Research On Detection Of Metallic Contraband Via Passive Millimeter Wave Sequence Image

Security is one of the important guarantees of country,especially for the safety inspection of areas with large passenger flow.Civil aviation,railway stations,subway stations,etc.all have announcements and security measures for prohibiting the carrying of contraband.However,the existing security-check mode not only cannot meet the securitycheck needs of the peak passenger flow,but also exist the large amount of security personnel input,the slow speed of personnel passing,etc.Checking mode for security purposes must be able to cover people,baggage and letters with a non-invasive manner to detect weapons,explosives biological or chemical contraband,etc.Therefore,it is urgent to propose a new security-check mode that can meet the current security-check of large passenger flows and can effectively detect contraband.The novel millimeter-wave based detection technology can cope with this problem well.The millimeter-wave has a certain ability to penetrate the clothes,and can image the concealed contraband under clothing.And millimeter-wave does not cause harmful radiation to the human body.Based on the 34 Ghz passive millimeter wave(PMMW)image acquired by the millimeter wave imaging system with the actual scientific research project,this paper studies the contraband object detection(concealed object)detection technology of sequence images.The main contents include:(1)Study the millimeter wave radiation characteristics and imaging principle,analyze and compare the radiation characteristics of human skin and typical materials;compare and analyze the existing passive millimeter wave image contraband identification and detection methods.These studies laid the foundation for the real-time detection of metal contraband in passive millimeter wave sequence images.(2)The traditional PMMW contraband detection method based on image denoising,image enhancement,image segmentation and edge detection is studied.(3)The target detection theory based on deep convolutional neural network,the basic unit of convolutional neural network,the structure design and optimization method of yolov3 neural network are studied.(4)Aiming at the low spatial resolution and low contrast of single-frame PMMW images collected by existing PMMW imaging systems,a method based on wavelet fusion for PMMW sequence image hiding detection is proposed.This method can effectively solve the above problems,and has certain processing power for sequence PMMW images,but the generalization ability of the algorithm is weak,and it can not meet the application requirements in the target recognition and detection real-time.(5)Aiming at the problems of low precision and low real-time detection speed of PMMW image target detection method based on wavelet fusion,a real-time detection method of PMMW image contraband based on yolov3 is proposed.The method combines the existing deep learning technology to realize the real-time detection of contraband in PMMW images efficiently and stably.