Video smoke detecting method based on color invariance

Abstract

The invention provides a video smoke detecting method based on color invariance. The method comprises the steps of firstly adaptively updating an image sequence background and obtaining a real-time background image; then performing normalization processing on a current image and the background image for obtaining the image with translation invariance and scale zooming invariance; performing motion area detection on the normalized current image and the background image by means of an optical gain, and filtering the motion area without a smoke color by means of an optical characteristic of the smoke; and finally confirming the smoke area of the motion area from the motion area with the smoke color by means of a color invariance descriptor. The video smoke detecting method has advantages of low calculation amount, high real-time performance, high efficiency, and high robustness for illumination change and noise corrosion which commonly occur in an outdoor monitoring video.

Description

technical field [0001] The invention belongs to the field of pattern recognition, in particular to a video smoke detection method based on color invariance. Background technique [0002] Tens of thousands of fires occur every day around the world, killing hundreds of people. In order to be able to detect fires in time, relevant companies have designed various smoke detectors such as temperature sensing, smoke sensing, and light sensing according to changes in physical quantities such as smoke, gas, and temperature. The law of fire shows that the appearance of fire smoke is earlier than that of open fire. When the flame is small, it is easy to be blocked by trees, but smoke is not easy to be blocked. Therefore, smoke plays an important role in early fire detection. Smoke detection based on video surveillance has caused extensive attention. [0003] The patent application number is CN201510170159.2, and the Chinese patent titled "Video smoke detection method based on multi-f...

AI Technical Summary

Problems solved by technology

The above two methods mainly rely on the motion characteristics of the smoke, and rarely use the essential feature of the smoke color, and the extraction of the motion area only relies on simple background subtraction. When the illumination changes strongly, the detection accuracy of the motion area will be greatly reduced. , so that the accuracy of smoke detection will decrease

Images