Design of a Portable Spectrophotometer Based on Raspberry Pi for Tea Type Classification Using Machine Learning

Abstract

The use of spectrophotometers in food and beverage quality analysis has become common. However, portability and high cost have long been a problem for food industry players and small-scale laboratories. In this study, a more compact and low-cost spectrophotometer has been developed using raspberry pi. To validate its performance, the prototype was used to classify green tea, black tea, and oolong tea types. The research started with designing and assembling the hardware using a raspberry pi camera with Complementary Metal Oxide Semiconductor (CMOS) sensor, Digital Versatile Disk (DVD) as diffraction grating, white LED as light source and 3D printer as casing. The prototype was then used to acquire data from green tea, black tea and oolong tea solutions. In the tea type identification process, Principal Component Analysis (PCA) was applied to the data obtained. The experiment involved 30 solution samples of each of the three types of tea. The light source was directed past the sample towards the slit and DVD, then the spectrum image of the light source was displayed through a user interface built using python programming. This research resulted in a spectrophotometer with dimensions of 260 mm x 120 mm x 63 mm that is capable of capturing light spectra in the range of 400 - 700 nm. Based on the experimental results, the classification accuracy of the three types of tea using CNN and Decision Tree reached 100%.