Research On Image Type Fire Identification Method Based On Convolutional Neural Network

Fire is one of the great inventions of mankind toward civilization.However,a forest fir-e will bring both material and spiritual losses to human beings.Therefore,research on fire is a very important task.With the development of computer vision technology,the current research methods for fire have evolved from traditional sensor to image processing.In recent years,the fire identification theory based on image processing technology has been continuously improved and developed,but there are still some problems such as incomplete segmentation,blind feature selection and high false detection rate.In response to these problems,this paper studies the image-based fire identification based on the adaptive pooling convolutional neural network.In this paper,the fire image is preprocessed and its color features are analyzed.In order to completely segment the flame region,a method of fire region segmentation based on color features and a method of fire region segmentation based on improved KNN are proposed.Aiming at blind feature selection existing in the fire image recognition process,a method of feeding the segmented flame region into the convolutional neural network for training is proposed,which can better learn the flame characteristics.Based on the research of convolutional neural networks,a convolutional neural network based on adaptive pooling is proposed based on the traditional pooling methods.The network can dynamically use different pooled domain to achieve accurate extraction of features.Finally,this paper uses PyQT tool to design a graphical user interface,which can visually display every step of fire image processing.Research and experiments show that the proposed algorithm can accurately identify the fire image,effectively reduce the false detection rate of fire image recognition,and has certain value for image-based fire identification,which has certain significance for further improving the fire monitoring system.