Fire Detection Method Based on Improved Glowworm Swarm Optimization-Based TWSVM

Abstract

In recent years, fires occurred frequently, and the solutions of fire detection and recognition are required automatically. Although there are various method of fire detection such as temperature and smoke sensor, the efficient real-time detection is not guaranteed. In order to solve this problem, a quick and effective fire detection method of twin support vector machine (TWSVM) based on improved glowworm swarm optimization algorithm (GSO) is proposed. Compared with SVM, twin support vector machine not only extends the range of the kernel function selection, but also obviously enhance the generalization ability of support vector machine. However, there are also some problems of twin support vector machine including the disability of dealing with the parameter selection problem properly which could reduce the classification capability of the algorithm. In this paper, we use improved glowworm swarm optimization algorithm to search the optimal penalty parameter and kernel parameter of TWSVM. Reasonable parameters will improve the performance of twin support vector machine and increase the accuracy. Experimental results show that the proposed approach is efficient and has high classification accuracy compared with traditional support vector machine and so on. Finally, the proposed method of twin support vector machine based on improved glowworm swarm optimization algorithm effectively improves the accuracy and real-time performance of flame identification, and settles the problems of TWSVM such as the difficulty of parameter selection in flame recognition and long optimization time of common parameter optimization algorithms.