The method of data fusion can make up for the results of single sensor data，which is incomplete and partial defects. A fire detection algorithm based on neural network and fuzzy theory is proposed and applied to the fire detection system in order to ensure the accuracy of the judgment . Through the use of neural network toolbox and fuzzy logic toolbox, MATLAB simulation and analysis are carried out on two kinds of algorithms. The conclusion is drawn that the combination of two algorithms is suitable for fire detection. Finally, using VS2010 to realize the function of fire detection algorithm based on the development platform of the detection system. The algorithm realizes the process of fire data, obtains the probability of fire occurrence, judges the possibility of fire and ensure the accuracy of the judgment.