Research On Recognition Of Fire Image Based On Video Surveillance System

The theories, steps and idiographic implement process of fire image recognition by utilizing the remote video monitoring system and image processing technology are discussed in this thesis. Firstly, the basic theory of colorized image processing is introduced, the perceptual color clustering for image segmentation is realized in the condition of HIS color former which places great emphases on fire image; secondly, the image character is analyzed by using many image processing technology, such as gray processing, two-value processing, filter and so on, the harp-angled number as one of the criteria has been proposed. According to above research, fire images are classified and recognized by BP neural network. A great deal of different conditional experiments illustrate that BP neural network has many advantages, such as simple computing, wide applicability and good processing results. At the same time, common interferences can be resisted and the combination of robustness and acuteness is achieved.