Video smoke recognition method based on dynamic features

Abstract

The invention discloses a video smoke recognition method based on dynamic features. The method has the advantages of both real-time performance and accuracy. The method comprises steps: (1) continuouskey frames of an input video are firstly extracted, noisy points are filtered through a median filter, and the definition is improved through contour mending; (2) through foreground difference processing on the continuous frames, modeling is carried out on moving targets, a VR feature classifier is used to extract dynamic features, the moving targets are classified in combination of typical dynamic features, and a smoke target is recognized; and (3) finally, the smoke target and a typical smoke image are subjected to similarity comparison, thereby improving the recognition accuracy. The videosmoke recognition method based on dynamic features is characterized in that as the VR feature classifier is designed, the target dynamic features in a complicated scene can be effectively extracted,interference of people, cars, trees, flags and the like is removed through target dynamic feature differences, the smoke target is accurately recognized, and a smoke recognition method with strong adaptability, real-time performance and accuracy is provided and can be used for smoke early warning at an early stage of fire in a complicated environment.

Description

technical field [0001] The invention belongs to the technical field of computer image processing, and in particular relates to a video smoke recognition method based on dynamic features, which can be applied to smoke and fire early warning and monitoring software systems. Background technique [0002] Fire is a frequent disaster, which often causes a large number of casualties and property losses. Therefore, timely detection of fire signs and early warning can save huge human and property losses and generate huge economic and social benefits. However, traditional manual monitoring is inefficient, costly, and has artificial false positives and false positives; although fire detection methods based on sensors such as smoke and temperature have made progress, they are seriously affected by factors such as physical environment, monitoring range, and hardware costs. , leading to problems such as high cost and high false alarm rate. For example, when the temperature sensor detect...

AI Technical Summary

Problems solved by technology

[0012] The purpose of the present invention is to propose an effective smoke recognition method based on dynamic features in view of the shortcomings of the existing smoke recognition technology, such as low accuracy, insufficient real-time performance and poor applicability.

Images