Array-optimized artificial olfactory sensor enabling cost-effective and non-destructive detection of...

Area of Science:

• * Agricultural Science
• * Sensor Technology
• * Artificial Intelligence

Background:

• * Mycotoxin contamination in maize poses significant food safety risks.
• * Traditional detection methods are often costly, time-consuming, and lack on-site capabilities.
• * Artificial olfactory sensors (AOS) offer potential for rapid, low-cost aroma detection.

Purpose of the Study:

• * To develop a low-cost, high-precision system for detecting mycotoxin-contaminated maize using an artificial olfactory sensor.
• * To optimize the sensor array size while maintaining classification accuracy through an improved sine-cosine algorithm (ISCA) integrated with MobileNetV3.
• * To achieve real-time performance for on-site food safety applications.

Main Methods:

• * Volatile organic compounds from maize were analyzed using an unoptimized AOS with 16 sensor elements (porphyrins and dye-attached nanocomposites).
• * A MobileNetV3 model was employed for classification, achieving over 98.5% accuracy.
• * The ISCA-MobileNetV3 algorithm optimized the AOS by reducing the sensor array from 16 to 6 elements with minimal accuracy loss.

Main Results:

• * The ISCA-MobileNetV3 system successfully reduced the olfactory array size by over 60% (from 16 to 6 sensors).
• * Classification accuracy remained high, with only a ~1% decrease compared to the full array.
• * Online evaluations were completed in under one second, demonstrating excellent real-time detection capabilities.

Conclusions:

• * The integration of AOS with ISCA-MobileNetV3 provides an effective and efficient method for mycotoxin detection in maize.
• * This approach significantly reduces cost and complexity by optimizing sensor array size.
• * The developed system facilitates the creation of affordable, on-site platforms for enhanced maize quality control and food safety.