A new multi-sensor fire detection method based on LSTM networks with environmental information fusion

Multi-sensor fire detection has been widely used, which allows monitoring multiple environmental indicators. However, most multi-sensor detection methods detect fires only by comparing the measurements of environmental indicators at each detection time with the preset thresholds. It is prone to fire false alarms due to neglecting the time series characteristics of environmental information. To improve the robustness and accuracy of fire detection, this paper proposes a new multi-sensor fire detection method based on long short-term memory (LSTM) networks, named EIF-LSTM. EIF-LSTM integrates environmental information fusion, which is divided into two steps. First, EIF-LSTM extracts the time series characteristics of the monitoring environment by processing multi-sensor time series readings, including environmental indicator variation information and environmental level information. Second, the normalized multi-sensor time series readings, environmental indicator variation information and environmental level information are fused together for fire prediction. The LSTM network realizes the extraction of environmental time series characteristics due to its ability to learn long-term dependencies. The addition of two kinds of time series information increases the detection dimension and enhances the fusion effect. Experimental results on a real-world fire dataset show that EIF-LSTM is capable of achieving state-of-the-art detection performance.