Recognition of Abnormal-Laying Hens Based on Fast Continuous Wavelet and Deep Learning Using Hyperspectral Images

Area of Science:

• Agricultural Science
• Biotechnology
• Computer Science

Background:

• Identifying low-egg-production laying hens is critical for poultry breeding enterprises.
• Current methods rely on breeder experience, lacking systematic accuracy.
• A precise and widely applicable identification method is needed.

Purpose of the Study:

• To develop a highly accurate and systematic method for identifying low-egg-production laying hens.
• To leverage hyperspectral imaging and deep learning for hen classification.
• To enhance efficiency and precision in commercial laying hen farms.

Main Methods:

• Hyperspectral imaging captured images of low-egg-production and normal laying hens.
• Vertex Component Analysis (VCA) extracted cockscomb spectral features.
• Fast Continuous Wavelet Transform (FCWT) processed spectral data into 2D images.
• A Convolutional Neural Network (CNN) deep learning model was trained on spectral image datasets.

Main Results:

• The developed CNN model achieved 97.5% accuracy in identifying low-egg-production laying hens.
• The combined method of hyperspectral imaging, FCWT, and CNN proved highly effective.
• FCWT demonstrated high efficiency and resolution in hyperspectral data analysis.

Conclusions:

• The proposed method offers a precise and effective solution for identifying low-egg-production laying hens.
• This approach significantly improves upon traditional experience-based identification.
• The application of FCWT in hyperspectral data analysis has broader implications for various scientific fields.