Alternative Non-Destructive Approach for Estimating Morphometric Measurements of Chicken Eggs from Tomographic Images...

Area of Science:

• Agricultural Science
• Biomedical Engineering
• Computer Vision

Background:

• Chicken eggs possess natural defenses against microbial contamination, crucial for food safety.
• Traditional methods for egg morphometric analysis are invasive and can be less efficient.
• Non-destructive techniques offer a promising alternative for accurate and efficient egg quality assessment.

Purpose of the Study:

• To demonstrate the comparability of non-invasive 3D computed tomography (CT) image analysis with conventional destructive methods for egg morphometrics.
• To evaluate the effectiveness of deep learning models in segmenting and analyzing 3D CT images of chicken eggs.
• To establish a scalable and reliable non-invasive method for egg quality assessment.

Main Methods:

• Development and application of two deep learning architectures: U-Net 3D and Fully Convolutional Networks (FCN) 3D.
• Creation and labeling of a dataset comprising real 3D CT images of chicken eggs.
• Extraction of morphometric parameters such as height, width, shell thickness, and volume from segmented images.

Main Results:

• Deep learning models achieved high accuracy (up to 98.69%) in segmenting and analyzing 3D CT egg images.
• The morphometric measurements obtained were comparable to those from traditional manual measurements.
• Demonstrated the effectiveness of the non-invasive approach in capturing key egg characteristics.

Conclusions:

• 3D CT image analysis combined with deep learning provides a viable non-invasive alternative for egg quality assessment.
• This approach minimizes the need for destructive testing in industrial and research settings.
• Offers a scalable, accurate, and reliable method for evaluating chicken egg morphometrics and quality.