Evaluation of freshness of lettuce using multi-spectroscopic sensing and machine learning

  • Akane Tsukahara Mie University, Japan
  • Shinichi Kameoka Mie University, Japan
  • Ryoei Ito Mie University, Japan
  • Atsushi Hashimoto Mie University, Japan
  • Takaharu Kameoka Mie University, Japan


We aimed to develop a method to evaluate lettuce freshness changes during storage using only the surface color. In the first experiment, the surface color of one lettuce were measured continuously for 6 days. At the same time, moisture contents, elemental composition and organic matter of lettuce leaves were measured by oven drying method, X-ray fluorescent analysis and Mid-infrared spectroscopy, respectively. Considering a combination of the surface color and moisture and elemental contents, it was found that there were several color change points before and after the time when the moisture contents and elemental balances in the lettuce changed. These results represented that the surface color could relate to the internal quality. Additionally, it is suggested that freshness of lettuce could be quantified and predicted using surface color information.
Furthermore, the data set and the method for freshness evaluation leading to machine learning were studied in the second experiment for the freshness judgement. In this experiment, 15 multispectral sensing data including lettuce color information were acquired, and the quality change point was determined using machine learning such as K-means and decision tree.