We explore a methodology to construct 3D maps of indoor environments using multiple heterogeneous robots. The robots acquire overlapping 3D range images at contiguous locations and fuse them together to form a complete range map. Our methodology involves three major steps in map construction: the first step is to estimate an initial pose between the multiple robots using a real-time camera pose estimation algorithm, the second step involves extracting polygon features from overlapping range images and the third step involves registration of the overlapping range images using extracted polygonal features. We evaluate the performance of several approaches to camera pose estimation, which is used to determine the initial estimate of pose between the multiple robots. We introduce a fast and robust segmentation algorithm, which extracts the polygonal features from the 3D range images. A polygon-based registration algorithm accurately fuses the overlapping range images. The algorithm further refines the initial pose estimate by minimizing the geometric distance between corresponding polygonal features extracted from overlapping range images. Our polygon-based registration combined with camera pose estimation executes in real-time and much faster compared to point-based scan registration techniques. We present experimental results obtained while building indoor maps and quantitative analysis of the algorithms used in map construction of indoor environments.