Efficiently reconstructing dense, complete, and accurate 3D models of long and complex pipelines remains a challenging task. The complex pipelines often contain geometrically unknown curved segments that are difficult to reconstruct by the existing monocular or stereo camera-based approaches. This paper proposes a dual-function depth camera array (DF-DCA) based inline inspection (ILI) system for automated 3D reconstruction of underground pipelines that contain arbitrarily curved segments. This newly developed DCA system consists of a central-positioned depth camera for pipes routes estimation based on a reference-assisted vSLAM framework and four oblique-viewed depth cameras for pipe surface mapping. A subvolume-based approach is used to merge the RGBD sequences from the cameras to obtain an accurate 3D map of the pipeline without using DCA calibration. The results showed that the system can reconstruct high-quality 3D models of the pipes with <0.2% distance error for straight pipes, and with centimeter-level geometric accuracy for pipelines with both straight and curved sections.