Temporal-spectral Data Fusion Based On Multi-dimensional Data Datasets(MDD)

Time series analysis has played an important role in the study of global changes,vegetation phenology,and land surface cover changes.However,due to the limitation of sensor hardware development technology,although a single index of remote sensing data(such as spatial resolution,spectral resolution,and time resolution)is getting higher and higher,it still cannot meet the demand of various remote sensing applications for comprehensive indicators of remote sensing data.Therefore,spatio-temporal integration of remote sensing data has become one of the research hotspots in recent years.In practical applications,analysis of long-term sequences is often based on the analysis of spectral index characteristic parameters.For example,by studying the time variation of NDVI and EVI indices,the analysis of vegetation changes can be realized.In the study of water changes,the water body The index is an important parameter.In traditional methods,when we need to construct long-term sequence cube data of a certain spectral index,we often need to reconstruct images of all missing time one by one to obtain long-time series of remote sensing images and then extract the spectral index of each single image.Though,the synthesis of long-time series spectral index cube is very tedious.Therefore,it is very important to directly obtain a spectral time index cube with high temporal and spatial resolution,which can effectively reduce the data operation time and improve the spectral index reconstruction accuracy.In order to compare the effectiveness of the proposed method,the spectral indices of remote sensing images were reconstructed for a long period of time based on two methods: STARFM space-time fusion method based on traditional remote sensing data format and SREM space-time spectrum fusion method based on MDD data set.Quantitative analysis of the efficiency and fusion accuracy of the two time-spectral fusion algorithms shows that the method can greatly improve the computational efficiency and save the data processing time,which provides a new means for the research of remote sensing time series.The construction and quality control of MDD dataset are the premise and basis of this study.In order to solve the requirements of remote sensing image quality control in the process of building MDD dataset,a hyperspectral image cloud removal method based on SREM fusion model is proposed.The method can remove the influence of the cloud and at the same time maximize the original spectral information of the cloudless region of the hyperspectral image,providing a high quality cloudless data set for quantitative application of remote sensing.The main results and innovations of the paper are as follows:1.Constructing remote sensing data quality control requirements for MDD data sets,proposes a remote sensing image cloud removal algorithm based on SREM spectral reconstruction technology.Experimental results show that this algorithm can remove cloud influences and better rebuild cloud coverage area.Spectral information.2.Developed a time-spectrum fusion algorithm based on MDD dataset.Compared with the traditional time-spectrum construction algorithm,the time-spectrum curve fitting accuracy and computational efficiency were improved.This study provides a new approach for the study of remote sensing time series,which can greatly improve the efficiency of data processing analysis and the accuracy of time series analysis in the study of remote sensing time series.