Smoke Detection Research Based On Video Image Processing

In recent years, with the high frequency of fire accidents, it's difficult to raise a fire alarm real-time through the conventional fire detection techniques which mainly focus on the physical phenomenon and/or the features of chemical reaction products when flammable materials are burning. For these conventional fire detection techniques which are easily affected by its surroundings and the equipments themselves, we found that not only did they hardly implement long-distance monitoring, but they require a massive investment of capital to update equipments.Therefore, how to improve the intelligent level of fire monitoring has become an important issue for today's study.In order to solve the problem of real-time performance of fire surveillance system, a fire alarm system based on video smoke detection was put forward to substitute the conventional fire detection techniques. The proposed fire detection system has an extremely extensive application prospect in the market and is of great significance. We developed the proposed fire detection system on Windows7 operating system and using MATLAB image software pack. Firstly, graying and de-noising pretreatment were carried out for images captured by the video. Secondly,getting the smoke region based on a single Gauss Background Model. Then,analyzing and extracting the features of suspected smoke, including short-term and stable movement characteristics, blurred background characteristics, irregular shape features, diffusion characteristics, color characteristics, etc. A two-dimensional wavelet transform was adopted to analyze the background features blurred by the smoke. The degree of circularity was used to measure the features of irregular shape.The growth rate of average absolute area was carried out to measure the diffusion characteristics. Finally, detecting smoke and making warnings with the integration of multiple features.The experimental results show that the algorithm of multi-feature fusion can identify smoke in video accurately and have a great effect in fire warning. At the endof this paper, we summarized the design for the whole system and discussed the related problems.