Infrastructure Engineering - Research Publications
Permanent URI for this collection
Search Results
1 - 4 of 4
-
ItemVIRTUAL ELEMENT RETRIEVAL IN MIXED REALITY(COPERNICUS GESELLSCHAFT MBH, 2022)Abstract. The application of mixed reality visualisation in construction engineering requires accurate placement and retrieval of virtual models within the real world, which depends on the localisation accuracy. However, it is hard to understand what this means practically from localisation accuracy alone. For example, when we superimpose a Building Information Model (BIM) over the real building, it is unclear how well does a BIM element fit the real one and how small a BIM element are we able to retrieve. In this paper, we evaluate virtual element retrieval by designing an experiment where we attempt to retrieve a set of cubes of different sizes placed in both the real and the virtual world. Furthermore, inspired by existing camera localisation methods for indoor MR being almost exclusively image-based, we use a localisation approach based solely on 3D-3D model registration. The approach is based on the automated registration of a low-density mesh model of the surroundings created by the MR device to the existing point cloud of an indoor environment. We develop a prototype and perform experiments on real-world data which show high localisation accuracy, with average translation and rotation errors of 1.4 cm and 0.24°, respectively. Finally, we show that the success rate of virtual element retrieval is closely related to the localisation accuracy.
-
ItemA PLATFORM for MULTILAYERED DOCUMENTATION of CULTURAL HERITAGE(Copernicus GmbH, 2021-06-17)Abstract. This paper presents a platform for multilayered documentation of cultural heritage, inspired by the current lack of a heritage BIM approach capable of creating models with both high geometric accuracy and high semantic richness. The platform is developed in the Unity game engine. It comprises several integrated and interconnected layers or datasets that can include data of different types, such as a point cloud, textured polygonal mesh, parametric information model and images, both 2D images and 360° panoramas. We present an overview of the platform concept, the benefits of the proposed multilayered representation and the details on the implementation and integration of datasets. Also, we present some of the innovative functions made possible by this integration, such as point cloud or mesh cutting and preforming measurements on 2D images and 360° panoramas. We perform and present the results of a preliminary analysis of platform functions, which indicates that the platform can be used for accurate measurement and retrieval of 3D coordinates.
-
ItemA multiclass TrAdaBoost transfer learning algorithm for the classification of mobile lidar data(Elsevier BV, 2020-08)A major challenge in the application of state-of-the-art deep learning methods to the classification of mobile lidar data is the lack of sufficient training samples for different object categories. The transfer learning technique based on pre-trained networks, which is widely used in deep learning for image classification, is not directly applicable to point clouds, because pre-trained networks trained by a large number of samples from multiple sources are not available. To solve this problem, we design a framework incorporating a state-of-the-art deep learning network, i.e. VoxNet, and propose an extended Multiclass TrAdaBoost algorithm, which can be trained with complementary training samples from other source datasets to improve the classification accuracy in the target domain. In this framework, we first train the VoxNet model with the combined dataset and extract the feature vectors from the fully connected layer, and then use these to train the Multiclass TrAdaBoost. Experimental results show that the proposed method achieves both improvement in the overall accuracy and a more balanced performance in each category.
-
ItemA Sparsity-Based Regularization Approach for Deconvolution of Full-Waveform Airborne Lidar Data(MDPI, 2016-08)Full-waveform lidar systems capture the complete backscattered signal from the interaction of the laser beam with targets located within the laser footprint. The resulting data have advantages over discrete return lidar, including higher accuracy of the range measurements and the possibility of retrieving additional returns from weak and overlapping pulses. In addition, radiometric characteristics of targets, e.g., target cross-section, can also be retrieved from the waveforms. However, waveform restoration and removal of the effect of the emitted system pulse from the returned waveform are critical for precise range measurement, 3D reconstruction and target cross-section extraction. In this paper, a sparsity-constrained regularization approach for deconvolution of the returned lidar waveform and restoration of the target cross-section is presented. Primal-dual interior point methods are exploited to solve the resulting nonlinear convex optimization problem. The optimal regularization parameter is determined based on the L-curve method, which provides high consistency in varied conditions. Quantitative evaluation and visual assessment of results show the superior performance of the proposed regularization approach in both removal of the effect of the system waveform and reconstruction of the target cross-section as compared to other prominent deconvolution approaches. This demonstrates the potential of the proposed approach for improving the accuracy of both range measurements and geophysical attribute retrieval. The feasibility and consistency of the presented approach in the processing of a variety of lidar data acquired under different system configurations is also highlighted.