Multimodal Sensors Classify Durian Ripeness With 96.91% Accuracy

*Important notice: This news reports on an unedited version of an accepted paper and is awaiting final editing. Therefore, the paper should not be regarded as conclusive or treated as established information.

Evaluating the exact ripeness of durian without damaging the fruit has long been a challenge, as traditional methods such as manual tapping and visual inspection are subjective and often lead to inconsistent quality control and post-harvest losses. In a recent study published in the journal Scientific Reports, researchers developed a non-destructive multimodal sensor system powered by artificial intelligence (AI) to classify durian maturity.

Durian fruit close-up in front of a market stall
Study: A multimodal IoT sensor platform analyzes durian gas emissions, surface temperature, and tapping sounds to support automated, non-destructive ripeness assessment. Image Credit: Marius Karp/Shutterstock.com

By integrating gas detection, thermal imaging, and acoustic analysis within an Internet of Things (IoT) architecture, the system assesses fruit maturity in real time. Using a neural network, it achieved a classification accuracy of 96.91% with an area under the curve (AUC) of 0.98, providing a reliable alternative to destructive methods such as fruit cracking.

Traditional Limitations in Durian Quality Control

Determining the maturity of climacteric fruits like Durio zibethinus (durian) is challenging due to internal physiological changes beneath a thick, spiky pericarp. Traditionally, post-harvest quality assessments have relied on subjective methods, including visual inspection, manual tapping, and odor evaluation, which often result in inconsistent quality control.

Saving this for later? Download a free PDF copy now.

Agricultural automation addresses these limitations through non-destructive multimodal sensor systems that integrate gas sensors, thermal imaging, and acoustic transducers. These sensors measure volatile chemical emissions, surface temperature distributions, and internal mechanical properties associated with fruit ripening.

Custom Sensor Framework for Durian Evaluation

To evaluate durian maturity non-destructively, researchers developed a custom acrylic chamber controlled by an ESP32 microcontroller for real-time data acquisition and wireless cloud transmission.

The system integrated four gas sensors: an MQ-3 sensor for alcohol detection, a TGS2620 sensor for volatile organic compounds (VOCs), an MG-811 sensor for carbon dioxide, and an oxygen sensor for monitoring oxygen consumption.

An MLX90640 thermal camera captured infrared surface temperature maps, while a servo-driven tapping arm paired with an Analog Sound V2 sensor recorded acoustic responses.

The study evaluated 90 Monthong durians across three maturity stages, measured in days after anthesis: unripe (80–100 days), half-ripe (101–110 days), and ripe (>110 days). Each fruit was measured at five surface locations, generating a total of 450 observations under controlled conditions (25 °C and 60–70% relative humidity).

Destructive testing validated the maturity, showing that flesh firmness decreased from 12.44 N in unripe fruit to 5.47 N in ripe fruit and 8.95 N in half-ripe fruit. Soluble solids concentration increased from 3.84 °Brix to 6.61 and 9.64 °Brix, and alcohol content also rose from 0.80% to 2.27% and 4.43%.

Data was preprocessed using min-max normalization and used to train a multilayer perceptron neural network with three hidden layers, dropout, and early stopping. To avoid data leakage from measurements of the same fruit, researchers applied fruit-level grouped 10-fold cross-validation and compared the model with six machine learning algorithms.

Performance and Contributions of Sensor Modalities

The multilayer perceptron neural network outperformed the six other supervised machine learning models, correctly classifying 157 of 162 validation samples for an overall accuracy of 96.91%.

The unripe fruits were identified with perfect precision, recall, and F1-score (1.00). The half-ripe class achieved precision and recall values of 0.94 and 0.96, respectively, while the ripe class recorded precision and recall values of 0.96 and 0.94. Misclassifications occurred only between adjacent half-ripe and ripe stages.

Ablation analysis revealed that the gas and VOC sensor array was the primary contributor to classification performance, achieving 100% standalone accuracy. The thermal imaging module provided complementary information, reaching 99.41% accuracy.

In contrast, the acoustic sensor achieved only 52.94% accuracy, indicating that sound pressure levels alone cannot reliably detect internal structural changes without frequency-domain analysis.

Applications in Commercial Durian Processing

With further development and real-world validation, this integrated sensor framework has applications in commercial fruit packing, processing, and international distribution. By combining gas, thermal, and acoustic sensors with cloud-connected microcontrollers, processing facilities can automate maturity assessment and sort durians by physiological readiness before packaging.

Real-time quality monitoring enables logistics managers to optimize shipping schedules, cold-chain storage, and long-distance transport. Therefore, this framework could help fruit reach consumers at peak quality while reducing post-harvest losses. Miniaturizing the sensor into portable handheld devices would support field harvesting and quality inspection, potentially allowing agricultural workers to assess tree-bound or freshly harvested fruit without cutting open the thick husk.

Directions for Advanced Agricultural Sensing

This study demonstrates that combining gas, thermal, and acoustic sensors with a neural network provides a scalable approach for non-destructive durian maturity assessment. Future work should focus on expanding the dataset to include additional durian cultivars, validating performance under variable field conditions, and incorporating frequency-domain acoustic features to improve characterization of internal fruit structure.

Overall, scaling this multimodal sensor platform into commercial equipment could help standardize automated quality assurance across the tropical fruit industry.

Journal Reference

Brilliantina, A., et al. (2026). Development of non-destructive durian fruit maturity detection tool based on multi-variable sensor for harvest quality optimisation. Scientific Reports. https://www.nature.com/articles/s41598-026-61395-9

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.

Muhammad Osama

Written by

Muhammad Osama

Muhammad Osama is a full-time data analytics consultant and freelance technical writer based in Delhi, India. He specializes in transforming complex technical concepts into accessible content. He has a Bachelor of Technology in Mechanical Engineering with specialization in AI & Robotics from Galgotias University, India, and he has extensive experience in technical content writing, data science and analytics, and artificial intelligence.

Citations

Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Osama, Muhammad. (2026, July 22). Multimodal Sensors Classify Durian Ripeness With 96.91% Accuracy. AZoSensors. Retrieved on July 22, 2026 from https://www.azosensors.com/news.aspx?newsID=16910.

  • MLA

    Osama, Muhammad. "Multimodal Sensors Classify Durian Ripeness With 96.91% Accuracy". AZoSensors. 22 July 2026. <https://www.azosensors.com/news.aspx?newsID=16910>.

  • Chicago

    Osama, Muhammad. "Multimodal Sensors Classify Durian Ripeness With 96.91% Accuracy". AZoSensors. https://www.azosensors.com/news.aspx?newsID=16910. (accessed July 22, 2026).

  • Harvard

    Osama, Muhammad. 2026. Multimodal Sensors Classify Durian Ripeness With 96.91% Accuracy. AZoSensors, viewed 22 July 2026, https://www.azosensors.com/news.aspx?newsID=16910.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type
Submit

Sign in to keep reading

We're committed to providing free access to quality science. By registering and providing insight into your preferences you're joining a community of over 1m science interested individuals and help us to provide you with insightful content whilst keeping our service free.

or

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.