Low-cost portable microplastic detection system integrating nile red fluorescence staining with YOLOv8-based deep learning

Abstract

Microplastic (MP) pollution presents considerable challenges to aquatic ecosystems and human health, yet cost-effective detection methods remain limited. This study presents a portable, low-cost MP detection device combining Nile Red (NR) staining with YOLOv8-based deep learning (DL). The compact system (22 × 23 × 20 cm) uses a digital microscope, optical filter, 395 nm UV source, and Raspberry Pi 4 (RPi4) as the central processing unit. This design provide a portable and affordable alternative to expensive laboratory-based detection methods. In testing six common polymers (ABS, Nylon, PE, PET, PS, PVC), the system achieves 94.8 % mean average precision at IoU threshold of 0.5 (mAP@50), with excellent performance for PE and Nylon (96.5 %). Each polymer exhibits distinct fluorescence patterns enabling robust automated classification. Economic analysis demonstrates 77.3 % cost reduction compared to conventional FTIR methods, from $0.44 to $0.10 per sample, with fixed costs of only $139. The 19-second processing time enables high-throughput analysis suitable for field applications, citizen science, and resource-limited settings. Detection is limited to particles >100 μm by microscope resolution. This technology enhances MP analysis accessibility, bridging the gap between expensive laboratory methods and practical environmental monitoring for widespread community use.