Motor Oil Classification Using Color Histograms and Pattern Recognition Techniques.

Area of Science:

• Analytical Chemistry
• Computer Science
• Materials Science

Background:

• Motor oil classification is crucial for quality control and detecting adulteration.
• Accurate identification methods are needed to ensure oil integrity.

Purpose of the Study:

• To develop a simple, rapid, inexpensive, and nondestructive method for classifying nine types of motor oil.
• To utilize image analysis and pattern recognition for oil classification based on color histograms.

Main Methods:

• Applied color histogram analysis in RGB, grayscale, and HSI color spaces.
• Utilized Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), and Support Vector Machine (SVM) for classification.
• Employed one-against-all (OAA) and multi-label classification strategies, with and without Principle Component Analysis (PCA).

Main Results:

• The OAA strategy achieved up to 97% accuracy, sensitivity, and specificity using LDA, QDA, and SVM.
• The multi-label approach yielded up to 93% accuracy with LDA and QDA, particularly using HSI color maps.
• PCA effectively reduced variables for modeling without significant performance loss.

Conclusions:

• The proposed image analysis and pattern recognition method shows significant promise for motor oil classification and adulteration identification.
• This non-destructive technique offers a viable solution for rapid and cost-effective quality assessment of motor oils.