Harvest smarter, not harder: Israeli machine learning meets tomato farming

This innovative approach offers a cost-effective, non-invasive method to predict critical quality parameters of tomatoes, including weight, firmness, and lycopene content, long before harvest, the researchers said.

By Pesach Benson, TPS

Revolutionizing harvesting, Israeli researchers harnessed artificial intelligence and technology to offer a cost-effective and non-invasive way to predict the quality parameters of tomatoes.

Led by Dr. David Helman of Hebrew University in collaboration with Bar-Ilan University and the Volcani Institute – Agricultural Research Organization, the researchers used a handheld hyperspectral camera to get detailed insights into tomatoes, then used machine learning algorithms to process the data. The team found the results were remarkably accurate.

“Our research aims to bridge the gap between advanced imaging technology, AI, and practical agricultural applications,” said Helman.

“This work has the potential to revolutionize quality monitoring not only in tomatoes but also in other crops. Our next step is to build a low-cost device, ToMAI-SENS, for use across the fruit value chain, from farms to consumers.”

This innovative approach offers a cost-effective, non-invasive method to predict critical quality parameters of tomatoes, including weight, firmness, and lycopene content, long before harvest, the researchers said.

The study’s findings were recently published in the peer-reviewed journal, Computers and Electronics in Agriculture.

Using a hyperspectral camera, the team analyzed 567 tomatoes from five different cultivars.

Hyperspectral imaging captures light across various spectral bands, enabling detailed insights into the physical and chemical properties of the fruit.

Machine learning algorithms such as Random Forest and Artificial Neural Networks processed this data to predict seven essential parameters, including total soluble solids, citric acid, ascorbic acid, and pH.

One key breakthrough was the efficiency of band selection. By requiring only five spectral bands for accurate predictions, the model opens the door to developing affordable, portable devices for widespread use.

The technology can be adapted for other produce and opens up the possibility of farmers monitoring produce quality during the ripening stages, enabling optimal harvest timing and enhancing crop quality.