Improving the Accuracy of Automatic Reconstruction of 3D Complex Buildings Models from Airborne Lidar Point Clouds

Abstract

Due to high requirements of variety of 3D spatial data applications with respect to data amount and quality, automatized, efficient and reliable data acquisition and preprocessing methods are needed. The use of photogrammetry techniques—as well as the light detection and ranging (LiDAR) automatic scanners—are among attractive solutions. However, measurement data are in the form of unorganized point clouds, usually requiring transformation to higher order 3D models based on polygons or polyhedral surfaces, which is not a trivial process. The study presents a newly developed algorithm for correcting 3D point cloud data from airborne LiDAR surveys of regular 3D buildings. The proposed approach assumes the application of a sequence of operations resulting in 3D rasterization, i.e., creation and processing of a 3D regular grid representation of an object, prior to applying a regular Poisson surface reconstruction method. In order to verify the accuracy and quality of reconstructed objects for quantitative comparison with the obtained 3D models, high-quality ground truth models were used in the form of the meshes constructed from photogrammetric measurements and manually made using buildings architectural plans. The presented results show that applying the proposed algorithm positively influences the quality of the results and can be used in combination with existing surface reconstruction methods in order to generate more detailed 3D models from LiDAR scanning.

A Cost-Effective Method for Reconstructing City-Building 3D Models from Sparse Lidar Point Clouds

“…The proposed algorithm has been created for the specific task of reconstructing simple 3D building models from sparse point clouds. It is based on several ideas first introduced by the 3D Grid conversion method [34], which performs special data preprocessing of the input point cloud, allowing for the generation of more detailed 3D meshes from sparse point clouds when using solutions such as Poisson surface reconstruction [35]. The algorithm proposed in this paper incorporates the ideas of preprocessing the input data in a way that increases the spatial regularity of the point cloud as well as dividing the data into a set of layers describing different height ranges.…”

Three-dimensional scanning system based on a low-cost infrared sensor

“…The authors tested their approach in several real datasets and compared it with other state-of-the-art approaches. In addition, a new algorithm to correct 3D point cloud data from airborne LiDAR scans of 3D buildings was presented in [12]. To verify the quality of the reconstructed objects, the authors used high-quality ground truth models from constructed meshes.…”

HUMANNET—A Two-Tiered Deep Neural Network Architecture for Self-Occluding Humanoid Pose Reconstruction

“…While the majority of methods focus on voxel based mesh representation [22][23][24][25][26][27], for object reconstruction due to their representation simplicity, voxels have one major flaw-exponentially increasing requirements to train them with increasing fidelity. Some papers tried to solve this ever-increasing memory requirements using smarter data representation styles like octrees [28,29].…”