Research On Fire Detection Technology Using Ordinary CCD Camera

With the development of social economy and the progress of technology, especially the sharp increase of the city population and fast growth of the citification, there have been more and more high buildings and large space. The high buildings have such features as fire spreading fast, difficulty of extinguishing a fire and rescuing human beings and properties, so it has been an increasing urgent demand that how to effectively protect the safety of high building and detect the fire as early as possible. As a new-style and effective measure for the detection of early fire, much attention has been focused on image fire detection technology. In this dissertation, by the aid of digital image processing techniques, several detecting algorithms and the image fire detection systems based on the nerve net have been designed and realized.The principle of fire occurring, three developing stages and characters of every stage have been discussed systematically in this dissertation. The difference is also studied between conventional fire detection systems and new-style fire detection techniques. The classification of image fire detection, detection devices, the composition of the fire detection system and basic principle of digital image processing have been introduced in detail as well.The algorithms of fire image enhancement, noise filtering, the maximum entropy threshold segmentation, region growth, Otsu threshold segmenting and their applications in the fire image processing have been studied in this dissertation. On the base of integration of several segmenting algorithms, a new method has been put forward that can track the fire flame in the fire image sequence and can obtain the fire target. In addition, the algorithm has been discussed and realized that distills the fire smoke envelope in fire smoke imagesequence by use of instantaneous motion analysis and cumulated motion analysis. For the fire image recognition, the fire flame image recognition and fire smoke image recognition have been discussed respectively. Stress has been put on the flame image recognition. The features of fire flame and other disturbing phenomena such as area variety, shape variety, edge variety, flame flicker frequency and the whole motion trend have been detailed studied in this dissertation. And the corresponding recognition algorithms have been given. So the early fire flame can be recognized and the disturbing phenomena can be distinguished. The algorithm how to recognize the fire smoke image on the base of the analysis of optical characters of fire smoke has been also discussed.The fire detecting scheme based on nerve net has been discussed in the lastpart of this dissertation. First, the basic concept of nerve net has been introduced.And then the detail structure of the BP nerve net and the detail design precept of input and output layer have been given. Finally the fire detection system based onBP nerve net has been realized. The experiment results show that the BPNN fire detection system has good feature of robustness.