A comprehensive understanding of the spatial distribution of urban settlements is significant to a series of research topics related to environmental and biological changes caused by the urbanization process. Among various technologies, Synthetic Aperture Radar (SAR) has been successfully applied in urban settlement mapping in the past two decades. Although much effort with varying degrees of success has been made in previous studies, the research work is still ongoing, and three challenges should be highlighted. First, the effect of de-speckling is usually underestimated, to the extent that the improvement of the SAR image quality is totally ignored in some studies. Second, a method that combines full Interferometric SAR (InSAR) information is as yet missing. Third, training samples are generally required to process SAR images to extract urban settlements, which is time-consuming and labor-intensive, or even impractical when classifying satellite data at the regional/global scale. To address these issues, this paper presents an automatic method for urban settlement mapping trained by multi-temporal InSAR using social media. To improve the detection performance and reduce false alarm ratio, intensity and coherence are first accurately estimated without loss of image resolution by homogeneous pixel selection and robust estimators. The homogeneous pixels will be also applied to measure urban characteristics from the geometrical prospective. After that, training samples are automatically generated from social media based on the fact that cities and urban areas are nowadays full of individual geo-referenced data such as social network data Finally, these multiple information sources will be fused to extract urban areas based on an improved one class classifier. Experimental results show that the proposed method is effective in extracting urban areas with good accuracy. This study provides a new de-speckling means to process multi-temporal InSAR and sheds new light on the applications of social media in the field of remote sensing.