Gai, M., Cho, Y., aŶd QiŶghua, X. ;ϮϬϭϯͿ. Target-free automatic point clouds registration using ϮD iŵages. ProceediŶgs of the ϮϬϭϯ A“CE IŶterŶatioŶal Workshop oŶ CoŵputiŶg iŶ Civil Engineering, June 23-25, Univ. of Southern California, LA, California. pp.865-872. Target-free Automatic Point Clouds Registration Using 2D images Mengmeng Gai 1 ; Yong K. Cho A.M.ASCE 2 ; Qinghua Xu 3 1 GRA, Charles Durham School of Architectural Engineering and Construction, College of Engineering, University of Nebraska - Lincoln, 1110 S. 67 th St. Peter Kiewit Institute (PKI) 118A, Omaha, NE 68182, USA; PH +1-402-206-7617; FAX +1-402-554-3850; Email: mengmeng.gai@gmail.com; 2 Associate Professor, Charles Durham School of Architectural Engineering and Construction, College of Engineering, University of Nebraska-Lincoln, 1110 S. 67 th St. Peter Kiewit Institute (PKI) 104C, Omaha, NE 68182, USA; PH +1-402-554-3277; FAX +1- 402-554-3850; Email: ycho2@unl.edu; (Corresponding author) 3 GRA, Charles Durham School of Architectural Engineering and Construction, College of Engineering, University of Nebraska - Lincoln, 1110 S. 67 th St. Peter Kiewit Institute (PKI) 118A, Omaha, NE 68182, USA; PH +1-718-664-8432; FAX +1-402-554-3850; Email: qinghua.xu@huskers.unl.edu ABSTRACT This paper presents a target-free automatic point cloud registration method which is based on identified common features between multiple images in point clouds. Target- free automatic registration of point clouds is an important research area in which several problems are still not addressed completely, such as registration accuracy, time, and size of overlapping area between point clouds. In this study, a series of point cloud data sets from different scan positions are registered based on the texture data in the overlapping area between point clouds sets. Both texture and point cloud data are obtained by a developed Laser Detection and Ranging (LADAR) system simultaneously. The automatic registration algorithm begins by identifying 2D common features of the texture, then constructing corresponding 3D common points, and afterwards finding out the optimal transformation matrix. The introduced method will reduce substantial amount of time for point clouds registration by using a few images within a small overlapping area between point clouds without the need of target references, thus significantly promoting as-built modeling process for existing infrastructure or construction progresses. INTRODUCTION Rapidly updating as-built design with detailed 3D scenario is useful in construction applications such as process monitoring and safety hazards detection