首页> 外国专利> Prediction method and the apparatus for onset time of sow farrowing by image analysis

Prediction method and the apparatus for onset time of sow farrowing by image analysis

机译:母猪产仔开始时间的图像分析预测方法及装置

摘要

A method of predicting a delivery time of a mother pig using image analysis according to an aspect of the present invention includes: collecting a delivery image data for each class of behavior patterns according to behavior characteristics of a pregnant mother pig by a delivery time prediction device; The delivery time prediction device normalizes the behavior pattern for each class of pregnant mother pigs to the predetermined coordinates with respect to the learning image data according to reference coordinates, and as learning processing data capable of recognizing the characteristics of the behavior pattern in the CNN recognition model. Processing to extract; A step in which the delivery time prediction device accumulates and learns processing data for learning for each class behavior pattern by applying a CNN recognition model; Storing the learned data in the behavior pattern database by the delivery time prediction device; Collecting the image data for recognition of the mother pig for recognition by the delivery time prediction device; The delivery time prediction device normalizes the recognition image data to a certain standard according to reference coordinates, and applies the CNN recognition model to recognize the behavior pattern of a mother pig for the recognition image data, and patterns for each class Classifying as one of; An analysis step of analyzing the expression pattern by time zone or the occurrence frequency and change trend by time zone for the classified behavior pattern; And a delivery prediction time notification step for determining the delivery prediction time point for the recognition mother pig from the analysis step and outputting the notification from 24 hours before.
机译:根据本发明的一个方面,使用图像分析来预测母猪的分娩时间的方法包括:通过分娩时间预测装置,根据怀孕的母猪的行为特征,收集每一类行为模式的分娩图像数据。 ;交付时间预测装置根据参考坐标,相对于学习图像数据将每种母猪的行为模式标准化为预定坐标,并且作为能够在CNN识别中识别行为模式特征的学习处理数据。模型。加工提取;交付时间预测装置通过应用CNN识别模型来积累和学习用于学习每个班级行为模式的处理数据的步骤;交付时间预测装置将学习到的数据存储在行为模式数据库中;收集图像数据以识别母猪,以通过分娩时间预测装置进行识别;交付时间预测装置根据参考坐标将识别图像数据归一化为一定标准,并应用CNN识别模型识别出识别图像数据的母猪的行为模式,并将每个类别的模式归类为其中之一。分析步骤,用于对分类的行为模式进行按时区的表达模式或按时区的发生频率和变化趋势的分析;还有一个交货预测时间通知步骤,用于从分析步骤中确定识别母猪的交货预测时间点,并从24小时之前输出通知。

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号