An Integrated Method Of Quasi-Real Time Processing And Anomaly Extraction Algorithm For Satellite Infrared Data And Thier Applications

Since a long time, the studies on seismic infrared anomalies, especially those based on satellitic remote sensing, have been concerned in the field of seismology. In the mean time, the development of infrared monitoring means and related systems has been thought by the earthquake administrations to be of great importance. However, limited factors, such as time, regions and many others led to great differences in phenomena in earthquake activities and more over, the data relationship is so stochastic that it is almost impossible to avoid biases in researches on isolated events. Because of the great randomness, it is nothing but statistical analyses that can discover the essential rule behind the phenomena. Hence a general research on the plenty, sequential historic data using temporal and spatial compositive analysis of statistics is of great necessity to find out the universalism. Nevertheless, the lack of massive professional data warehouse needed by seismologists has seriously restricted the research. Consequently, only when a seismology oriented extensive satellite data processing system is developed can the research progresses be made.For the purpose of solving the problem in technique, we established a project to meet the needs of seismic infrared anomaly researches, including the development of automated satellite data processing system and the construction of infrared data warehouse, followed by a statistical temporal-spatial compositive analysis. Upon the project is the research on anomaly extracting methods aiming at monitoring means developing.The thesis introduces the solutions in automated noise and cloud elimination, data calibration and correction, encoding / decoding and geo-info dealing by studies of data processing algorithms. Based on the researches, a quasi-real time satellite data auto-processing system has been developed and through the operation of the system, a massive cloud-eliminated database and a suited imagery info database have been set up, containing data from April 2003 up to the present. The data were utilized for statistical analysis, which demonstrated that the brightness temperatures in the original data are largely discrete, whereas those in the cloud eliminated ten days'data are quite regular. Concretely, the variation of brightness temperatures shows temporal linear-correlation in short term and low-order correlation in long term.Based on the conclusion of above-mentioned statistical analysis, a set of models, respectively called linear trend deviation model and 2-order trend deviation model, were established to extract anomaly information from the processed satellite data. The models were set on the assumption that in normal cases the temperatures in ten days'data well meet the linear regressive equation or the 2-order one, along with a normaly-distributed random error. The differential value between a 50-day's trend predicted temperature and a sample temperature is regarded as a token of...