A Research On Algorithm Of Clear-Sky With FY-3 Mersi Data

Satellite remote sensing data had a wide range of uses, for a long time, including the atmospheric observation, terrestrial observation, oceanographic observation, resource exploration, land use, environmental monitoring and protection, as well as disaster monitoring, etc.. The acquisition satellite remote sensing data are seriously affected by unclear-sky data, such as clouds, cloud shadows and abnormal data. Since the end of 2008, FY 3A carried Moderate Resolution Imaging Spectrometer(MERSI) has started earth observation. The sensor has as many as twenty spectral channels, which contain rich and accurate Surface vegetation information. The use can not synchronously acquire MERSI L1 B data and CLM, cause they are separated of each other. So the goal of this study is to develope a practical solution of clear-sky data acquisition. According that the user can quickly get the high quality clear-sky data, using it to do research and other things. According to characteristics of MERSI, clear-sky data recognition algorithm has been explored further and by that remote sensing data for research and application has been acquired.At first, background knowledge, purposes and meaning of clear-sky-data recognition are introduced. Research status at home and abroad has been analyzed in a deep and extensive way. Methods of clear-sky-data recognition have been summarized accordingly. MERSI platform and its characteristics of Fengyun has been introduced.Second, The introduction of the internal structure and pre-processing of MERSI has been taken. To make preparations for the research on recognition algorithm of Clear-Sky, geometric correction, reflectivity, brightness temperature have been taken or calculated.Third, In the research on recognition algorithm of “unclear-sky” of abnormal pixels, including three cases( abnormal stripes, “dash lines” and isolated pixels), threshold method and morphology method have been compared and verified. Morphology method could do recognition better cause it Suitable for more scenarios.Fourth, in the paper, the recognition algorithms of clear-sky has been researched deeply. The best spectral characteristics have been selected by J-M distance value. The main job is that the research on recognition algorithm of clear-sky of normal pixels. Different threshold values of method of Decision Tree, which use reflectivity, brightness temperature and the index portfolio to recognize, could lead to missing and false detection problem in the recognition procession. The fixed threshold method for pixel identification according to the reflectivity, multi channel brightness temperature and the index portfolio, the change of threshold value will result in missing and false detection problem. The dynamic threshold method based on Ostu can solve the problem of one-time detection and improve its effect.In this paper, the recognition system has been developed after the research on recognition algorithms of clear-sky. By this, users can get the product of clear-sky data quickly and efficiently.