Smoke/flame detection method based on video image analysis

Abstract

The invention relates to a smoke / flame detection method based on video image analysis. The existing fire monitoring mode is limited by a plurality of geographic conditions and infrastructure constructions. The method comprises the following steps: firstly performing the motion detection on an acquired video image by using a Kalman filter; secondly analyzing the colors of smoke and flame to obtain a smoke pixel region and a flame pixel region; then using a decision machine to determine whether a corresponding region is detected, and extracting the feature of the determined region; and finally performing normalization conversion on the extracted feature so as to adapt to hidden Markov model flame detection and neural network smoke detection. The outdoor smoke and flame can be detected simultaneously by using the smoke / flame detection method based on the video image analysis so as to warn the forest fire; and the effect predicting the forest fire at an early stage is achieved, and the economic loss caused by the fire can be reduced.

Description

technical field [0001] The invention belongs to the technical field of intelligent fire monitoring, and in particular relates to a method for simultaneously detecting intelligent flame and smoke based on video images. Background technique [0002] With the development of social productivity and increasing social wealth, the scope of fire losses and fire hazards is constantly expanding. Fire not only causes huge losses to human economic construction and destroys the ecological environment, but also threatens the safety of people's lives and property, and affects the stable development of society. At present, my country's forest fire monitoring mainly includes ground patrol, watchtower monitoring, cruising aircraft and meteorological satellite remote sensing monitoring, etc. These fire monitoring methods have their own shortcomings: on the one hand, due to the complex terrain and vast area of ​​the forest area, it is unrealistic and unreliable to rely solely on patrol personn...

AI Technical Summary

Problems solved by technology

These fire monitoring methods have their own shortcomings: on the one hand, due to the complex terrain and vast area of ​​the forest area, it is unrealistic and unreliable to rely solely on patrol personnel to monitor the fire situation. Due to the limitation of facilities, relying on aircraft patrol observation is not only costly but also inefficient; on the other hand, due to the complex terrain conditions in most forest areas in my country, lack of power supply and corresponding communication facilities, etc. Moreover, the above fire monitoring methods are not suitable for the scattered characteristics of small and medium-sized forest areas in my country, and their practicability and scalability are limited.

Images