VitalAir: IoT-Based Intelligent System for Harzard Detection in Vehicles

Abstract

This paper presents VitalAir, a novel standalone IoT safety system that provides critical lifesaving assistance inside a closed vehicle cabin. Automobiles left idle with commuters onboard are exposed to two life-threatening yet underserved hazards: cabin heatstroke and toxic accumulation of CO2 inside a sealed cabin. VitalAir addresses this gap through a multi-sensor, AI-enabled embedded architecture built around an MH-Z19E NDIR CO2 sensor, an AMG8833 8×8 thermopile thermal imaging array, and a DHT22 temperature-humidity sensor, all interfaced to a low-cost ESP32 dual-core microcontroller. An on-device quantized decision-tree classifier analyzes the fused sensor data to detect occupant presence and breathing state without cloud dependency, triggering a graded three-level alert (Warning, Danger, Emergency) delivered via Wi-Fi and GSM within 10 seconds of hazard detection. The system was validated under varying temperature, humidity, and occupancy conditions across different geographical settings, achieving 88.6% classification accuracy with 50 ms inference latency and zero false alarms in unoccupied vehicles. With a target retail price below Rs. 3,000 and compatibility with any vehicle make or model, VitalAir offers an affordable, technologically differentiated solution to a preventable cause of death.