Spatiotemporal enabled Content-based Image Retrieval

Remote sensing has emerged as a powerful tool in a number of applications, but many challenges such as the development of “smart’ (knowledgeable), effective and efficient Earth observation (EO) image-content extraction and content-based image retrieval systems in response to big sensory data acquisitions by ever-increasing spaceborne EO imaging sensors remain unsolved. In this paper, we discuss the need to explicitly specify a priori 4D spatiotemporal scene domain knowledge to be mapped onto the image domain in terms of 2D image features and spatial constraints. This 4D-to-2D mapping capability implies the solution of the vision problems, where the semantic gap from sensory data to high-level information products must be filled in.