Abstract: The authors propose a method to real-time render large amount of LIDAR points clouds. In order to fast culling invisible data and make result of progressive rendering reasonable, a sequential coded Quardtree is built with points clouds evenly distributed to nodes of every layer from top to down. The authors discussed the procedure of building a sequential coded QuardTree, explain how to add points clouds to nodes of tree. The authors also implement a system to prove the method proposed in this paper, which can real-time render 1GB raw point clouds( about 20 million points) data with common commodity computer.