【摘要】 目前火灾检测只是单一判断火灾是否发生,无法为火灾扑救提供更多参考。提出一种基于燃烧状态信息的火灾燃烧物种类识别方法。该方法采用STM32F搭建数据采集平台,离线阶段采集燃烧物在不同时刻的燃烧状态,建立样本库。在线阶段利用具有良好非线性映射能力和建模速度快的极限学习算法,对采集到的燃烧物状态数据进行识别。实验结果表明该方法能有效地判断出火灾燃烧物的种类,准确度达到90%以上。相对于BP神经网络和贝叶斯网络方法,该方法具有训练时间短、准确率高的优点。更多还原

【Abstract】 At present,fire detection is only a single judgement whether the fire has occurred.It can not provide more reference information for fire fighting.A method based on combustion state information is proposed to identify the types of fire burning.A platform is set up with STM32F to collect data.During the offline stage of the method,combustion state data in different time are gathered and stored as a sample database.During the online stage,the ELM with good nonlinear learning ability and the fast modeling,is used to intelligently recognize the collected combustion state data..Experimental Result shows that the ELM can accurately identify the types of burning material with an accuracy of more than 90%.Compared with BP neural network and Naive Bayesian,the method proposed has the advantages of short training time and high accuracy.更多还原