基于STM32F和极限学习机在火灾检测中的应用

针对传统单一信号的火灾检测方式存在误判问题，以及布线复杂并且性价比低的弱点，提出了基于STM32F和极限学习机火灾检测方法；该方法首先通过STM32F模块采集多个传感器的值（烟雾传感器，甲烷传感器，可燃气体传感器，一氧化碳传感器），WLAN为载体进行数据发送，然后采用加权滤波对数据进行去噪处理，获得极限学习机的训练和测试样本库，模型训练结束后，以测试数据进行方法验证，并对验证结果进行评估。结果表明，该方法能够准确判断火灾类型，准确度达到90%以上。在火灾处理算法方面，极限学习机相对于BP神经网络、支持向量机和贝叶斯网络训练时间短，准确率高,具有较高的应用于推广价值。

Application of STM32F and ELM in fire detection

Traditional fire detection mechanisms which aims at using a single signal method results in misjudgements, complex wiring and low performance-to-price ratio. Aiming at solving these problems, a method of fire detection based on STM32F and extreme machine learning algorithms is proposed. For our model, the value of multiple sensors by STM32F module is collected, a WLAN is used as the carrier to transmit data and then denoised by weighted filter to obtain the training data for the ELM. After the model training, a simulation experiment on fire detection is finally carried out on a test data to evaluate and verify the resulting performance. The result shows that, our method can accurately identify fire types with 90% accuracy. In fire?signal?processing?algorithms, the proposed model is faster and achieves higher accuracy when compared with several state-of-the-art methods such as BP neural network, Naive Bayesian and SVM, and it is practical and worthy of using abroad.