Fire Image Detection Technology

The traditional methods for fire detection usually include temperature detection, light detection and smoke detection. These technologies are limited by the factors such as environment and space so that mistakes often are appeared. The fire detection technology based on image processing is a new and effective detection method, which can effectively overcome the disadvantages of traditional methods. So, the fire detection technology based on image processing has broad application prospects.In this paper, the principle of image detection technology is introduced. The characteristic of the fire image is analyzed. On this basis, a new fire flame detection algorithm base on three stores multi-feature and RBF neural network is proposed.To solves the problems of the fire image are difficult to be segmented, and the background is complex, fire image need be enhanced by using some processing technologies such as noise filtering, histogram equalization, image sharpen. The three stores segmentation model is established by using threshold segmentation, RGB space model, morphology characters segmentation in order to exclude most of the disturbance and get the suspicious flame zone. And then, some characteristics such as perimeter, area, degree of roundness, center of excursion distance, circularity of the fire image are studied.Finally, the model of fire detector algorithms is established by using RBF neural network net in which flame characters are taken as inputs and by which is used to classify and recognize fire image. The experiment results show that the algorithm has higher reorganization rate in different environments.