A web-based semi-automated method for semantic annotation of high schools in remote sensing images

The overwhelming volume of routine image acquisition requires automated methods or systems for feature discovery instead of manual image interpretation. While most existing researches focus on extracting elementary features such as basic terrains and individual objects, the detection of compound feature is still a challenge. This paper proposes a semi-automated approach integrating supervised image classification and geo-processing workflow to discover and annotate compound objects within RS images. Taking the high school in U.S. as an example, we developed a web-based prototype system to detect compound objects. Experimental results by the prototype show that the approach is capable of annotating high schools with an acceptable accuracy. This paper demonstrates a novel way to leverage existing technologies in completing the semantic annotation of RS images.