Liquid-Level-Detection

Vision System

Problem D of the 28th Electronics Design Contest: A beverage cup measuring device based on monocular vision.

This project features a monocular vision system and STM32 communication pipeline. I implemented a dual-model YOLOv5 detection framework with IoU-based tracking, EMA smoothing, and a custom liquid surface detection algorithm using blob analysis and gradient scoring. On the STM32 side, I established a reliable UART communication protocol at 115200 baud for real-time vision data transmission, fused with INA240A1 current sensor readings, and drove an SSD1315 OLED display for status output. The integrated pipeline processes high-resolution camera frames and delivers stable measurement results every three frames.

MaixCAM YOLOv5 STM32 Computer Vision UART Embedded