FEgg3D: A Non-Contact and Dynamic Measuring Device for Egg Shape Parameters and Weight Based on a Self-Designed Laser Scanner

Abstract

In large-scale poultry farming, real-time online measurement of egg weight and shape parameters remains a challenge. To address this, we developed FEgg3D, a non-contact dynamic measuring device based on a self-designed laser scanner. The device employed a subset of the point cloud generated to predict the shape parameters and weight of eggs using machine learning algorithms. Different colors and sizes of eggs on various backgrounds were scanned using FEgg3D mounted on a gantry system. Our results demonstrated the following: (1) The Support Vector Regression (SVR) model was optimal for major axis length estimation, with an R2 of 0.932 using six laser lines and eight points per line. (2) The Gaussian Process Regression (GPR) model excelled in minor axis length estimation, achieving an R2 of 0.974 with six laser lines and 16 points per line. (3) SVR was optimal for volume estimation, attaining an R2 of 0.962 with six laser lines and 16 points per line. (4) GPR showed superior performance in weight prediction, with an R2 of 0.964 using five laser lines and 16 points per line. Including density features significantly improved accuracy to an R2 of 0.978. This approach paves the way for advanced online egg measurement in commercial settings.

Keywords

• laser scanner
• machine learning
• dynamic
• egg parameters
• egg weight