Detecting Human Activities in the Arctic Ocean by Constructing and Analyzing Super-resolution Images from Modis Data

In this paper, we propose a framework to detect human activities by constructing super-resolution images from the MODIS data. The highest resolution of the MODIS images is 250 meters per pixel, which is usually not enough to detect human activities. By magnifying and de-blurring the low resolution MODIS image through the SupportVector Regression, the constructed super-resolution image can achieve 4 to 8 times higher resolution than the original MODIS image. To evaluate the feasibility of the super-resolution MODIS images for the application of human activity detection, we collect a dataset by selecting four land cover types through Google Earth: the land with human activities, the land without human activities, the water without ice, and the land covered with snow and ice. Using a learning-based method, surface reflectance from the super-resolution MODIS image predicts land cover type of a geo-location specified by latitude and longitude. Experimental results demonstrate feasibility of the proposed approach for human activity detection using the super-resolution MODIS images.