Studies On The Optimal Image Matching Technology Of Meteorological Satellite In The Calculation Of Cloud Top Height

The study of cloud is very important to the meteorology and climate, it will be difficult because of the variability and mobility of cloud. Snow exiting at high latitudes in winter, makes difficult to distinguish cloud and snow in remote sensing image. So many errors exit in cloud cover between remote sensing observation and ground observation. A new research method is needed to distinguish cloud and snow effectively. This paper brings digital photogrammetry into distinguishing cloud and snow, and the key technology is image matching between conjugate point. Compared with the high-resolution image, the middle-resolution image has poorer resolution, poorer gray characteristics, and serious’ bowtie effect’, which makes image matching difficult. To solve these problems, this paper puts forward an image matching method based on dynamically eliminating the bowtie effect, integrated with correlation coefficient method, least squares method, relaxation algorithm, or image feature matching algorithm, which provides data support for the coming solution of cloud top height and identification between cloud and snow. Then, this paper builds an calculation model of cloud top height using the digital photogrammetry. In conclusion, a solution method to distinguish cloud and snow based on digital photogrammetry is put forward, which has important scientific value and practical significance to meteorological observation, synoptic analysis, and climate forecast.The major works of this paper includes the follows:(1) Study the characteristics of middle-resolution image like AQUA/MODIS and FY3B/MERSI, and the regularity of the overlapping problem sufficiently. Use the method integrating with correlation coefficient matching, statistics and curve-fitting to eliminate the bowtie effect well. To ensure the integrity of the constitutive parameters of the image, put forward a getting value method of dynamically eliminating the bowtie effect, which has been proved to be practical and reliable by experiment.(2) Mainly research the image matching method suitable to middle-resolution image. Based on dynamically eliminating the bowtie effect, put forward a matching method integrating correlation coefficient with relaxation algorithm, a bidirection least squares method, and a new SIFT image feature matching method using chessboard distance. Specific tests and analysis are done to prove the modified method stable and transplantable, the modified method is suitable for the ground and cloud field.(3) Bring the digital photogrammetry into the calculation of cloud top height, structures the imaging equation of middle-resolution image of meteorological satellite-AQUA/MODIS and FY3B/MERSI. Distinguish cloud and snow effectively using the calculating model based on digital photogrammetry. Specific tests, analysis, and comparison with MODIS cloud cover product, this paper’s result can coincide with MODIS cloud cover product well. Especially, our result has higher resolution and more fine effect when distinguish cloud and snow.