Measuring Shape Parameters of Pearls in Batches Using Machine Vision: A Case Study

Area of Science:

• Computer Vision
• Image Processing
• Metrology

Background:

• Accurate pearl shape parameter measurement is crucial for grading and sorting.
• Existing methods struggle with low precision due to pearl contact and sorting errors.
• Contacting pearls present a significant challenge in automated analysis.

Purpose of the Study:

• To develop an accurate method for segmenting mutually contacting pearls.
• To improve the precision of pearl shape parameter measurement.
• To enhance the accuracy of pearl shape-based sorting.

Main Methods:

• Backlit imaging and image preprocessing to enhance pearl visibility.
• Connected domain analysis to extract contacted pearl areas.
• Edge tracking and concave point detection for precise pearl segmentation.
• Euclidean distance metric for segmenting tangent pearls.
• Establishing a pearl shape parameter model and classification standard.

Main Results:

• The proposed pit detection method achieved over 95% segmentation accuracy.
• The average loss rate for pearl segmentation was within 4%.
• Sorting accuracy based on the derived shape information reached 94%.

Conclusions:

• The pit detection-based segmentation method offers superior performance compared to watershed and morphological algorithms.
• This method effectively addresses the challenges of measuring and sorting contacting pearls.
• The improved accuracy in segmentation and sorting has significant implications for the pearl industry.