Flame dynamic behavior recognition method based on multi-modal time sequence modeling

By using multimodal temporal modeling technology, the microscopic dynamic features of flames are extracted and global temporal correlation is performed, which solves the problems of response delay and high false alarm rate in traditional fire detection. It enables accurate identification and quantitative analysis of fire development trends, and improves the reliability and accuracy of fire early warning.

CN122454239APending Publication Date: 2026-07-24SHANXI ZHIPU TECHNOLOGY CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202610317642.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-16
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

Existing fire detection technologies suffer from problems such as response delay, high false alarm rate, and lack of fire evolution trend recognition capabilities in complex environments. Especially in industrial scenarios with complex lighting and fire-like interference, traditional visual detection algorithms cannot effectively distinguish between flames and interference objects, and cannot provide dynamic trend analysis of fire development.

Method used

A multimodal temporal modeling approach is adopted, which extracts the micro-dynamic features of flames through convolutional long short-term memory networks and transformer networks. Combined with global temporal correlation modeling, a binary mask of the flame region is generated, and the flame area change rate and acceleration are calculated to construct a fire situation threat index, thereby realizing graded early warning of fire risk.

Benefits of technology

It significantly advances the warning time point, reduces the false alarm rate, can accurately detect fires in their infancy, provides a quantitative assessment of fire development, and provides a scientific basis for fire protection decisions.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a flame dynamic behavior recognition method based on multi-modal time sequence modeling, which comprises the following steps: preprocessing a video frame, and extracting multi-modal spatial features and spatial detail features; flame edge microcosmic dynamic features are extracted through a convolutional long short-term memory network, and global time sequence correlation modeling is carried out through a converter network to obtain global features; generating a binary mask of the flame region based on the global features and the spatial detail features; calculating the pixel area of the flame area based on the binary mask, and differentiating the change of the pixel area to obtain a spreading rate and an acceleration; and constructing a fire situation threat index based on the spreading rate and the acceleration, and carrying out graded early warning on the fire risk. The method has the beneficial effects that the early warning time node is obviously advanced, the false alarm rate of a security system is greatly reduced, scientific data support is provided for fire emergency decision making, and the accuracy of area statistics and trend analysis is ensured.
Need to check novelty before this filing date? Find Prior Art