An Ensemble Algorithm Combining Multi-models and Proposed Chaotic Harris Hawks Optimization for Fire Flame Recognition

Abstract

Fire recognition and early prevention are of great significance to reduce the loss caused by fire. In this paper, an ensemble algorithm combining multi-models and proposed chaotic Harris hawks optimization (CHHO) is proposed for fire flame recognition. The combined multi-models include decision tree (DT), K-nearest neighbor (KNN), least squares support vector machine (LSSVM) and extreme learning machine (ELM). Aiming at the problem that improper parameter will seriously affect the classification performance of the combined models, a chaotic Harris hawks optimization (CHHO) is proposed to optimize the parameters of models. Tent mapping, improved exploration mode and improved exploitation mode are introduced into CHHO to improve the performance of Harris hawks optimization (HHO). Finally, the output of each optimized model are obtained, then the final output are obtained by weighted average method. Experiment on a set of flame images shows that the proposed model is effective and has good classification performance.