Mozzarella cheese is popular worldwide, and its production has greatly increased in the last decade. During mozzarella cheese production, a process named stretching is essential to the cheese quality. This stretching process plays an important role in fiber structure formation to reach a unique fibrous texture. Thus, it is crucial to monitor the microstructure transition of mozzarella during production.
This work is part of a research project on real-time monitoring of continuous mozzarella production, which includes nearby colleagues as well as external academic and industrial partners. The purpose is to find methods for acquiring optical properties and for correlating these with microscopic and functional properties of the observed product. Currently, our research experiments will be conducted by SR (spatially resolved) HIS (Hyperspectral imaging) technique. CT scan will be used to get the detailed microstructure of mozzarella cheese and used as reference to compare with the image get from hyperspectral. Other methods for model-based estimation of microstructure from optical measurements will be also studied. Another research group will utilize the results to correlate the microstructure to functional cheese properties. Subsequently, a vision company associated with the project will apply the research outcome to implement the inline use. Research on light-matter (hyperspectral or CT scan) interaction and machine learning techniques will be applied to develop robust models for real-time monitoring of product microstructures.
The overall focus of the research will be computer vision with industrial applications for real-time monitoring of microstructures. A hyperspectral camera will be set to the manufacture line to collect the real-time image of the mozzarella cheese during the production. The computer vision system should be able to recognize the change of cheese microstructure to judge if the cheese is ready to be sent to the next production phase. To accomplish this, it is necessary to create a model that connects macroscopic camera-based observations with microstructural features in the observed product. The research results can have a direct impact on the industrial machine vision used for mozzarella production in the future.