Research on a Dynamic Algorithm for Cow Weighing Based on an SVM and Empirical Wavelet Transform.

Area of Science:

• Animal Science
• Agricultural Engineering
• Data Science

Background:

• Accurate dairy cow weight monitoring is crucial for growth assessment.
• Traditional static weighing is labor-intensive and costly.
• Existing dynamic weighing methods neglect the impact of cow motion on accuracy.

Purpose of the Study:

• To develop and validate a dynamic weighing algorithm for dairy cows.
• To classify cow motion states (low, medium, high activity) using SVM.
• To accurately determine dynamic cow weights using EWT and filtering techniques.

Main Methods:

• Acquisition of dynamic weight curves using a corridor-based weighing device.
• Data preprocessing including signal acquisition, feature extraction, and normalization.
• Classification of cow motion states via Support Vector Machine (SVM).
• Application of mean filtering, Empirical Wavelet Transform (EWT), and a combined periodic continuation-EWT algorithm for weight calculation.

Main Results:

• Achieved a classification accuracy of 98.6928% for cow motion states.
• Calculated dynamic weights with average error rates of 0.1838% (mean filtering), 0.6724% (EWT), and 0.9462% (combined EWT).
• Demonstrated high precision in dynamic weight estimation across different motion levels.

Conclusions:

• The proposed SVM and EWT-based dynamic weighing algorithm offers a highly accurate and efficient solution for dairy farms.
• This method overcomes limitations of traditional weighing and existing dynamic algorithms by accounting for cow motion.
• The findings contribute to advancing precision livestock farming and dairy herd management practices.