Research On Pre-processing Method Of Passive Millimetre-wave Image Based On Partial Differential Equations

Passive millimetre-wave (PMMW) systems own an extensive practicingforeground in fields of battlefield reconnaissance, aircraft landing, airportsecurity-check and mine sweeping due to the technical superiorities on highanti-reconnaissance capability, great ability of working at any time in any kind ofweather and penetrating common clothing, good performance on metal target detecting.However, MMW image suffered from severe noise pollution and accompanied withblurred edge, furthermore the channel responses of array in PMMW systems areinconsistent seriously, which make the research on methods of MMW imagepre-processing to be the premise and the key-part of the PMMW imaging technology.The methods of image processing based on partial differential equations (PDE)make a discrete image to be a continuous mathematical model, and use the completedtheories of numerical analysis on PDE to process the image. These methods have higherlocal adaptability and greater flexibility, compared with traditional methods.The pre-processing of PMMW image based on PDE is proposed in this paper bycombining the features of MMW image and those of image processing technologybased on PDE. The main contents of the paper can be listed as follow:1. Analyses are made on characteristics of PMMW imaging system and PMMWpicture, a powerful supplement is made based on existing evaluation factors toeffectively quantify the performance of the algorithms that used in this paper.2. To deal with the problem that MMW image suffers from severe noise pollution,a fourth-order PDE model based on Laplacian and gradient operator (LG model) isproposed here to remove noise. The advantages of PM and YK model are integrated inLG model to eliminate the blocky effect which exists in results of PM model, and tosolve the problem that weak ability of noise removal in YK model, as well as to make agood trade-off between noise removal and edge information keeping.3. To deal with the problems that MMW image accompanied with unclear edge andbad performance on rate of convergence, an improved image enhancement methodbased on adaptive bidirectional shock diffusion (IABSD method) is brought forth here. This method introduces four-orientation gradient and Laplacian operator to traditionalmethod which improves the convergence speed and enhances the real-time featurecompared with old ways.4. To deal with the problem that channel responses of array in PMMW systems areseriously inconsistent, a scene-based channel equalization method based on channelrelations and thermal radiation uniformity of a scene (SBCT method) is proposed. Thecold and hot calibration processes in source-based PMMW imaging can be leaved outby utilizing this method, which will cuts down the cost of system development andimproves the real-time feature of imaging systems. Contrasted with traditionalscene-based channel equalization methods, this method possesses stronger capacity ofarray non-uniformity correction.Experimental completed with simulated and measured data demonstrate theeffectiveness of LG model, IABSD and SBCT method, the results demonstrated thatcompared with the existing ways these three methods perform better on noise removal,image enhancement and channel equalization respectively.