Get the Latest News Updates

Pakistani scientists use artificial intelligence to figure out how sweet orange fruits are.

A group of Pakistani scientists has made a big scientific step forward by creating a visual classification method based on artificial intelligence (AI) that can correctly tell how sweet local citrus fruits are.

The team, which was led by Dr. Ayesha Zeb of the National Centre of Robotics and Automation at the National University of Science and Technology (NUST), was able to predict the sweetness of fruit with more than 80% accuracy without hurting the fruit.

For their experiment, the researchers chose 92 citrus fruits from a farm in the Chakwal area, such as Blood Red, Mosambi, and Succari. They used a portable spectrometer to get spectra, which are patterns made by the way light bounces, from certain spots on the skin of the fruits. The team used near-infrared (NIR) spectroscopy, a method for analyzing light spectra that can’t be seen, to look at the fruit samples. Out of the 92 vegetables, 64 were used to set up the spectrometer and 28 were used to make predictions.

The Pakistani team took a new method by using NIR spectroscopy to model the sweetness of local fruits. This is not the first time that NIR spectroscopy has been used to classify fruits that have not been damaged. They also used formulas based on artificial intelligence to directly classify the sweetness of oranges, which improved the accuracy.

Usually, chemical and taste tests are used to figure out how sweet a fruit is. Total sugars, which are measured by Brix, show how sweet an orange is, while titratable acidity (TA) shows how much citric acid it has. For the AI model, the team got reference values for Brix, TA, and fruit sweetness by pulling off samples from the marked areas used for spectroscopy.

The real Brix and TA values were found when the juice from the samples was tested in a lab. Also, people tasted the fruits and sorted them into three groups: flat, sweet, and very sweet.

The team taught the AI algorithm on a total of 128 samples using the spectrum, reference values, and sweetness labels they had found. Based on spectral data, the AI model was made to be able to predict Brix, TA, and sweetness levels. To find out how accurate the model was, researchers used data from 48 new fruits to test it. They compared the values predicted by the model with the real values found through sensory evaluations and chemical analysis.

The results were surprising because the AI model correctly predicted the values of Brix, TA, and overall sweetness. It also did a better job of predicting sweetness than standard methods. The model was able to tell the difference between sweet, mixed, and sour tastes 81.3% of the time.

This scientific breakthrough is important for the citrus business, especially for figuring out how good citrus fruit is. Oranges don’t keep getting sweeter like bananas and mangoes do after they’ve been picked from the tree. So, this new AI-based method could make it easier and better to judge how sweet citrus fruits are, which would help the business and make sure customers are happier.

Pakistan, which will be the sixth-largest exporter of citrus goods in the world in 2020 with 0.46 million tons, stands to benefit from this change.

The results of this study were published in Nature, which is a well-known research magazine.

Dr. Ayesha Zeb and Dr. Mohsin Islam Tiwana from the National Centre of Robotics and Automation at the National University of Sciences and Technology (NUST); Dr. Waqar Shahid Qureshi from the School of Computer Science at Technological University Dublin, Ireland; Drs. Abdul Ghafoor, Muhammad Imran, and Alina Mirza from the Military College of Signals at NUST; and Dr. Amanullah Malik from the Institute of Horticultural Sciences