I am developing a real-time air pollution monitoring system using the Maixduino microcontroller with MicroPython. The system incorporates a custom 3D-printed enclosure with a black, dark chamber that contains a laser source positioned perpendicular to the Maixduino’s camera module. This setup utilizes Dynamic Light Scattering (DLS) to measure particle sizes. In this configuration, the laser illuminates particles that pass through the dark chamber, causing them to reflect light and appear as white dots on the camera feed. The camera captures these reflections as glowing contours against the dark background.

The Maixduino processes the camera feed frame by frame to detect and count these contours. By analyzing the number and size of the reflections, the development board can estimate the concentration and size of airborne particulates. This data is crucial for accurate air pollution prediction.

A Regression model will be trained using this data from the microcontroller along with pollution level data from an existing high accuracy air pollution sensor. The data from an existing pollution sensor will act as label for each and every individual frame of time.
This model, once trained, will be deployed on the Maixduino to predict pollution levels in real-time. TF lite will be used for the model training and will be deployed on the Maixduino.

The pollution levels will be displayed on the Maixduino’s screen, providing immediate visual feedback. Additionally, the system transmits the data to a Google Firebase-based map dashboard, enabling large-scale pollution monitoring and aggregation from multiple sensors, thus facilitating comprehensive environmental analysis.