Design And Implementation Of The Smoke Detection System Based On Video Images

The fire accidents are often caused by the loss of economic and environmentalaspects, and even casualties. Especially in recent years, a greater range of fire is evencatastrophic fire accident-prone, and caused economic losses, casualties and damage tothe environment which is increasing year by year. Therefore, early warning anddetection of fires in fire prevention will reduce losses.In such case, the video-based fire detection has increasingly high degree ofconcernr. In view of the large amount of smoke is often first appeared in the early stagesof fires this feature.Video smoke detection can be widely applied to many aspects ofautomatic fire monitoring, security detection; forest fire prevention has become thefocus on attention of pattern recognition and image processing.Video smoke detection can be divided into the preprocessing of the input videostream image, the motion region extraction and smoke characteristics of the videoidentification. Smoke characteristics identify part-depth study of comparative analysis,on a variety of smoke recognition method, which summed up a combined two or threekinds of smoke recognition algorithm serial match smoke characteristics.Graspable, equilibration and other aspects of the preprocessing stage of the image,the image processing, and the wavelet transform (DWT, discrete wavelet transform)provide valid data in smoke recognition energy matching.Motion region extraction stage,frame difference, background subtraction and run in the background subtraction ofmeans method, the median method, good algorithm and real-time extraction of thebetter prospects for movement. Determined through experiments, the basic use of theimproved approximate median algorithm (Approximate median method), motion regionextraction.The smoke feature recognition stage is a central part the smoke feature recognitionalgorithm a direct impact on the accuracy and timeliness of the alarm. Existing smokefeature recognition method is roughly divided into the identification of the colorcharacteristics, the identification of the static characteristic of the smoke, and theidentification of the dynamic characteristics of the smoke. Specific recognition algorithm on the RGB color gamut, recognition algorithm is based on HIS modelcumulate and identify the main direction of movement.The experiments show that, the accumulated amount and the main direction ofmotion recognition combined with the motion history image (MHI) and unchangedfrom the identification of the characteristics can get good results. In addition, themethod of support vector machines (SVMS) has been used in smoke detectionalgorithm (SDA) to produce smoke alarm.