Presentation,
Autonomous Time Series Generation of High Spatial Resolution Images - A Feasibility Study
(2017)
Abstract
Ävailability of daily high spatial resolution remote sensing data will soon become reality with the development of
optical sensor systems and the addition of new remote sensing satellites in a well-planned constellation. This
development will boost the growth of more remote sensing application in the fast growing remote sensing field. The
amount of data available for the remote sensing applications will definitely increase. Even with the fully developed functional remote sensing constellation, the remote sensing satellite architecture has some limitations in providing the required daily high spatial resolution images. Cloud cover over a particular area is one such limitation which blocks the remote sensing satellites from observing the ground surface. Also the forecasted remote sensing satellite constellation will only provide daily high spatial resolution images over the northern and southern hemispheres. Daily observation over the equatorial zone is still far from achieving. These major factors result in temporal gaps in high spatial resolution data and these gaps needs to be filled by other means. Because the performance of remote sensing applications such as crop monitoring and rapid changes in the ecosystem greatly depends on the availability of daily high spatial resolution data. Time series generation of high spatial resolution data by data fusion algorithms seems to be a better alternative for this issue. Using various data fusion algorithms, time series generation produces synthetic high spatial resolution images by fusing the high temporal moderate spatial resolution images with the low temporal high spatial resolution images. Time series generation is a complicated process which requires lot of human interventions. This is because, time series generation as a whole involves multiple processes. The main processes are downloading of respective satellite data, pre-processing of the downloaded data, removal of cloud pixels from the images and filling the gaps caused during this process, time series generation by data fusion and finally assessing the quality of generated data. Having a single tool which can perform the time series generation autonomously will be a great benefit for the remote sensing community. So far there is no such tool available as it is quite complicated to perform all the processes autonomously on the images from different sources used for the time series generation.
This paper provides detailed description about the above mentioned processes, investigates the challenges of
automating the processes and finally provides a possible solution to develop a tool which can perform the time series
generation autonomously with less to no human interaction."
