Single-View Measurement Method for Egg Size Based on Small-Batch Images.

Area of Science:

• Agricultural Engineering
• Computer Vision
• Metrology

Background:

• Egg size is critical for consumer perception and quality assessment in the food industry.
• Accurate and efficient egg measurement methods are essential for automated grading systems.
• Traditional methods can be labor-intensive and prone to inaccuracies.

Purpose of the Study:

• To develop and validate a deep learning-based single-view metrology system for measuring egg dimensions.
• To assess the performance of the Segformer algorithm in segmenting egg images for accurate outline extraction.
• To establish a reliable method for determining egg major and minor axes.

Main Methods:

• Designed a specialized egg-carrying component to capture clear egg outlines.
• Employed the Segformer deep learning algorithm for image segmentation of eggs in small batches.
• Utilized single-view metrology principles to measure egg major and minor axes from segmented images.

Main Results:

• The Segformer algorithm achieved high segmentation accuracy, with a mean Intersection over Union (IoU) of 96.15% and mean pixel accuracy of 97.17%.
• The proposed single-view measurement method demonstrated strong performance, yielding R-squared values of 0.969 for the major axis and 0.926 for the minor axis.
• Experimental results confirm the effectiveness of the deep learning approach for egg dimension measurement.

Conclusions:

• The developed deep learning-based single-view measurement system provides an accurate and efficient solution for egg size determination.
• The Segformer algorithm is highly effective for egg image segmentation, facilitating precise dimensional analysis.
• This method holds significant potential for application in automated egg quality grading systems.