Grape growers usually need to perform time-consuming lab test on sugar content to determine the best time for harvest. This project aims to replace sugar content lab test with a machine vision solution that comprises a multi-spectral imaging system and a machine learning model to estimate sugar content based on multi-spectral reflectance.

  • Develop a multi-spectral imaging prototype with raspberry-pi, picamera and camera filters
  • Collect multi-spectral images of grapes using the imaging prototype and the corresponding sugar content using a BRIX meter
  • Pre-process and align channels of spectral images with traditional computer vision techniques (keypoint detection, template matching) using Python and OpenCV
  • Develop regression-based machine learning models to estimate sugar content based on multi-spectral reflectance using Python and Scikit-learn