Research On Active And Passive Hybrid Millimeter Wave Imaging Target Recognition Method Based On Multi Feature Fusion

In the global public transport places,such as railway stations,subway stations and airports,some violent or terrorist incidents often occur.Therefore,it is necessary to exclude passengers carrying knives,guns,drugs,explosives and other dangerous goods.Millimeter wave has strong penetration ability to clothing,high resolution imaging image,and can detect dangerous goods hidden under clothing.Therefore,it is of great significance to research and make security inspection equipment and target classification and recognition technology based on it.Active millimeter wave imaging results and passive imaging results have their own advantag es and disadvantages.Therefore,this paper proposes an active passive hybrid millimeter wave imaging target recognition method based on multi feature fusion,which makes the measured results of the two imaging methods fuse at the feature level,and improves the recognition ability of millimeter wave imaging system for different dangerous goods targets.This paper studies the experimental system and results of the active and passive millimeter wave imaging,image preprocessing technology,feature extraction,multi feature fusion and target classification and recognition methods.The following four aspects are included:Firstly,on the one hand,the passive millimeter wave imaging technology is discussed,the blackbody radiation theory and the linear relation ship between the received power and temperature of the receiving antenna are derived,the structure of the passive millimeter wave imaging system and the optical photos of the measured system are introduced,and the experimental results of passive millimet er wave imaging are given.On the other hand,the active millimeter wave imaging technology is discussed,and the structure diagram of the active millimeter wave imaging system is given.Two active millimeter wave holographic imaging reconstruction algorit hms are derived in detail,including back propagation(BP)algorithm and reconstruction algorithm based on fast Fourier transform.The optical photos and measured data of the active millimeter wave imaging system are given.Secondly,the image preprocessing techniques for active and passive millimeter wave imaging are discussed,including image enhancement,morphological transformation,binarization and edge detection.This paper introduces the functions,methods and principles of image enhancement,morphological transformation and binarization,as well as the algorithms of Canny,Sobel,Prewitt,Roberts and Laplace operator.This image preprocessing process can effectively highlight the target,eliminate the background interference and clearly depict the ta rget contour.Thirdly,the methods of image feature extraction,data feature extraction and multi feature fusion are discussed.According to the image feature selection criteria,Hu second moment,eccentricity,rectangularity,image gradient and other imag e features are selected to complete the image feature extraction of passive millimeter wave imaging of different targets.According to the data feature selection criteria,the mean value and maximum value of regional amplitude are selected as the active da ta features,and the scattering data feature extraction of different targets in active millimeter wave imaging is completed.On this basis,a multi feature fusion method of weighted series fusion is proposed.According to the correct recognition rate and feature dimension of active and passive millimeter wave imaging,the weighted coefficient of feature vector is determined,and then the feature vector form of active and passive millimeter wave imaging fusion is obtained.Finally,the target classification and recognition methods for millimeter wave imaging are discussed,including support vector machine and BP neural network.For SVM,15 classifiers are constructed by one-to-one algorithm,and the feature vectors of six different targets are input into SVM;Aiming at BP neural network,a two-layer neural network is constructed,and the feature vectors of six different targets are input into BP neural network,and the conclusion that the fused feature vector is better than the original feature vector is obtained.In addition,the correct recognition number,correct recognition rate,training time and recognition time are used for quantitative analysis of the two classification and recognition methods.It is concluded that the effect of classification and recognition of millimeter wave imaging target using support vector machine is better than that of BP neural network,which provides ideas for the fusion of main and passive millimeter wave imaging at the feature level.In summary,the research on active and passive millimeter wave imaging experimental system,image preprocessing technology,feature extraction,multi feature fusion and target classification and recognition methods in this paper lays a theoretical foundation for the application of security equipme nt based on this.