Study Of Temporal And Spatial Change Of Vietnam Rice Cultivation Based On Multi-source Satellite Remote Sensing Data

Food is a necessity for the people to survive. The characteristic of basis and specificitymakes food security have a greater significance. Food security relate to humanity survival needscontention and a significant issue faced by many countries around the world.Most of Southeast Asian countries are traditional agricultural countries. Agriculture is thetraditional industrial in those countries. Except Singapore and Brunei, other Southeast Asiancountries’ more than60%population is lived based on agricultural production. In addition, theappropriate climatic and convenience irrigation make Southeast Asian region the most importantrice producing region in the world. Located in the same developing region, China and SoutheastAsian countries’ resources are quite complementary. We should understand rice cropping area,monitoring rice cropping area change and realize food production potential change. This issignificant to food security to China.Currently, remote sensing technology use for agricultural monitoring has become one of themost common methods for crop acreage census. MODIS data is characterized by high spectrumresolution and high temporal resolution, and it is very easy to get the data. MODIS data had beenwidely used for vegetation phenology detection and changing information extraction. However,effected by the cloudy and rainy weather in the study area and the multiple cropping system ofrice, in addition, the variance of rice planting time, limited the traditional method of remotesensing classification.In this paper, we studied paddy rice distribution extraction using remote sensing technology,and fragmented paddy rice area extraction using the combination of MODIS time-series data andTM image information. We studied the change of rice cropping pattern in Vietnamfrom2000-2010.Firstly, facing the cloud contamination in remote sensing images of Vietnam, this paperproposed a cloud remove method using of the filtering and replacing to de-noising the MODISNDVI timing series. Exploring the specific parameters set in the SG filter can effectivelyreconstruct invalid pixels NDVI maximum gap length.Based on the reconstructed MODIS NDVI time-series data, we proposed a similaritymeasurement method to extract paddy rice area that based on DTW (dynamic time warping) algorithm, and combined this similarity measurement with a fuzzy classification method. Theclimate and moist in study area are optimum for rice cropping all the time in the rainy season(May.-Oct.) which bring limitation to traditional remote sensing classification method. UsingSequence similarity to classify vegetation cover can not only take full advantage of thecharacteristics high temporal resolution of MODIS data, but also relatively accurately monitorthe rice acreage. Applying fuzzy classification in classify vegetation cover using MODIS datacould reduce the possibility of over-classification and misclassification by using a singlethreshold. The DTW distance which is calculated from DTW algorithm is treated as amembership in fuzzy classification method. The NDVI time series that extracted fromMOD09A1data in Vietnam of the year2000and2010were used to extract rice acreage andobtained the distribution of single cropping and multi-season rice area. It was proved that,83%of paddy rice area is correctly extracted.Facing paddy rice extraction in places where the rice cropping are scattered, in the supportof eCognition software, we based on the object-oriented approach and fused the similarity data inMODIS NDVI with the medium resolution TM remote sensing image and DEM image. Weprecisely classified medium north region at Vietnam in2000and2010. We use the Google earthand wild survey point as validation data to validate the classification result, and throughvalidation, it was proved that the correctly classification accuracy is89%, this method improvedMODIS NDVI time-series classification accuracy.The research in this paper shows that the combination of filtering and replacement methodwhich use of combined filter and replacement method can better reconstruct MOD09A1NDVItime series of NDVI product. The reconstructed MDODIS NDVI time series can provide moreprecise data can be better used for large area monitoring of agriculture.In this paper, the use of sequential similarity in combination with fuzzy classificationmethod can take full advantage of the characteristics of high temporal resolution MODIS data,through analyzing the change of crop growth calendar to improve the rice planting extractionaccuracy. In addition, through combination of the low spatial resolution MODIS time-series datawith TM remote sensing image data, we can significantly raise the classification accuracy inscattered rice planting area.This research proved a efficiently technology of cropping area monitoring in cloudy andrainy region and places of complicated human activity region.