Design and implementation of a scenic fire detection system based on wireless network control

Abstract

Abstract: Most of the current fire detection systems are based on Zig Bee, WLAN or operator's 4G connections, which have the short communication distance, low capacity and high cost, and cannot meet the current and future needs of the Internet of Things for low power consumption, long distance, high capacity and low cost. In addition, these fire detection systems do not provide a good visualisation of the data and lack subsequent processing and analysis of the data. To address the shortcomings of current fire detection systems, this paper designs a Lo Ra wireless network-based fire detection system for scenic areas. (1) The Lo Ra terminal node is designed and built, including hardware design and software design. (2) The back-end system for fire detection in scenic areas is designed and built based on the Things Board framework.(3) A CNN-BLSTM algorithm model based on the attention mechanism was designed and built for environmental temperature prediction. In order to further enhance the analysis and utilisation of the data, the LSTM algorithm model was firstly built based on the characteristics of the existing collected data sources, and two levels of optimisation were carried out on the basis of the algorithm model. Firstly, to address the shortcomings that the LSTM model lacks the ability to deal with individual anomalous data disturbances and can only be trained forward to mine data links, a CNN and a BLSTM are introduced. and BLSTM pairs were introduced to extract the trend and pre/post features of temperature, respectively. Secondly, an attention mechanism is introduced to reconstruct the model for the final optimization of multiple environmental features that have different weights on temperature. The final reconstruction was carried out by combining with FNN, RNN.LSTM and CNN-BLSTM, it is concluded that the attention-based CNN-BLSTM .The results were compared with FNN, RNN, LSTM and CNN-BLSTM time-series models. Keywords: Fire Detection Systems, Lo Ra, Data Visualisation, Time Series Prediction