Air Pollution Measured with Camera and ML in Particulate Matter Sensor

This advanced air quality monitoring system operates by combining optical scattering with machine learning regression analysis executed on a dedicated AI vision platform. At the heart of the system is a custom 3D printed enclosure that aligns the 445-455 nm 5V DC laser module and the Maixduino development board's camera module. A 4010 5V DC fan module draws a steady, controlled stream of air through the sensing chamber, passing the airborne particulate matter directly across the path of the laser beam. When individual particles intercept the coherent light from the laser, they scatter the beam in various directions. The onboard camera captures these optical scattering events in real time. Programmed using MicroPython via the MaixPy IDE, the Maixduino board processes the visual data stream through a pre-trained machine learning regression model. Instead of relying on traditional binary thresholds, the ML model analyzes the distinct diffraction patterns and intensity of the scattered light, allowing the system to accurately detect, count, and estimate the concentration of particulate matter. Power is supplied reliably to the entire circuit via a portable 1S 18650 power supply, while connection cables ensure secure electrical interfacing between the microcontroller, laser, and fan modules.