Research On Target Detection Of Passive Millimeter Wave Imageing

Passive millimeter wave (PMMW) image can be obtained by detecting the radiationenergy difference of scene in millimeter-wave band. Passive millimeter wave imaging isemerging as military and civil imaging technology thanks to its penetrability, all-timeand all-weather operation without emitting microwave actively and so on. Passivemillimeter wave image target detection technology, extracting the target from thebackground using the radiation characteristics of objects and providing more validinformation about the target for the application of the system, is a very important issuein the modern information acquisition technique，and becomes one of the internationalresearch focus recent years.The essay, surrounding the PMMW imaging system application requirements of themilitary scene surveillance and security applications, mainly researches stationary targetdetection algorithms and the human concealed moving target detection algorithm. Themain contents of the thesis are as follows.1. Analysis the millimeter wave radiation characteristics of the object to guarantee thedelectability; analysis the detection range of the millimeter wave focal plane arrayimaging system and obtain the quantitative expression according to parameters ofthe system and imaging principle; introduce the reduced-scale test principle, whichsupplies the theoretical basis of the verification of the detection range.2. In view of the situation that PMMW moving target detection in different applicationbackgrounds are inseparable from characteristics of target and background,investigate the characteristics of them in different backgrounds, and establisheddetection framework respectively for stationary targets and human concealedmoving targets.3. Aimed at limited radiation brightness of target and not adequately used informationof PMMW image, propose a new method with Gaussian Weighted averaging toestimate the CFAR threshold, and propose the detection algorithm based on thistwo-dimensional cell weighted average CFAR (2D CWA-CFAR)for stationary targetof the PMMW images. In contrast to the detection algorithm based ontwo-dimensional cell average CFAR (2D CA-CFAR), the proposed algorithm canaccurately detect target.4. In view of the detection problem caused by movement and scale-Invariant of theconcealed targets of the millimeter sequence images, extract the body area, andestablish the Gaussian mixture model (GMM) based on the extracted body. Putforward the concealed target detection algorithm based on Gaussian mixture model and expectation maximization (EM) estimation. Analysis the performance of thealgorithm by comparing with multi-level EM detection method using the BayesianError Rate (BER) and Frame Processing Time (FPT).The detection result of the stationary targets in measured millimeter wave images,which obtained from the91.5G Hz band16-channel millimeter wave focal plane arrayimaging system shows the effectiveness of the algorithm; while the concealed movingtargets detection of the millimeter sequence images for human security demonstrates thedetection algorithm based on Gaussian mixture model and EM estimation can detect theconcealed target accurately and efficiently.