Vision-based mixed color detection of plastic particles

Area of Science:

• Materials Science
• Polymer Engineering
• Image Processing

Background:

• Color mixing in high-end PP-R pipe production leads to aesthetic defects and compromises physical properties.
• Current inspection methods may lack efficiency and accuracy in detecting these color variations.

Purpose of the Study:

• To develop a visual, non-destructive inspection solution for identifying color inconsistencies in PP-R pipes.
• To enhance detection accuracy and efficiency using advanced image processing techniques.
• To ensure the quality, durability, and safety of high-end PP-R pipes.

Main Methods:

• Utilized K-Means image segmentation to remove complex backgrounds and improve initial image segmentation.
• Employed Gaussian mixture models for automatic color threshold extraction from foreground images.
• Applied the mean value algorithm for swift and accurate identification of different colored plastic particles.

Main Results:

• The proposed image processing method effectively identifies different colored plastic particles.
• The system demonstrated a high average detection accuracy of 99.3%.
• The approach successfully supports the rejection of impurity particles, enhancing product quality.

Conclusions:

• The developed visual inspection solution accurately detects color mixing issues in PP-R pipe production.
• This non-destructive method improves product quality control and material integrity.
• The algorithm's adaptability and high accuracy offer a robust solution for industrial applications.