Method of fire image identification based on optimization theory

In view of some distinctive characteristics of the early-stage flame image, a corresponding method. of characteristic extraction is presented. Also introduced is the application of the improved BP algorithm based on the optimization theory to identifying fire image characteristics. First the optimization of BP neural network adopting Levenberg-Marquardt algorithm with the property of quadratic convergence is discussed, and then a new system of fire image identification is devised. Plenty of experiments and field tests have proved that this system can detect the earlystage fire flame quickly and reliably.