AgriTera: Accurate Non-Invasive Fruit Ripeness Sensing via Sub-Terahertz Wireless Signals

  • Sayed Saad Afzal ,
  • Atsutse Kludze ,
  • Subhajit Karmakar ,
  • ,
  • Yasaman Ghasempour

ACM MobiCom |

Published by ACM | Organized by ACM

Related File

The ability to assess the quality of fruit and vegetables at scale can revolutionize the agriculture sector and significantly reduce food waste. In this paper, we present AgriTera, a novel solution for accurate non-invasive, and contract-free fruit ripeness sensing via sub-terahertz wireless signals. The key idea is that sugar and water concentrations in fruit (that are associated with fruit ripening) leave unique non-uniform footprints in the wide band spectrum of the reflected signal off of the fruits. AgriTera utilizes the sub-THz bands for its wide bandwidth, sensitivity to water, mm-scale penetration depth, and non-ionizing features that offer high-resolution inferences from the peel as well as the pulp underneath the peel. We develop a chemometric model that translates the reflection spectra to well-known ripeness metrics, namely Dry Matter and Brix. We conduct extensive over-the-air experiments with commercially available sub-THz transceivers. We compare our results with ground truth values captured by a specialized quality sensor and a vision-based scheme that infers ripeness based on changes in the appearance of the fruit. We demonstrate that AgriTera can accurately estimate Brix and Dry Matter in three different types of fruit with an average Normalized RMSE value of 0.55%, an error that yields a negligible impact on taste and is imperceivable by the consumer.