高光谱图像物理信息提取与目标检测识别方法研究-光学工程专业论文.docxVIP

Abstract Abstract III Abstract Hyperspectral images，which are of the property of摇nity of image and spectrum? provide spatial，radiative and spectral information of ground objects，these characteristics endow hyperspectral remote sensing with unique advantages in both data mining of the information and detectionrecognition of small differences of the objects，and make it a hot spot in the field of remote sensing information processing． This dissertation focuses on how to use the abundant information of hyperspectral images to retrieve the physical information and to detect and recognize the targets．The major works carried out are as follows： (1)A method of retrieving water-leaving radiance based on the radiative transfer theory was proposed．Firstly,the interference of the SUn glint on water surface was removed，the water area was obtained using texture segmentation，avoiding obscuring land pixels when the glint Was removed and facilitating the subsequent treatment；the major factors which．influenced the degree of the glint contamination were analyzed in detail，the interference of the sun glint was removed using own information of the scene．Based on the glint removal and the atmospheric radiative transfer theory,a retrieval model of water-leaving radiance Was developed and the water-leaving radiance was obtained． (2)A method of retrieving surface reflectance based on the radiative transfer theory was proposed．Firstly,by analyzing the effect of the dispersion phenomenon on the retrieval of surface reflectance and expressing the fresnel reflectance as the function of wavelength,the accuracy of the results was greatly improved through the whole bands；secondly,the effect of the sky radiance on the retrieval of surface reflectance was analyzed，the retrieval accuracy in visible bands was greatly improved thanks to the elimination of extra effect caused by sky radiance；Finally,．a retrieval model of reflectance was developed and the surface reflectance was obtained． (3)A method of