Research On Forest Fire Image Recognition Method Based On PCNN

Forest fires are devastating and seriously threaten the safety of forests.Traditional forest fire detection methods mainly use various sensors to detect the temperature,spectrum,smoke particles,combustible gas content and other indicators in the area.However,these sensing devices have large information loss,poor stability and real-time information cannot be stored.And other defects.The forest fire detection technology based on image processing technology can effectively make up for the shortcomings of traditional detection methods and realize real-time monitoring of forest fires.Aiming at the problem of forest fire image recognition,this paper uses pulse coupled neural network(PCNN)and bilateral filtering theory on the basis of in-depth study of the characteristics of forest fire images to process mixed noise in forest fire images,and uses PCNN-based background differential The motion detection deformation algorithm extracts the suspected fire area.In order to solve the problem that PCNN parameters are complex and difficult to select,this paper uses traditional genetic algorithms to automatically obtain parameters,and uses optimal family genetic algorithms for optimization.The simulation results show that the optimized algorithm can realize the automatic parameter setting,and can better remove the noise and extract the suspected fire area more accurately.According to the characteristics of the fire image,static feature variables and dynamic feature variables are extracted.Texture feature as an important feature of forest fire smoke image is one of the key issues studied in this paper.Using non-subsampled wavelet transform and gray-level co-occurrence matrix theory to extract texture features from each image block of the segmented suspected flame area image,simulation experiment The results show that the extraction method is more effective than the LBP method and the GLCM method.Finally,the rough set theory and the support vector machine theory are combined to establish the RS-SVM classification model.The attribute reduction algorithm is used to reduce the dimensions of the six feature variables input by the classifier,to remove redundant information,and to shorten the recognition time and improve the recognition efficiency under the premise of ensuring the recognition rate.Simulation experiments are carried out for the forest fire image recognition algorithm based on PCNN designed in this paper.The experimental results show that the recognition accuracy can reach 95.00%,the recognition speed is 43% higher than before the classifier reduction,and the missed detection rate is less than 3.4%.The missed detection rate is below 0.8%.Therefore,the image recognition algorithm designed in this paper meets the design requirements.