Video-based Fire Detection and Reporting System for Immense Urban Areas

Abstract

Fire disasters cause significant physical and emotional trauma worldwide, resulting in the loss of thousands of lives and billions of dollars in property damage. Delayed detection and reporting of fire incidents lead to slower response times, exacerbating the impact of fires. This study presents a video-based fire detection method integrated with a Geographic Information System (GIS)-based reporting platform aimed at enhancing real-world firefighting efforts. The fire detection method employed a hybrid approach, combining rule-based fire detection with a trained Support Vector Machine (SVM) classifier. The rule-based method utilized motion and color-based segmentation techniques, while the SVM classifier was trained using the chromatic and temporal features of fire objects. Test results showed that the module could detect fire in video streams with an overall classification accuracy of 99.66%, along with precision and recall rates of 99.38% and 99.97%, respectively, supported by an F1-score of 99.68%. This approach demonstrated excellent performance in predicting fires with minimal false alarms. Additionally, the GIS-based platform provides real-time information to responding agencies, further enhancing the effectiveness of firefighting operations.