On the assessment of automatically processing HR/VHR imagery using low-resolution global reference data

Recently, we presented a general framework for processing high and very high resolution images (HR/VHR) with the prospect of automatically constructing a Global Human Settlement Layer (GHSL) [1]. This work constitutes an integral part of this effort and focuses on finding fairly reliable modes for the assessment of the produced image information. We describe the validation process together with the performance measures and we apply regression analysis in order to investigate the relationship, if any, among the results derived from human interpretation, the outcome of our methodology and established global reference data as MODIS 500-m urban.