Research On Fire Detection Algorithm Based On Image Processing

With the urbanization process and the rapid development of the economy and society,the urban population density and the hidden fire hazards have increased.As one of the frequent major disasters,fires have caused more and more serious economic losses and personnel safety problems to the society.Therefore,it will be of great theoretical,practical and economic value to improve the accuracy of fire detection and take effective early warning measures in time.Compared with the traditional fire detection technology,the fire detection method based on image processing has better performance in terms of recognition accuracy and real-time performance,so it has become one of the research hotspots nowadays.In the current stage of fire recognition research based on image processing,when using traditional intelligent optimization algorithms such as particle swarm optimization(PSO)to optimize support vector machine(SVM),the classification models obtained generally will have the problems of low fire sample recognition rate and low algorithm efficiency.In order to improve this situation,this paper applies two new intelligent optimization algorithms in recent years,artificial intelligence colony(ABC)and gray wolf algorithm(GWO),to the research of image fire recognition to obtain better model parameters and improve the model classification accuracy of fire samples.The main research contents of this article are as follows:(1)The main theories of image processing,related technologies and commonly used processing methods were analyzed in deep.In order to improve the image quality and the accuracy of recognition,the methods of graying and image filtering were determined through comparative analysis and simulation verification.(2)Based on the research and comparison of the characteristics of mainstream moving target detection methods such as optical flow method and background subtraction,the detection method of background difference combined with mixed Gaussian background modeling was selected to obtain the suspected target area;At the same time,the color moment and the gray level co-occurrence matrix were used to extract the color and texture characteristics of the target area respectively.By comparing the features with the main interferers,the validity and rationality of each feature extracted were further explained.(3)The study discussed the performance characteristics of particle swarm optimization algorithm,artificial bee colony algorithm and gray wolf algorithm to optimize support vector machine parameters,and used Libsvm toolbox based on Matlab platform for experimental verification.Finally,the experimental research of PSO-SVM,ABC-SVM and GWO-SVM on fire image recognition was carried out,and their recognition performance was verified and compared.The experimental results showed that,compared with the traditional PSO-SVM,the ABC-SVM and GWO-SVM used in this paper can more accurately identify fires and major sources of interference.Among them,compared to PSO-SVM,the fire sample recognition accuracy of GWO-SVM has been improved from 96% to 99%,and its classification speed has been increased by 32.4%.It has shown better performance in terms of recognition accuracy and operating efficiency,which confirmed the effectiveness of this study.