Milk Spoilage Classification through Integration of RGB and Thermal Data Analysis

Abstract

In Malaysia, milk consumption is commonly associated with family households, specifically children. The nutrition in milk is fundamental for children’s growth which is why the parents will ensure their children have adequate milk intake from an early age. Various kinds of milk are available on the market, but pasteurized milk and UHT milk are the most consumed. Without proper storage and packaging conditions, milk could spoil quickly; hence an early detection method is needed to detect milk staleness and spoilage. Much research and study has been done regarding the classification of milk spoilage. However, factors such as the unreliability of data and time-consuming methods prove that a better working model with high accuracy needs to be developed. Efficient detection methods are crucial for ensuring milk quality. This project is targeted to develop and introduce image-based analysis to detect the spoiled milk in various packaging and storage conditions using Deep Learning and Python programming language to cater for the problem stated above. A dataset containing both RGB and thermal images of milk was self-obtained. The proposed model in this paper has achieved the accuracy of 99% for classification of RGB images of milk and 98% for the thermal images of milk.