Fire passage occupation image recognition and cloud linkage alarm method

Abstract

The application belongs to the field of intelligent security and protection, and discloses a fire passage occupation image recognition and cloud linkage alarm method; video stream and environmental illumination data are synchronously collected through an edge computing node, motion foreground is extracted by using adaptive background modeling, and cross-frame association and coarse-grained classification are performed; targeted image enhancement is implemented in combination with real-time illumination level, edge preliminary judgment confidence score is calculated from multiple dimensions such as spatial occupation, continuous duration and motion state; a cloud platform performs fine-grained semantic segmentation and multi-modal feature fusion analysis on a suspected event package, and introduces a historical false alarm feature library to complete fine filtering and occupation confirmation; finally, a hierarchical alarm linkage strategy is triggered according to the type and duration of the occupation object. The application effectively reduces the recognition false alarm rate in a complex environment, and significantly improves the intelligent level and disposal efficiency of fire safety management.

Images