Detecting the sweetness of fruits has always been a challenge, requiring physical examination or taste testing. However, Pakistani experts have revolutionized this process by developing a cutting-edge system. That can determine the sweetness of citrus fruits without the need to open them. This groundbreaking method utilizes artificial intelligence (AI) and visual properties to accurately predict the sweetness of fruits, such as oranges and kino, with an impressive 80% accuracy rate.
Led by Dr. Ayesha Zeib from the National Center for Robotics and Automation at the National University of Science and Technology (NUST). A team of scientists and researchers undertook this significant endeavor. Their objective was to devise a non-invasive technique that would safeguard the fruits while allowing for reliable sweetness assessment.
To accomplish this, the team employed infrared spectroscopy, a process that analyzes the interaction between light and matter. By using a manual spectrometer that emitted near-infrared light. The scientists collected the spectral patterns of 92 fruits, including mosambi, red-orange, and sweet oranges. Circle marks were placed on the fruits, and the corresponding spectrum was recorded from these areas. Sixty-four fruits were used for calibration, while the remaining 28 were used for prediction.
Spectroscopy, a powerful tool in scientific analysis, enables the examination of objects based on their interaction with light. Although traditionally used for a wide range of applications, including medical diagnoses and astronomical observations, spectroscopy had not been extensively applied to assess fruit quality until now. Pakistani scientists broke new ground by successfully utilizing spectroscopy to directly determine the sweetness of local oranges, marking a significant advancement in the field.
Typically, fruit quality is measured through chemical and sensory analysis, with Brix representing the sugar content and Titratable Acidity (TA) indicating the fruit’s sourness. To establish reference data, the team used NIR spectroscopy to obtain readings and then compared them to conventional chemical and taste tests. By training an AI algorithm with data from 128 samples, the team achieved exceptional results. The AI model accurately estimated the sweetness of fruits, surpassing expectations.
To further validate the AI model’s predictions, an additional 48 fruits were introduced to the system. Through a meticulous process involving chemical analysis and taste testing, the researchers confirmed the accuracy of the model’s estimations. The reliability of the system was reinforced by the fact that external factors like low light or dust on the fruit’s surface did not hinder the spectrometer’s measurements. Thanks to the near-infrared rays’ ability to penetrate the fruit’s peel and reach the sensor.
Notably, the computer model not only provided reliable predictions of sweetness comparable to Brix and TA measurements. But also achieved an impressive 81% accuracy in determining overall flavor profiles, including sweet, mixed, and sour tastes. Such advancements are vital for the citrus industry. As fruits with high sweetness levels are in high demand for various applications. Such as juice production, beverages, and preserves. By adopting this innovative method. Pakistani exports of oranges, cano, and mosambi can be further enhanced, contributing to the growth of the industry.
Unlike mangoes and bananas, which continue to ripen after harvesting, oranges and canoes do not undergo further ripening. This makes the Pakistani innovation particularly valuable, as it ensures the production of high-quality fruits. With the successful implementation of this technology, Pakistani citrus exports have the potential to experience substantial growth. By leveraging scientific expertise and cutting-edge AI, Pakistani scientists have opened new horizons in the realm of fruit quality assessment, paving the way for enhanced productivity and economic opportunities.