Detecting Repeated Frying on Cooking Oils based on its Visual Properties using Embedded System

Cooking oil has been widely used to conduct heat from the pan to the food in order to fry the food to affect its taste, color and texture. The exposure to high temperature repetitively can degrade the quality of cooking oil. Moreover, dangerous chemical reaction may affect human health who consume it. The quality of cooking oil can be detected visually from its color and its clarity. In this study, an embedded system to detect cooking oil frequency of use based on its visual properties was designed. The proposed system used color and photodiode sensor to extract the visual information of cooking oil. Then, k-Nearest Neighbor (k-NN) algorithm was implemented on the embedded system platform to predict and classify the cooking oil into 5 classes. There were 49 dataset that were used as training dataset. Using 10-fold cross validation process, k=3 were selected for its lowest misclassification error. Finally, the system was tested using real data test while simultaneously measure its computation time performance. The result shows 100% classification accuracy from 20 test data and on average, k-NN require 24.25 ms to perform the classification on Arduino UNO board.