This technology is an autonomous robotic system that uses optical scanning, near-infrared spectroscopy, and machine learning to detect, identify, and map microparticles in the environment in real time, enabling efficient and accurate monitoring of microparticles, including microplastics.
Background
The field of environmental monitoring has become increasingly important as concerns about pollution, particularly microplastic contamination, continue to grow. Microplastics—tiny plastic fragments less than 5 millimeters in size—are pervasive in terrestrial and aquatic ecosystems, posing risks to wildlife, human health, and the broader environment. Accurate, large-scale mapping and classification of these particles are essential for understanding their sources, distribution, and ecological impact.
Traditional monitoring methods, which often rely on manual sample collection and laboratory analysis, are labor-intensive, time-consuming, and limited in spatial coverage. As a result, there is a pressing need for technologies that can provide rapid, reliable, and comprehensive data on microplastic pollution directly in the field.
Current approaches to microplastic detection and classification face several significant challenges. Manual sampling and laboratory-based spectroscopic analysis, such as Fourier-transform infrared (FTIR) or Raman spectroscopy, are slow and require specialized equipment and expertise, making them impractical for routine or large-area surveys. These methods are further hindered by the complexity of environmental samples, which often contain a mixture of plastics and natural materials like water and plant matter.
Optical identification alone is unreliable due to the visual similarity between plastics and other particulates, while spectroscopic techniques can be confounded by overlapping signals and environmental noise. Additionally, most existing systems lack automation and real-time data processing capabilities, limiting their scalability and responsiveness to dynamic environmental conditions.
These limitations underscore the need for more robust, autonomous, and accurate solutions for in situ microplastic monitoring.
Technology description
This technology is an integrated, autonomous system designed for the real-time detection, identification, and mapping of microparticles, especially microplastics, in environmental settings. It features a two-stage scanning process: an initial optical scan uses machine vision to rapidly identify potential microparticles within complex natural backgrounds, followed by a targeted near-infrared (NIR) spectroscopic analysis that determines the chemical composition of each detected particle. The system is mounted on a robotic rover capable of autonomous navigation across diverse terrains, such as beaches and coastlines.
Advanced machine learning models, including neural networks and support vector machines, process both the optical and NIR data, enabling accurate classification of polymer types and differentiation from environmental interferents like water and plant matter. Adaptive path planning algorithms further optimize the rover’s survey routes, focusing data collection on areas with higher concentrations of microplastics, while all findings are logged for detailed mapping and ongoing model improvement.
What differentiates this technology is its robust, field-ready approach to microplastic monitoring, overcoming the limitations of traditional, labor-intensive laboratory methods. The use of machine learning models trained with synthetically augmented data—including environmental noise and interferents—ensures high accuracy and resilience in real-world, contaminated conditions, achieving reliable detection even at low signal-to-noise ratios. The fiber-based NIR spectrometer, designed for reflection and backscattering measurements, allows for effective analysis of irregularly shaped particles as small as 100 microns directly in situ.
By combining autonomous operation, real-time data analysis, and adaptive navigation, the system enables scalable, cost-effective, and comprehensive environmental surveys. Its modular design and data-driven approach also allow for future adaptation to other types of chemical pollutants and continuous improvement through iterative field deployment, making it a significant advancement in environmental monitoring and pollution mitigation.
Benefits
- Enables real-time, in situ detection and classification of microplastics, eliminating the need for manual sample collection and laboratory analysis
- Autonomous robotic operation allows efficient, large-area environmental surveys with minimal human intervention.
- High accuracy and robustness in complex, noisy environments through machine learning models trained with augmented data including environmental interferents
- Two-stage scanning system combines wide-area optical pre-screening with precise near-infrared spectroscopic analysis for reliable microparticle identification down to 100 microns.
- Adaptive path planning optimizes data collection by focusing on areas with higher microplastic concentrations, improving survey efficiency.
- Comprehensive data logging supports detailed pollution mapping, statistical analysis, and iterative improvement of detection algorithms.
- Scalable platform adaptable for surveying various chemical compositions and deployable in diverse environmental settings.
- Supports environmental monitoring, regulation, and mitigation efforts by providing crucial data on microplastic pollution distribution and composition
Commercial applications
- Autonomous beach microplastic pollution mapping
- Real-time waterway microplastic monitoring
- Industrial site polymer contamination detection
- Large-scale environmental plastics surveying
- General water treatment, image acquisition & preprocessing, image analysis, image enhancement or restoration, investigating characteristics of particles, neural networks, optical analysis of materials, other data recognition, sensor fusion, spectrometry
Additional information
This autonomous robotic system surveys and classifies microparticles. It employs a two-stage process: optical pre-screening with machine vision, followed by focused near-infrared spectroscopy for chemical composition analysis. Real-time machine learning identifies polymer types, supported by robust data augmentation for environmental noise. Adaptive path planning optimizes large-area mapping of particles down to 100 microns.
Publications
