Oestrus Analysis of Sows Based on Bionic Boars and Machine Vision Technology

Area of Science:

• Animal Science
• Agricultural Engineering
• Artificial Intelligence

Background:

• Accurate oestrus detection is crucial for efficient sow reproduction.
• Traditional methods for oestrus detection can be labor-intensive and less precise.
• Technological advancements offer potential for automated and improved monitoring systems.

Purpose of the Study:

• To develop an intelligent mobile monitoring system for detecting oestrus in sows.
• To utilize a bionic boar model and machine vision for automated oestrus behaviour recognition.
• To enhance the accuracy and efficiency of determining optimal conception times in swine production.

Main Methods:

• A recyclable bionic boar model was designed, mimicking real boar stimuli (sound, smell, touch).
• Machine vision technology was employed to analyze sow-bionic boar interaction behaviors.
• Deep belief network (DBN), sparse autoencoder (SAE), and support vector machine (SVM) models were trained for oestrus recognition.

Main Results:

• The DBN, SAE, and SVM models achieved high accuracy rates of 96.12%, 98.25%, and 90.00%, respectively.
• A strong correlation was found between sow-bionic boar contact duration and static ear behaviors during oestrus.
• Oestrus sows showed an average contact duration of 29.7 s/3 min and static ear duration of 41.3 s/3 min.

Conclusions:

• The proposed intelligent system accurately identifies sow oestrus states based on interaction behaviors.
• The recyclable bionic boar design and machine vision offer an innovative and efficient approach to oestrus monitoring.
• This method provides a more accurate oestrus duration and a scientific basis for optimizing sow conception timing.