Flame detection device based on video image analysis

Abstract

The invention discloses a flame detection device based on video image analysis. The flame detection device based on video image analysis comprises an acquisition system for acquiring video images, a video image analysis system and a fire alarming system. The video image analysis system comprises a video image preprocessing module, a moving target recognition module, a flame pixel recognition module and a fire judgment module. The acquired video images are converted from RGB (red, green and blue) color space to CIE (international commission on illumination) xyY color space and are compared with distribution characteristics of marked flame colors on a CIE xy chromaticity diagram, so that flame pixels are recognized. Converted CIE xyY data are input into the flame detection module and the moving target detection module, and binary images output from the two modules are fused for detection of a fire. The flame colors has high resolution and uniqueness in the CIE xy color space, so that the flame detection device has high flame recognition precision and wide range of application.

Description

technical field [0001] The invention relates to a flame detection device based on video image analysis, in particular to perform flame detection and alarm by analyzing pixel color features in real-time collected video images, and belongs to the technical field of intelligent fire monitoring. Background technique [0002] Fire monitoring and early warning device systems play a very important role in many fields, including large building fire prevention, forest fire prevention, and natural environment monitoring. Traditional fire monitoring technologies and devices include particle-type smoke sensors, infrared and laser technologies, etc. Particle-type smoke sensors require smoke particles to enter the sensor to cause an alarm. Infrared and laser technology also require smoke occlusion to trigger an alarm. In addition, for large-scale buildings and outdoor environments, a large number of sensor devices need to be deployed to achieve high monitoring coverage. rate and monitori...

AI Technical Summary

Problems solved by technology

However, in these methods, some algorithms use traditional color spaces such as RGB and YUV for flame color feature recognition, and the misjudgment rate is high; some algorithms first perform flame feature analysis, and then perform moving target recognition, resulting in slow calculation speed. higher operating costs

Other algorithms analyze and detect dynamic features such as flame shape, and the algorithm structure is complex and the false positive rate is high.

Images