As urbanization accelerates, data on the diverse aspects of urban life, including the environment, finance, and transportation, are increasing exponentially. Single-domain data analysis falls short for complex tasks, necessitating multi-source data integration to offer decision-makers a holistic view of interconnections and trends. However, challenges arise in effectively representing data associations and swiftly identifying relevant datasets. This paper proposes techniques to integrate multi-domain urban data relationships and facilitate the acquisition of target datasets. We introduce a method to transform data entities into association graphs and a data extraction method to aid users in understanding data relationships and acquiring target datasets. Using these methods, we developed DM-Vis, a data reconnaissance visual analysis system, to facilitate data integration and exploration from multiple sources. Real-data case studies and positive feedback from users and experts validate the utility and reliability of our approach.