Fire Flame Tracking Research Based On Video Image

The fire does great harm to human beings'life and property safety,so how to detect and put out the fire effectively has always been a problem that human beings research and probe. In recent years, as the development of computer figure and image processing technology, image type fire detection technology has got a lot of development. In this dissertation, by the aid of digital image processing techniques, dynamic flame tracking based on video image has been researched, which is a preparation for the work of putting out the fire.According to the features of flame image, the method of image difference based on dynamic sequential images and the method based on Mean-Shift segmentation algorithm combined with the threshold segmentation have been studied. Before the detection of flame target, pre-processing such as enhancement and filter must been done which in order to smooth and reduce the noise the video images. The method of flame recognition based on image difference detects the moving information by differencing two sequential frames or every two frames according to the speed of object. The Mean–Shift algorithm , which has the feature of fast iteration and clustering segment images according to detecting the pattern of density, is mainly used in static images. In this dissertation, the method of segmentation based on the Mean-Shift combined with threshold segmentation algorithm is applied to segment moving flame images.After confirming the flame region by image segmentation, by the analysis of general information of present tracking methods, this dissertation try to use a kind of improved Camshift algorithm, namely improved Continuously Adaptive Mean-Shift, to track the dynamic flame,which can overcome the shortcomings of artificial orientation and divergence problem of traditional Camshift algorithm during tracking. Emulation results show that the method can effectively track moving flame.