Detect Counterfeit Olive Oil with Light and AI

ISEF Category: Chemistry

Subcategory: Analytical Chemistry  ·  Difficulty: Intermediate  ·  Setup: School Lab  ·  Time: 1 to 2 Months

The Hook

Counterfeit olive oil can hide in plain sight. A bottle can look perfect, smell normal, and still be mixed with cheaper oils. You can test for that kind of fraud with light, a cheap refractometer, and a machine learning model. That makes this project part chemistry detective work, part data science.

What Is It?

This project looks at whether you can tell real olive oil from adulterated oil by measuring two signals. One signal is refractive index, which tells you how much a liquid bends light. The other signal is UV-induced fluorescence, which is the glow some oils give off when you shine 365 nm light on them. Think of it like fingerprinting a liquid. One test gives you a number from a refractometer, and the other gives you image patterns from the oil’s glow.

Olive oil is not a pure, single compound. Its composition changes with variety, age, storage, and mixing. That means you are not testing one molecule. You are testing a chemical mixture. A random-forest classifier can learn patterns from several measurements at once and sort samples into groups. In plain terms, it looks at many clues, then votes on the most likely label.

Why This Is a Good Topic

This is a strong science fair topic because you can measure real differences with accessible tools and turn them into a clear classification problem. You also connect to a real-world issue, food fraud, which affects consumers, labeling, and supply chains. You can study one variable at a time, then improve the model with better features or better sample prep. That gives you room to show both chemistry knowledge and data analysis skill.

Research Questions

• How does adulteration level change the refractive index of olive oil mixtures?
• What is the effect of adding different seed oils on UV-induced fluorescence image features?
• Does combining refractive index and fluorescence improve classification accuracy compared with either method alone?
• To what extent can a random-forest classifier separate authentic olive oil from adulterated samples?
• Which image features, such as brightness, color ratios, or texture, best predict olive oil adulteration?
• How does storage time affect the refractive index and fluorescence signature of olive oil samples?

Basic Materials

• Cheap handheld refractometer.
• 365 nm UV flashlight.
• Clear glass or quartz sample dishes.
• Authentic extra-virgin olive oil.
• Common adulterant oils, such as sunflower, canola, or soybean oil.
• Smartphone with a fixed camera app.
• White background or light box.
• Metric ruler or coin for scale.
• Disposable pipettes or droppers.
• Nitrile gloves.
• Notebook or spreadsheet for sample labels.

Advanced Materials

• Digital refractometer.
• UV-safe imaging enclosure.
• UV-transmitting cuvettes or optical glass vials.
• Controlled light source with fixed geometry.
• Spectrophotometer, if available.
• Laboratory balance with 0.01 g precision.
• Food-grade reference oils with certificates, if available.
• Image calibration target.
• Reference standards for mixture ratios.
• Analytical software for feature extraction.

Software & Tools

• ImageJ: Measures brightness, color channels, and texture features from fluorescence images.
• Python: Lets you clean data, build features, and train a random-forest classifier.
• scikit-learn: Provides random-forest models, cross-validation, and accuracy metrics.
• Google Sheets: Organizes sample labels, refractive index values, and image features.
• PubChem: Helps you check oil-related compounds and background chemistry for fluorescence behavior.

Experiment Steps

1. Define the sample set and decide which oil mixtures you will compare first.
2. Choose your measurement order, so refractive index and fluorescence images stay matched to the same sample.
3. Plan a fixed imaging setup that keeps camera position, lighting, and background constant.
4. Build a feature list before you collect data, so you know which image measurements matter.
5. Decide how you will label samples for the classifier and how you will split training and test data.
6. Select evaluation metrics that show whether the combined method beats either signal alone.

Common Pitfalls

• Changing the distance between the flashlight, the sample, and the camera, which makes fluorescence features shift between photos.
• Mixing oils by eye instead of by mass or volume, which weakens the link between adulteration level and your labels.
• Letting fingerprints, bubbles, or dust sit in the sample dish, which adds fake texture to the images.
• Training and testing on nearly identical samples from the same bottle, which inflates classifier accuracy.
• Using only one authentic olive oil source, which makes the model learn that brand instead of oil quality.

What Makes This Competitive

A stronger project does more than sort two or three bottles. It tests many mixture ratios, compares several adulterant oils, and checks whether the model still works on new brands or storage conditions. You can also compare feature sets, like refractive index alone, fluorescence alone, and both together. If you include cross-validation and a blind test set, your results look much more credible.

Project Variations

• Test whether the same method can detect counterfeit avocado oil or sesame oil instead of olive oil.
• Compare UV fluorescence under 365 nm and blue-light excitation to see which gives cleaner class separation.
• Add simple texture metrics from phone images and test whether they improve the classifier more than color alone.

Learn More

• US FDA Food Fraud resources: Search the FDA site for food fraud, economically motivated adulteration, and olive oil identity standards.
• NIH PubMed: Search review articles on olive oil adulteration detection, fluorescence spectroscopy, and chemometrics.
• USDA FoodData Central: Look up fatty acid and composition data for common edible oils.
• NIST Chemistry WebBook: Use it to review optical and physical properties background for food chemistry projects.
• MIT OpenCourseWare: Search for introductory materials on data analysis, statistics, and machine learning basics used in classification projects.
• Food Chemistry: Search this journal for peer-reviewed papers on olive oil authenticity and optical methods.