The High-frequency Disturbance Detection And Identification Of MODIS Image

The remote sensing image provides a great convenience for human to obtain land cover variation of the earth's surface. But the data is effected by the cloud, aerosol, snow and other factors, the real land surface information is often hidden in all kinds of noise. If these disturbance factors are not eliminated, it will be very difficult to accurately depict the regulation of the land surface characteristic products by means of remote sensing images.The effects of natural disasters in remote sensing image also belong to a kind of high-frequency disturbances. The detection of high-frequency disturbances can detect the influence of natural disasters. This work is conducive to the assessment of natural disaster damage, facilitate the nation to timely understand disasters. In addition, systematically study the spatial and temporal variation of clouds, aerosols, snow and other high-frequency disturbances is an important component of quantitative remote sensing.MODIS has many bands and wide applications, it can be used to monitor land cover change, carry out cloud detection and identification, calculate aerosol optical depth, extract marine physical and chemical indicators, study the changes of snow cover and so on. The high temporal resolution of MODIS images is easy for the detection and identification of high-frequency disturbance information. This work creates an advantage for the high-frequency disturbances of land cover study, disaster monitoring and assessment.The preprocessing for original image data of MODIS HDF format includes destriping, geometric correction and atmospheric correction. Based on the MODIS image of same time every month, achieve the land clear sky synthesis to use the ratio of one/seven band method, the sea clear sky synthesis to use the maximum chrominance method, set the land and sea sky image mosaic together to obtain the clear sky image on North China region. Based on pixel, take the daily image with the standard of clear sky image to build correlation analysis, based on the correlation coefficient to determine whether there is a disturbance, if the correlation coefficient is greater than 0.8 then determined to be non high-frequency interference pixel, or high-frequency interference pixel. Based on the spectral characteristics of the disturbed pixels, to determine the disturbance type. At present, carry on the the work of identifying cloud, snow, red tide and other types of high-frequency disturbances, detailed discuss the identification method of these three high-frequency disturbances, propose a set of detection and identification method of high-frequency disturbances for MODIS images in north China region.Cloud detection algorithm is improved based on multi-spectral threshold method. Use 7 bands such as band 1, 6, 8, 26, 31, 32 and so on. It can effectively detect thick clouds, cirrus, thin cloud and other cloud type by setting a particular threshold. Snow detection make use of NDSI, the apparent reflectance of band 2 and band 4. It can effectively detect the land surface snow coverage by setting the threshold. Red tide detection method using sea surface temperature (SST) to calculate the distribution of sea surface temperature. If abnormal high temperatures in some marine areas last a few days, significantly higher than the near surrounding water, the area can be considered to be red tide. The red tide waters can be selected by using that the reflectance ratio of band 3 and band 4 is greater than 1.5. The final distribution of red tide is obtained by intersecting the test results of these two conditions.