了声音监测技术、机器视觉技术、无线传感网络技术在牛行为监测的研究与应用现状,分析了支持向量机(SVM)、K均值聚类(K-means)、人工神经网络(ANNs)等3种牛行为识别分类算法的优缺点,结果表明:(1)机器视觉技术具有无接触的识别,不外带装置,可以对动物行为进行识别,对牛活动影响小,但对图像视频环境要求苛刻,动物行为识别精准度不高;(2)无线传感器技术应用广、技术成熟,可以监测畜禽采食、反刍、休息、活动等行为,但适合动物穿戴、长期可靠工作的无线网络传感器技术有待突破;(3)支持向量机计算简单,理论完善,所需样本数据少,且识别精度高,分类效果好。人工神经网络算法的学习规则简单,非线性拟合能力较强,但数据不足易出现运算时间长、过学习、容易陷入局部最小值等情况。更多还原

【Abstract】 The research and application of sound monitoring technology, machine vision technology and wireless sensor network technology in cattle behavior monitoring were summarized, and the classification of animal behavior recognized by support vector machine, K-means, artificial neural network was described in the paper. The results showed that:(1) Machine vision technology had no contact recognition, did not take out the device, could identify the behavior of animals, and had little impact on the activities of livestock.(2) Wireless sensor technology was widely used and mature, which could monitor the behavior of livestock, such as feeding, ruminating, rest, activity and so on, but the wireless network sensor technology which was suitable for animals to wear and work reliably for a long time needed to be broken through.(3) Support vector machine had the advantages of simple calculation and perfect theory, which was suitable for less sample data to get higher recognition rate. The learning rule of artificial neural network algorithm was simple, and the nonlinear fitting ability was strong, but the lack of data was prone to long time, over-learning, local minimum and so on.更多还原