Research on the Evaluation Method of Eggshell Dark Spots Based on Machine Vision

Chunshan Wang, Ji Zhou, Huarui Wu, Jiuxi Li, Chunjiang Zhao, Rong LiuIEEE Access 2020 · DOI

Introduction

Eggshell "dark spots," or eggshell translucency, are watermark-like patches that typically manifest 5–14 days after an egg is laid. While often viewed as a cosmetic defect, these spots represent a significant concern for both the poultry industry and public health. Dark spots indicate increased shell porosity, which facilitates the penetration of bacteria such as Salmonella into the egg's interior. Furthermore, these defects lower the commercial grade and consumer appeal of the product.

Manual inspection for dark spots is labor-intensive, subjective, and requires high-intensity transillumination to make the spots visible. To address these challenges, researchers have developed an automated, high-speed machine vision system designed to identify and quantify these defects early in the supply chain.

Methodology

The proposed system utilizes a specialized experimental apparatus to capture high-quality images of eggs under transillumination. The hardware consists of a conveyor belt with rollers, an LED cold light source positioned beneath the eggs, and two CCD cameras to capture images of both sides of the egg shell.

The experimental system design includes (a) a top view of the image acquisition area, (b) a sectional view showing the LED transillumination setup, and (c) the resulting images of egg surfaces A and B.

Image Processing Pipeline

The software architecture for evaluating the eggs involves three critical stages:

  1. Background Segmentation: The system employs K-means clustering (K=2) to isolate the egg from the mechanical background. While deep learning models like DeepLab-v3 were tested, K-means was selected for its superior balance of speed and accuracy in a real-time industrial context.
  2. Feature Enhancement: To highlight subtle dark spots, the system uses unsharp masking. A Gaussian low-pass filter creates a blurred reference image, which is then subtracted from the original to isolate high-frequency details—the dark spots themselves.
  3. Quantitative Evaluation: The system calculates the total count of individual spots using 8-connected region labeling and determines the ratio of the dark-spot area relative to the total projected area of the egg.

Experimental Results

The system was validated using a dataset of 360 eggs from Jinghong and Jingbai laying hens. The machine vision approach demonstrated significant improvements over manual methods:

  • Efficiency: The system processes a single frame in just 0.5 seconds. Compared to the 480 seconds required for a human to manually mark the same image, this represents a 960x speedup.
  • Throughput: The experimental device achieved a detection capacity of 3,600 eggs per hour, making it suitable for commercial production lines.
  • Sensitivity: The algorithm proved more sensitive than human inspectors, identifying tiny, subtle spots that are often overlooked during manual marking.

Discussion and Future Work

While the system is highly effective, certain limitations remain. The use of 2D imaging on 3D ellipsoidal objects causes distortion of spots near the egg's edges. Future research will explore 3D surface reconstruction to improve area calculation accuracy. Additionally, the system currently requires clean eggs, as surface contaminants like dust or stains can lead to false positives.

This automated evaluation method provides a robust tool for the poultry industry to ensure food safety and maintain product quality standards through objective, high-speed grading.

Summary updated

BibTeX
@article{wang2020research,
  author = {Chunshan Wang and Ji Zhou and Huarui Wu and Jiuxi Li and Chunjiang Zhao and Rong Liu},
  title = {{Research on the Evaluation Method of Eggshell Dark Spots Based on Machine Vision}},
  journal = {IEEE Access},
  year = {2020},
  doi = {10.1109/access.2020.3020260},
}
Figure