MLX90640 - Student job at Advionics nv

Built during a student job at Advionics nv, this project is a production-deployed application for ingesting, visualising, and analysing live data from multiple MLX90640 far-infrared thermal camera modules simultaneously. The system went from prototype to active professional use on the factory floor.

What I Built

The application continuously polls each sensor, reconstructs the 32×24 pixel thermal frames using the melexis driver library, and renders colour-mapped heat maps with Matplotlib — all inside a responsive Kivy GUI. An automated alerting layer watches every frame and notifies operators the moment a pixel cluster crosses a configurable temperature threshold.

Key Features

• Real-time multi-sensor dashboard with per-camera heat-map views
• Configurable threshold alerts for abnormal temperature detection
• Change-point detection algorithm that identifies sudden shifts in thermal trends over time
• Packaged as a reusable Python package for straightforward deployment and version management
• Persistent logging of temperature data for post-hoc analysis

Technical Highlights

NumPy underpins all frame processing — vectorised operations keep per-frame computation fast enough to sustain smooth live rendering across multiple camera feeds. The change-point detection module applies statistical methods to rolling temperature windows, distinguishing genuine anomalies from sensor noise without producing excessive false positives.

The project taught me how to bridge low-level hardware I²C communication with a polished desktop UI, and how to structure a Python project for maintainability in a professional engineering context.