A Novel Image Processing Approach for Colloid Detection in Saturated Porous Media

Area of Science:

• Environmental Science
• Colloid Science
• Image Processing

Background:

• Colloids and nanoparticles are increasingly prevalent in surface and subsurface environments.
• Colloid presence impacts contaminant transport and groundwater quality, necessitating accurate detection methods.

Purpose of the Study:

• To evaluate various image processing techniques for accurate colloid detection in microscopy images.
• To develop an improved colloid detection method for porous media research.

Main Methods:

• Four categories of image processing methods were applied: segmentation-based, background-detection-based, filter-based, and morphology-based.
• Eight specific methods were analyzed for their advantages and disadvantages.
• An ensemble approach using a majority vote of the top three methods was proposed.

Main Results:

• Morphology-based methods demonstrated the highest detection performance, especially for small colloids.
• The proposed ensemble approach achieved perfect scores across all evaluation criteria (Precision, Recall, F-measure, TCR).

Conclusions:

• Image processing methods are highly accurate for colloid recognition.
• The developed ensemble approach offers superior colloid detection accuracy for porous media studies.