| Fire disaster is one of the biggest threats to human security. Fire detection ways based on traditional touching sensors can’t be applied in large-space building fires, plant specialties with bad environment, or forest fire detections. It is of great importance to research and exploit embedded analysis equipments based on video image to forecast fires on the open space to decrease the fire damages.In this paper, the fire detection system was based on high-performance processor TMS320DM6446and embedded Linux operating system. Compared with PC, embedded devices have low cost, low power consumption, small volume. And it is convenient to transport, install and maintain. The system mainly achieved the fire video image acquisition, image processing and fire recognition.This paper discussed the principle and process of video image fire detection algorithm. Firstly, background subtraction detection was made on those images from the camera, as with those detected suspicious graphics primitive, fire image segmentation was made on ways of YCbCr space, then flame characteristics from the segmentation were extracted and recognized.Also research on fire detection techniques was made using dual-core processor TMS320DM6446. This part mainly included:the Linux operating system was ran on ARM9kernel, programs like video image acquisition, image processing, and displaying were made with multithreaded programming ideas to coordinate the whole system; video image segmentation and recognition algorithm were studied on DSP kernel; the communication between that two cores was completed by Codec Engine.During the research, experiments were conducted on both PC and DM6446platform. Matlab software programming experiment was made on PC, it concludes that flames even in complicated environments can be well segmented out, and even those small areas can achieve good results. Moreover, this paper showed the experimental results on DM6446platform, that is, it can almost excellently segmented those fire flames out. |
PDF Full Text Request