A Low-Cost Image Histogram and Machine Learning Approach for Detection of Cow Milk Adulteration.

Area of Science:

• Food Science
• Analytical Chemistry
• Computer Science

Background:

• Milk adulteration poses significant food safety and public health risks.
• Limited access to rapid, inexpensive testing methods hinders detection, especially in certain regions.
• Traditional laboratory analyses are often costly and time-consuming.

Purpose of the Study:

• To develop a simple, affordable, and accessible method for detecting common milk adulterants.
• To utilize image analysis and machine learning for milk quality assessment.
• To provide a practical alternative to expensive spectroscopic techniques.

Main Methods:

• Collected digital images of pure milk samples and samples adulterated with water, detergent, starch, and synthetic milk.
• Extracted image statistics including brightness, variance, skewness, and kurtosis to analyze light scattering and turbidity.
• Employed Principal Component Analysis (PCA) for data dimensionality reduction.
• Utilized Support Vector Machines (SVMs) for classification of adulterated versus pure milk samples.

Main Results:

• The developed model achieved approximately 85% accuracy in identifying various types and amounts of adulteration.
• Specific adulterants induced unique changes in image statistics: water increased mean intensity, while detergent, starch, and synthetic milk altered skewness and kurtosis.
• The method demonstrated effectiveness in distinguishing between pure and adulterated milk based on optical properties.

Conclusions:

• Image analysis combined with machine learning offers a viable, economical approach for milk adulteration detection.
• This method is faster, requires no chemicals, and is more accessible than traditional spectroscopic methods.
• The technique is suitable for rapid quality control checks at points of sale or in remote areas.