Multi-sensor Data Fusion Algorithm for Indoor Fire Detection Based on Ensemble Learning

Among all the disasters, fire is one of the most catastrophic events that frequently and universally threaten public safety and social development. In recent years, indoor fires have resulted in an increasing number of casualties and extensive damages. Therefore, in order to mitigate the impact of indoor fire disasters, it is crucial to propose an indoor fire detection method that can quickly and accurately detect the fire. In the research, a multi-sensor data fusion algorithm based on LogitBoost ensemble learning was designed for indoor fire detection. To improve detection accuracy and robustness, the proposed model utilizes an S-G filter and Min-Max normalization method, then uses Logitboost to synergistically integrate four classifiers, including Naïve Bayes, backpropagation neural network (BPNN), support vector machine (SVM) and k-nearest neighbor (KNN) classifier. A dataset containing various fire scenarios from the National Institute of Standards and Technology (NIST) was adopted to evaluate the effectiveness of the proposed algorithm. The experimental results demonstrated that the proposed method outperforms single models in terms of accuracy, stability, and efficiency.