Robust Biharmonic Skinning Using Geometric Fields

Predicting the likelihood of collapses and landslides on the German Baltic Sea coast cliffs requires a wide range of geological, hydrological, and climate data. Point clouds and images from drone surveys constitute a significant part of the data.The cliffs predominantly consist of Quaternary sediments of glacial origin with highly variable properties, often intricately interwoven. The glacial processes that contributed to these sediments' formation, shaping, and modification left heterogeneous deposits and various glacial-tectonic structures such as joints, shear planes, and oriented stones. These structures are crucial for assessing failure probabilities in cliff areas and are necessary for engineering geological slope stability analysis.CloudCompare is an open-source software supporting various point cloud analyses. It includes a FACETS plugin for extracting planes from 3D point clouds of rock bodies. The identification of discontinuities has been performed and validated by various authors using the FACETS plugin on hard rock exposures. We are testing the plugin for mapping discontinuities in unconsolidated sediments.Unconsolidated sediments like glacial till and varved silts reveal glacial discontinuities in cliff exposures. These can be documented in the field but require significant time. In point clouds, facets can be calculated using the plugin in a single step. However, unlike joints measured with a compass, these are always open surfaces on the cliff. While their formation may relate to joint systems, additional factors such as flaking, rolling, erosion, drying, frost wedging, and root growth may contribute to or independently cause the formation of these facets.We use point clouds generated from drone surveys of three cliff locations. These sites differ significantly in their geological structure and glacial deformation history. The facets are calculated from the point clouds and validated using structural data from engineering geological coastal surveys conducted in the past and our recent fieldwork. The FACETS plugin is suitable for capturing open joint surfaces on cliffs in unconsolidated sediments. However, care must be taken to ensure that the exposure of the steep coastal section does not dominate the measured discontinuity data. Slope-parallel planar surfaces in unconsolidated sediments are not always open joints. Also, shear planes and oriented stones are challenging to detect. Shear planes rarely form open surfaces due to frost wedging, and the long axes of stones cannot be calculated with the plugin method due to their rounding.The method is well-suited for rapid and reliable documentation of joints. Given the considerable annual coastal retreat of several meters at some locations, the FACETS method makes it possible to create a time series for joints to find potential changes in orientation, dip, and joint density. These structural datasets are particularly valuable for engineering geological slope stability calculations. Specifically, these data could be integrated into training deep learning algorithms as additional features to support the automatic identification of sediments forming the cliffs.

Direct immersogeometric fluid flow and heat transfer analysis of objects represented by point clouds

Nowadays, point clouds are considered as a promising representation for future immersive 3D applications. However, to recreate a 3D object or scene with high fidelity, a large number of points is required, often with high coordinates precision. Thus, efficient compression schemes are much needed for transmission and storage of point cloud data. Several types of coding techniques (e.g. 2D mapping, graph transforms, etc.) have been proposed in the literature to solve this problem. Octree-based solutions provide compression efficiency and level-of-detail scalability, especially for large static point clouds. The upcoming MPEG Geometry based Point Cloud Coding (G-PCC) solution provides high coding performance and relies on a static pruned octree, combined with a trisoup (for triangle soup) surface reconstruction. In this paper, the geometry coding process of G-PCC is enhanced by introducing mode decisions, thus enabling the trisoup in leaf nodes at multiple levels of the octree, i.e. allowing a more adaptive octree partitioning. Experimental results report average BD-rate gains of 5.3% (up to 7%) for the geometry component (point-to-plane error) on dense point clouds, with no significant impact on the color coding performance.

This paper presents how to utilize the open source Point Cloud library (PCL) to develop a series of computational registration processes efficiently and robustly. The registration is the key process of the Digital Face-Inspection (DFI) system. In 2009, the DFI system was developed to assist dentists to detect deviations of patients' face shapes before/after an orthodontic treatment. The system combined the technologies of 3D scanning and Reverse Engineering together to achieve the goal of creating a better quality environment. Regarding the visual analysis made by the DFI system, dentists can adjust their treatments to guarantee that their treatments are on the right track. This is definitely very helpful to create a high-quality dental environment. The inputs of the DFI system are a set of point clouds generated by 3D scanning systems. The data type of point cloud present an object by a huge amount of surface points. A completed DFI process requires two scanning independent scanning routines and those scanning routines generated point cloud by their reference coordinate systems. This event causes that generated point clouds can't be applied into the deviation analysis directly. Before the analysis, the DFI system introduced the Iterative Closest Point (ICP) algorithm to align two point clouds as close as possible. This is to force two point clouds in a same coordinate system. However, the ICP algorithm is a local operation process. Therefore, a coarse registration is required for finding an optimal initial alignment. The registration includes a lot of computational algorithms and makes the DFI development very complicated. Fortunately, an open source called Point Cloud Library is available for helping us to develop this registration easily and efficiently. PCL collected hundreds of functions and algorithms for handling point cloud data in various applications. This paper demonstrates how to introduce PCL to build up the DFI system. In addition, we also presented how to utilize PCL to improve the efficiency of the DFI system.

Extracting a computer model of a real scene from a sequence of views, is one of the most challenging and fundamental problems in computer vision. Stereo vision algorithms allow us to extract from the images a sparse 3D point cloud on the scene surfaces. However, computing an accurate mesh of the scene based on such poor quality data points (noise, sparsity) is very difficult. Here we describe a simple yet original approach that uses both the stereo vision extracted point cloud and the calibrated images. Our method is a three-stage process in which the first stage merges, filters and smoothes the input 3D points. The second stage builds for each calibrated image a triangular depth-map and fuses the set of depth-maps into a triangle soup that minimize violations of size and visibility constraints. Finally, a mesh is computed from the triangle soup using a reconstruction method that combines restricted Delaunay triangulation and Delaunay refinement.

Segmentation based building detection approach from LiDAR point cloud

The uprooting of trees is one of the most important biogeomorphological processes in temperate forests around the world. The direct cause of this process is the influence of hurricane-speed winds. However other factors (such as the features of soil, bedrock, topography, and stand) also play a significant role. Tree uprooting leads to the formation of root plates. Detecting the location of root plates may widen our knowledge about the contemporary impact of winds on forest ecosystem evolution. Tree uprooting involves the displacement of soil and weathered bedrock, and therefore may be considered in terms of biotransport (ie. transport of material caused by the impact of living organisms). Estimating the volume of the root plates can allow a better understanding of the scale and factors influencing the process of biotransport.Uprooted trees may be investigated with the use of LiDAR (Light Detection and Ranging) point clouds. Such data can provide a basis for creating Digital Surface Models (DSMs), which may be used to extract the location and estimate the volume of root plates. Previous research has focused on applying point clouds to detect 1) stems of fallen trees and 2) pit-mound topography. To date, as far as we know the LiDAR data were not applied to investigate root plates.The aim of the current project was to develop an automatic method for the detection of root plates and for biotransport estimation, based on point cloud data. Analysis was performed for two 100x100 m research plots located in the Babia Góra National Park (BgNP; Western Carpathians). The study plots were situated within the monitoring area established in 2005 following a catastrophic windstorm event that damaged the forest in November 2004. For the analysis two types of point clouds were applied: 1) open access point cloud from the Polish Institute of Geodesy and Cartography (minimal density: 4 points / m2, acquisition year: 2014) and 2) point cloud from BgNP (density: 40 - 55 points / m2, acquisition years: 2019-2020). Locations (GNSS receiver) and dimensions of 150 root plates measured in the field were used as validation data. DSMs in 0.25 m spatial resolution were created on the basis of three point classes: ground, low vegetation (< 0.2 m), and medium vegetation (< 2 m). Contour lines were generated every 0.1 m. Closed contours occurred frequently on the convex forms of root plates and therefore were used to extract the boundaries of potential root plates. Polygons created from closed contours were filtered using various criteria and compared to validation data to increase the accuracy of the method. The volume of root plates was estimated on the basis of DSMs and validated against the volume calculated from root plate dimensions measured in the field. The results underline the significance of LiDAR point clouds in the research on the tree uprooting process and the importance of GNSS technology to develop precise validation data with centimeter-level measurement accuracy.The study has been supported by the Polish National Science Centre (project no 2019/35/O/ST10/00032).

Neural implicit surfaces are a promising tool for geometry processing that represent a solid object as the zero level set of a neural network. Usually trained to approximate a signed distance function of the considered object, these methods exhibit great visual fidelity and quality near the surface, yet their properties tend to degrade with distance, making geometrical queries hard to perform without the help of complex range analysis techniques. Based on recent advancements in Lipschitz neural networks, we introduce a new method for approximating the signed distance function of a given object. As our neural function is made 1‐Lipschitz by construction, it cannot overestimate the distance, which guarantees robustness even far from the surface. Moreover, the 1‐Lipschitz constraint allows us to use a different loss function, called the hinge‐Kantorovitch‐Rubinstein loss, which pushes the gradient as close to unit‐norm as possible, thus reducing computation costs in iterative queries. As this loss function only needs a rough estimate of occupancy to be optimized, this means that the true distance function need not to be known. We are therefore able to compute neural implicit representations of even bad quality geometry such as noisy point clouds or triangle soups. We demonstrate that our methods is able to approximate the distance function of any closed or open surfaces or curves in the plane or in space, while still allowing sphere tracing or closest point projections to be performed robustly.

Today’s methods of acquiring Earth surface data, namely lidar and unmanned aerial vehicle (UAV) imagery, non-selectively collect or generate large amounts of points. Point clouds from different sources vary in their properties such as number of returns, density, or quality. We present a set of tools with applications for different types of points clouds obtained by a lidar scanner, structure from motion technique (SfM), and a low-cost 3D scanner. To take advantage of the vertical structure of multiple return lidar point clouds, we demonstrate tools to process them using 3D raster techniques which allow, for example, the development of custom vegetation classification methods. Dense point clouds obtained from UAV imagery, often containing redundant points, can be decimated using various techniques before further processing. We implemented and compared several decimation techniques in regard to their performance and the final digital surface model (DSM). Finally, we will describe the processing of a point cloud from a low-cost 3D scanner, namely Microsoft Kinect, and its application for interaction with physical models. All the presented tools are open source and integrated in GRASS GIS, a multi-purpose open source GIS with remote sensing capabilities. The tools integrate with other open source projects, specifically Point Data Abstraction Library (PDAL), Point Cloud Library (PCL), and OpenKinect libfreenect2 library to benefit from the open source point cloud ecosystem. The implementation in GRASS GIS ensures long term maintenance and reproducibility by the scientific community but also by the original authors themselves.

Abstract. Today’s methods of acquiring Earth surface data, namely lidar and unmanned aerial vehicle (UAV) imagery, non-selectively collect or generate large amounts of points. Point clouds from different sources vary in their properties such as number of returns, density, or quality. We present a set of tools with applications for different types of points clouds obtained by a lidar scanner, structure from motion technique (SfM), and a low-cost 3D scanner. To take advantage of the vertical structure of multiple return lidar point clouds, we demonstrate tools to process them using 3D raster techniques which allow, for example, the development of custom vegetation classification methods. Dense point clouds obtained from UAV imagery, often containing redundant points, can be decimated using various techniques before further processing. We implemented and compared several decimation techniques in regard to their performance and the final digital surface model (DSM). Finally, we will describe the processing of a point cloud from a low-cost 3D scanner, namely Microsoft Kinect, and its application for interaction with physical models. All the presented tools are open source and integrated in GRASS GIS, a multi-purpose open source GIS with remote sensing capabilities. The tools integrate with other open source projects, specifically Point Data Abstraction Library (PDAL), Point Cloud Library (PCL), and OpenKinect libfreenect2 library to benefit from the open source point cloud ecosystem. The implementation in GRASS GIS ensures long term maintenance and reproducibility by the scientific community but also by the original authors themselves.

Enforcing essential boundary conditions on domains defined by point clouds

Real-time 3D data processing is important in robotics, video games, environmental mapping, medical and many other fields. In this paper we propose a novel optimisation approach for the open source Point Cloud Library (PCL) that is frequently used for processing 3D data. Three aspects of the PCL are discussed: point cloud creation from disparity of colour image pairs, voxel grid downsample filtering to simplify point clouds and passthrough filtering to adjust the size of the point cloud. Additionally, rendering is examined. An optimisation technique based on CPU cycle measurement is proposed and applied in order to optimise those parts of the processing chain where measured performance is worst. The PCL modules thus optimised show on average an improvement in speed of 2.4x for point cloud creation, 91x for voxel grid filtering and 7.8x for the passthrough filter.

R67 dihydrofolate reductase (DHFR) catalyzes the reduction of dihydrofolate (DHF) to tetrahydrofolate using NADPH as a cofactor. This enzyme is a homotetramer possessing 222 symmetry, and a single active site pore traverses the length of the protein. A promiscuous binding surface can accommodate either DHF or NADPH, thus two nonproductive complexes can form (2NADPH or 2DHF) as well as a productive complex (NADPH.DHF). The role of water in binding was monitored using a number of different osmolytes. From isothermal titration calorimetry (ITC) studies, binding of NADPH is accompanied by the net release of 38 water molecules. In contrast, from both steady state kinetics and ITC studies, binding of DHF is accompanied by the net uptake of water. Although different osmolytes have similar effects on NADPH binding, variable results are observed when DHF binding is probed. Sensitivity to water activity can also be probed by an in vivo selection using the antibacterial drug, trimethoprim, where the water content of the media is decreased by increasing concentrations of sorbitol. The ability of wild type and mutant clones of R67 DHFR to allow host Escherichia coli to grow in the presence of trimethoprim plus added sorbitol parallels the catalytic efficiency of the DHFR clones, indicating water content strongly correlates with the in vivo function of R67 DHFR.

<p>Coastal dunes are both a vital natural coastal defence and a key ecological habitat; therefore, understanding their evolution is important to inform coastal management. Megatidal environments are the world largest tidal ranges and hence provide a unique endmember of the tidal range continuum. A study site at Crymlyn Burrows, Swansea Bay, UK is monitored here; the area was originally of applied interest due to its identification as a key receptor of the Swansea Bay Tidal Lagoon project. The study site comprises of 2km of dune frontage bounded to the west by hard sea defences and to the east by the River Neath estuary. The intertidal is characterized by a shallow slope and crescentic intertidal bars. Mean spring tidal range at the nearby Mumbles tide gauge is 8.46m; mean wave heights at a wave buoy offshore of the site (depth 9m LAT) are 0.66m and storm wave heights exceed 3m; predominant wind direction is in an alongshore – onshore direction.</p><p>A Sensefly Ebee-RTK drone with a Sony RGB camera has been used to map the dune system and the mid to upper intertidal beach on a monthly – bimonthly frequency since October 2018. Initial post-processing was conducted in the Sensefly Emotion3 software; Pix4D was then used to generate a point cloud from the georeferenced images. RTK-GPS surveyed ground control points distributed over the study area were used to improve the accuracy of the solution. Point clouds were cleaned to remove noise using Cloud Compare, an open source point cloud editor, before being interpolated onto a gridded surface. Comparison of the gridded surface against RTK-GPS surveyed points gave a vertical mean absolute error (MAE) of 0.05m over the beach area. Comparison in the dune area is more complex since the raw point cloud includes the vegetation and hence over-estimates height compared to the bare earth. Based on the raw point cloud, MAE over the dune area was 0.22m; however, when vegetation points were removed using artificial neural network based colour discrimination, the MAE was 0.05m.</p><p>Longshore variation in dune evolution is clearly evident. At the eastern and western ends of the dune system, dune progradation can be observed whereas in the central portion the frontal dune is cliffed and the dune foot position is static or eroding landward. Pressure transducers have been deployed in a longshore array at the neap high tide level to assess variation in wave energy reaching the upper intertidal over the study area.</p><p>This presentation will explore whether this variation in behavior is due to longshore variation in wave energy (erosion potential), variation in sediment availability (accretion potential) or the persistence of antecedent morphology.</p>

Abstract. This paper presents an automated and effective method for detecting planes and their intersections as edges in unorganized point clouds. The edges are subsequently extracted as vectors to a CAD environment. The software was developed within the Microsoft Visual Studio and the open source Point Cloud Library (PCL, http://pointclouds.org/) was used. The Point Cloud Library is a standalone, large scale, open project for 2D/3D image and point cloud processing. The code was written in C++. For the detection of the planes in the point cloud the RANSAC algorithm was employed. Subsequently, and according to the standard theory of Analytic Geometry the edges were determined as the intersections of these planes with each other. A straight line in 3D space is defined by one of its points, which was determined with the Lagrangian Multipliers method and a parallel vector, which was determined with the help of the cross product of two vectors on space. Finally, the algorithm and the results of the implementation of the process with real data were evaluated by performing various checks, mainly aiming to determine the accuracy of the edge detection.