Research On The Recognition Of Fire Smoke Based On ARM And OpenCV

With the characteristics of sudden and randomness, the fire has not onlyhigh occurrence rate but also extreme destructiveness. How to make fire alarmquick and effective and how to reduce the losses caused by fire become theimportant content of the prevention and research of the fire.Smoke is one of the obvious characteristics which is produced in the earlystage of the fire and could be detected easily; smoke detection based on videois an important topic both on theoretical significance and practical value. Withthe rapid development of embedded technology, it becomes more and morewidely used in reality that the front-end detection based on the embeddedplatform with the advantage of convenient installation, low hardware cost andgood real-time performance.The thesis studies a detecting method of smoke recognition which isbased on ARM and OpenCV. The circuit of video capturing and datacommunication is designed taking ARM as the core. Linux operating systemand OpenCV is transplanted to S3C2440development board, and the smokefeatures are extracted and the smoke is recognized by using the OpenCV function. Firstly, an algorithm of smoke segmentation was proposed based onfrequency field and fractal property, with the collected image4-layerdecomposed using DB4wavelet basis and the proper processing of waveletcoefficient of each layer, which is combined with differential box-counting(DBC) fractal dimension operation. Then according to the characteristics offire smoke, the features of smoke are extracted, including the fractaldimension(DBC), the direction of motion, the color feature and densitydistribution. Next, the principle of the algorithm, preferences and the modelbuilding of Support Vector Machine (SVM) are introduced and the algorithmof smoke recognition is proposed based on SVM. Lastly, the SVM gives thepredicted classification of the rest of the samples after trained by training set.The results show that the effect of the algorithm of smoke recognition is good.The thesis makes full use of the advantage of embedded system anddesigns both software and hardware, making the smoke monitoring based onvideo carried out. It provides certain significance for video monitoring ofsmoke based on ARM and OpenCV.