Design And Implementation Of Smoke Detection And Recognition System Based On Dark Channel And Wavelet

With the improvement of the level of science and technology,most of the substations are unattended.In the substation monitoring center,the monitoring of the target of attention is completely based on manual staring,and it is inevitable that there will be omissions.In particular,fires that are not detected and saved early will cause significant casualties and property damage.Video smoke detection has the advantages of fast speed,low environmental impact,low cost,etc.,which provides a strong guarantee for early warning of fire.This is often accompanied by the generation of smoke in the early stages of the fire.Based on this,it is particularly important to design a system for early detection and identification of smoke in substation fires to assist in fire alarms.The smoke detection and recognition system based on dark channel and wavelet described in this thesis is divided into video preprocessing module,motion region detection module,motion region screening module,smoke feature extraction module and support vector machine(SVM)smoke recognition module.The smoke pre-processing module uses grayscale and bilateral filtering to reduce the amount of post-calculation while reducing the edge information of the noise-preserving image.The motion region detection module is based on mixed Gaussian background modeling,morphological filtering,median filtering,stability determination and other processes to extract the foreground.In the motion region,the motion region screening module uses dark channels and wavelet transform to eliminate high-transmittance interferences and boundary-specific interferences in the motion region.The smoke feature extraction module mainly extracts the LBP texture features of the filtered motion regions,and the support vector machine(SVM)The smoke recognition module performs prediction based on the feature information provided by the extraction module.Based on the needs of early warning of substation fires,it is designed to design a system that can quickly identify smoke areas and improve the accuracy of smoke identification as much as possible.