Adaptive Spatiotemporal Fusion Model Of Remote Sensing Data Based On Object Features

With the development of remote sensing satellite technology,remote sensing images are widely used to study surface dynamics.However,there are lack of the high temporal and high spatial resolution satellite images due to frequent cloud pollution,cost control,revisiting cycles and frequent cloud pollution,limiting load on satellite platforms and technical bottlenecks in the development of sensors.Spatiotemporal data fusion is the most flexible,convenient and effective way to solve such problems.Based on the ESTARFM algorithm,this study introduces phenological information and surface spatial structure information into the spatiotemporal fusion algorithm,then a spatiotemporal fusion algorithm based on phenological information and a spatiotemporal fusion algorithms based on surface spatial information were obtained.The results of the two modified algorithm are compared with the result of ESTARFM.It's found that the accuracy of the two modified algorithms were higher than ESTARFM.This study not only provides new ideas for improving the spatiotemporal fusion algorithms,but also provides technique support for constructing time series data.The main research contents and conclusions are as follows:(1)Based on the ESTARFM algorithm,the phenological information is introduced to modify the spatiotemporal fusion algorithm.A spatiotemporal fusion algorithm based on phenological information is proposed to eliminating the assumption that the change rate of each endmember is linear for paddy rice.According to the different phenology periods of paddy rice,a new rule for selecting similar pixels was established.Then the proposed algorithm was test through an experiment using real remote sensing data in the study area.Compared with the well-known fusion method ESTARFM,our algorithm showed better performance based on visual inspection and quantitative metrics.According to the robustness verification,the spatiotemporal fusion algorithm based on phenological information has good robustness.(2)Based on ESTARFM algorithm,the spatial space structure information is introduced to modify the spatiotemporal fusion algorithm.A spatiotemporal fusion algorithm based on surface spatial information is proposed to solve the contradiction between surface heterogeneity and moving window size.We take the local variance to indicate the spatial heterogeneity of the surface and calculate the local variance in the moving window of different sizes respectively.Then the moving window of the size which is corresponding to the minimum value of the local variance was selected as the moving window of the central pixel in the spatiotemporal fusion algorithm.The results of spatiotemporal fusion algorithm based on surface spatial structure information are compared with ESTARFM and real image.It is found that the overall accuracy of spatiotemporal fusion algorithm based on surface spatial structure information is higher than ESTARFM,and according to the robustness verification,the spatiotemporal fusion algorithm based on surface spatial structure information has good robustness.(3)The study found that the three common accuracy evaluation indicators,which were correlation coefficient,regression coefficient and root mean square error,were not applicable to all cases.Therefore,this study used the mean reflectance difference of six band of image to evaluate the accuracy of the fusion result.he mean reflectance difference of six band of image can clearly show the fusion accuracy of different regions on the image.