This paper describes how model-specific constraints and domain specific knowledge can be utilized to develop efficient sampling based algorithms for robust model estimation in the presence of outliers. As a special case, a robust algorithm for homography estimation is proposed that exploits the invariance of collinearity under homography to improve efficiency in noisy scenarios. Unlike most existing approaches, the proposed algorithm does not make any assumption regarding the data distribution, data specific properties or availability of large amount of data. The proposed estimation algorithm is applied for multiple applications involving large sensor networks. These include estimation and maintenance of geo-registration by fusing observations from different modalities, such as RADAR and Automatic Identification System (AIS), and data-driven estimation (using target observations) of the relative topology of cameras with overlapping fields of view. Qualitative and quantitative results are presented that show the ability of the proposed algorithm to handle large fraction of outliers in the data, spatial noise, and high traffic densities, which are defining characteristics of these applications.