A Smoke Detection Algorithm For Indoor Environment

According to recent statistics,a large number of fire accidents happen every year in the world,which cause immeasurable damage.China is a fire-prone country,almost severe fire accidents occur every year.Obviously,fire mitigation work has become a major issues that maintain social stability and keep people's lives and property safe.To address this important issue,we propose a image indoor fire smoke recognition algorithm,which is combined with the current video surveillance platform in each building and can quickly and accurately identify the fire smoke information to remind the relevant staff to confirm and process fire accident.This paper describes the shortcomings of traditional fire detection equipment in fire detection,such as lack of detection speed and large dependences on environment.This thesis sets forth the advantages of new image fire detection systems,and analyzes the research status of fire smoke algorithm at home and abroad.Moreover,a new indoor fire smoke image recognition algorithm is proposed and implemented.Indoor fire smoke recognition algorithm proposed in this paper mainly includes three modules: video images preprocessing,moving object extraction and characteristics of smoke extraction.The main purpose of video image pre-processing module is reducing image noise in order to get reliable image source.The moving object extraction algorithm can provide target recognition analysis with an object of study.In order to take into account the needs of real-time algorithm,moving object detection algorithm used in this paper is ViBe algorithm.Fire smoke characteristics contains static and dynamic characteristics,including smoke color characteristics,smoke ambiguity characteristics,smoke energy curves characteristics and smoke movement upward direction characteristics.A combination of these four characteristics can distinguish between smoke and non-smoke object,which ultimately completes identification of fire smoke.Finally,we give the system distributed clusters architecture design and the algorithm structure design.Many video experiments and field environmental testing laboratories are carried out with the proposed algorithm.Three performance indicators are put forward to measure the performance of the algorithm: system response time,the average time cost by processing each frame and smoke occupied areas of the image pixel ratio.The experimental results show that: proposed algorithm in the overall performance is excellent and better real-time.