Leveraging the power of multi-core platforms for large-scale geospatial data processing: Exemplified by generating DEM from massive LiDAR point clouds

Processing the massive LiDAR point cloud is a time consuming process due to the magnitude of the data involved and the highly computational iterative nature of the algorithms. In particular, many current and future applications of LiDAR require real- or near-real-time processing capabilities. Relevant examples include environmental studies, military applications, tracking and monitoring of hazards. Recent advances in Graphics Processing Units (GPUs) open a new era of General-Purpose Processing on Graphics Processing Units (GPGPU). In this paper, we seek to harness the computing power available on contemporary Graphic Processing Units (GPUs), to accelerate the processing of massive LiDAR point cloud. We propose a CUDA-based method capable of accelerating processing of massive LiDAR point cloud on the CUDA-enabled GPU. Our experimental results showed that we are able to significantly reduce processing time of constructing TIN from LiDAR point cloud with GPGPU based parallel processing implementation, in comparison with the current state-of-the-art CPU-based algorithms.

Individual tree detection from airborne laser scanning data based on supervoxels and local convexity

This paper presents a point-based rendering approach to visualize massive sets of 3D points in real-time. In many disciplines such as architecture, engineering, and archeology LiDAR technology is used to capture sites and landscapes; the resulting massive 3D point clouds pose challenges for traditional storage, processing, and presentation techniques. The available hardware resources of CPU and GPU are limited, and the 3D point cloud data exceeds available memory size in general. Hence out-of-core strategies are required to overcome the limit of memory. We discuss concepts and implementations of rendering algorithms and interaction techniques that make out-of-core real-time visualization and exploration of massive 3D point clouds feasible. We demonstrate with our implementation real-time visualization of arbitrarily sized 3D point clouds with current PC hardware using a spatial data structure in combination with a point-based rendering algorithm. A rendering front is used to increase the performance taking into account user interaction as well as available hardware resources. Furthermore, we evaluate our approach, describe its characteristics, and report on applications.

Compressing massive LiDAR point clouds in real-time is critical to autonomous machines such as drones and self-driving cars. While most of the recent prior work has focused on compressing individual point cloud frames, this paper proposes a novel system that effectively compresses a sequence of point clouds. The idea to exploit both the spatial and temporal redundancies in a sequence of point cloud frames. We first identify a key frame in a point cloud sequence and spatially encode the key frame by iterative plane fitting. We then exploit the fact that consecutive point clouds have large overlaps in the physical space, and thus spatially encoded data can be (re-)used to encode the temporal stream. Temporal encoding by reusing spatial encoding data not only improves the compression rate, but also avoids redundant computations, which significantly improves the compression speed. Experiments show that our compression system achieves 40× to 90× compression rate, significantly higher than the MPEG’s LiDAR point cloud compression standard, while retaining high end-to-end application accuracies. Meanwhile, our compression system has a compression speed that matches the point cloud generation rate by today LiDARs and out-performs existing compression systems, enabling real-time point cloud transmission.

3D modelling method and application to a digital campus by fusing point cloud data and image data

It takes much time to do TIN data I/O operation when converting Li DAR points cloud to raster data via TIN. An efficient streaming algorithm based on triangle driver is proposed in this paper. Firstly, a new sub algorithm for converting TIN to GRID based on the straight line positive and negative zone discriminant principle is presented to make the streaming algorithm feasible, for it performs by each triangle rather than by each grid node. Based on the sub algorithm, the streaming algorithm combines constructing triangles, converting triangles to GRID and freeing memory into a pipeline by traversing triangles to simulate the streaming computation. Instead of creating TIN and converting TIN to GRID separately in the CPU, the streaming algorithm integrates them together and creates GRID from points cloud directly. The results show that the sub-algorithm performs efficiently and there is a linear relationship between time consuming and the number of triangles. The streaming algorithm has a better performance in terms of time owing to no TIN data I/O operation and it improves RAM memory utilization. It offers an efficient way to create GRID from a great amount of Li DAR point cloud data and supports parallel computing.

Based on the whole process of computer-aided technology, a 3D animation data processing development platform based on artificial intelligence is designed and implemented. A random forest model for animation data processing and development is designed to mine the experience that can guide animation generation from the accumulated animation data. Based on the design goal and implementation principle of animation data processing and development platform, the attributes and categories of random forest model are abstracted. After standardizing a large number of historical data, the training sample set is obtained, and the random forest model is obtained after training. The parameters of the random forest model are continuously optimized by experiments, so that the learning model can better guide the dynamic animation data processing and development platform to generate animation to the satisfaction of users. The designed three-dimensional animation data processing and development platform interacts with the animation generation module, users, and system administrators. It can continuously receive the sample data of the animation generation module, automatically expand the number of training samples, analyze the status of the sample database, and put forward suggestions to the system administrator to update the learning model, so as to realize the initiative of learning. The experimental results show that the designed 3D animation data processing and development platform is effective and feasible.

Point cloud datasets provided by LiDAR have become an integral part in many research fields including archaeology, forestry, and ecology. Facilitated by technological advances, the volume of these datasets has steadily increased, with modern airborne laser scanning surveys now providing high-resolution, (super-)national scale, multi-terabyte point clouds. However, their wider scientific exploitation is hindered by the scarcity of open source software tools capable of handling the challenges of accessing, processing, and extracting meaningful information from massive datasets, as well as by the domain-specificity of existing tools. Here we present Laserchicken, a user-extendable, cross-platform Python tool for extracting statistical properties of flexibly defined subsets of point cloud data, aimed at enabling efficient, scalable, distributed processing of multi-terabyte datasets. We demonstrate Laserchicken’s ability to unlock these transformative new resources, e.g. in macroecology and species distribution modelling, where it is used to characterize the 3D vegetation structure at high resolution (<10 m) across whole countries or regions. We further discuss its potential as a domain agnostic, flexible tool that can also facilitate novel applications in other research fields.

Faced with the deficiency of storing the massive unstructured LiDAR point cloud data with local file system, a cloudizing storage method is proposed for unstructured LiDAR point cloud data based on MongoDB database system. Firstly, this paper designs the architecture of cloudizing storage server cluster for LiDAR point cloud data in the light of distributed storage framework of MongoDB. Secondly, this paper puts forward an organization model of point cloud data file chunks, in accordance with MongoDB distributed file system GridFS, based on which the point cloud file chunks' parallel sharding is achieved. Finally, the prototype system is designed and accomplished for point cloud data cloudizing storage, which is used to carry out experiments on accessing the point cloud data in GridFS, and the results approve that the cloudizing storage proposed in this paper performs better than local file system in the accessing point cloud data.

The combination of 3D acquisition (terrestrial and airborne LiDAR, structured light, structure-from-motion) and 2D imaging (photographic, multispectral, panoramic, orthorectified, reflectance transformation) techniques allows the geometry, appearance and other aspects of culturally significant sites to be objectively documented. Traditionally, these data are usually transformed into models such as 3D textured meshes before they are visualized or analyzed—an often time- and effort-intensive process. We propose a system for the direct visualization and analysis of such data, allowing the different aspects recorded to be layered together, and co-visualized with annotations and other relevant information. We describe the required technical foundations, including gigapoint and gigapixel visualization pipelines that enable the dynamic layering of high-resolution imagery over massive minimally-processed LiDAR point clouds that serve as the base spatial layer. In particular, we introduce the pointbuffer—a GPU-resident viewdependent point cache—as the foundation of our gigapoint pipeline, and outline the use of virtual texturing for draping of gigapixel imagery onto point clouds. Finally, we present case studies from sites in Jordan and Italy.

This document introduces a new application for rendering massive LiDAR point cloud data sets of interior environments within highresolution immersive VR display systems. Overall contributions are: to create an application which is able to visualize large-scale point clouds at interactive rates in immersive display environments, to develop a flexible pipeline for processing LiDAR data sets that allows display of both minimally processed and more rigorously processed point clouds, and to provide visualization mechanisms that produce accurate rendering of interior environments to better understand physical aspects of interior spaces. The work introduces three problems with producing accurate immersive rendering of Li-DAR point cloud data sets of interiors and presents solutions to these problems. Rendering performance is compared between the developed application and a previous immersive LiDAR viewer.

Abstract. Build on soft soil, close to sea level the Netherlands is at high risk for the effects of subsidence and deformation. Interferometric Synthetic Aperture Radar (InSAR) is successfully used to monitor the deformation trends at millimetre level. Unfortunately the InSAR deformation trends suffer from poor geolocation estimates, limiting the ability to link deformation behaviour to objects, such as buildings, streets or bridges. A nationwide, high resolution, airborne LiDAR point cloud is available in the Netherlands. Although the position accuracy of this LiDAR point cloud is to low for deformation estimates, linking the InSAR location to the geometries outlined by the LiDAR point can improve the geolocation estimates of the InSAR trends. To our knowledge no such integration is available as of yet. In this article we outline methods to link deformation estimates to the LiDAR point cloud and give an outlook of possible improvements. As a test we link 3.1 million TerraSAR-X InSAR Persistent Scatterers to 3 billion LiDAR points, covering the city of Delft and surroundings.

We present MCF3D, a multi-stage complementary fusion three-dimensional (3D) object detection network for autonomous driving, robot navigation, and virtual reality. This is an end-to-end learnable architecture, which takes both LIDAR point clouds and RGB images as inputs and utilizes a 3D region proposal subnet and second stage detector(s) subnet to achieve high-precision oriented 3D bounding box prediction. To fully exploit the strength of multimodal information, we design a series of fine and targeted fusion methods based on the attention mechanism and prior knowledge, including &#x201C;pre-fusion,&#x201D; &#x201C;anchor-fusion,&#x201D; and &#x201C;proposal-fusion.&#x201D; Our proposed RGB-Intensity form encodes the reflection intensity onto the input image to strengthen the representational power. Our designed proposal-element attention module allows the network to be guided to focus more on efficient and critical information with negligible overheads. In addition, we propose a cascade-enhanced detector for small classes, which is more selective against close false positives. The experiments on the challenging KITTI benchmark show that our MCF3D method produces state-of-the-art results while running in near real-time with a low memory footprint.

This paper proposed an organization method of laser point cloud data for the efficient management and rapid visualization of the massive point cloud data of vehicle-mounted laser scanner. In this paper, both the original point cloud data and its trajectory are sectioned for the fast indexing firstly, then the LOD(levels of details) index of each section is build based on octree structure. With a tile type and multiresolution storage mode based on folder system, the depth of octree structure is represented by the level of folder directory, and in every node folder, the corresponding point cloud data file and its node properties file are both include. The storage method of this paper decreases the preprocessing time greatly and can support concurrent access owing to mutually independence of each node. Moreover, with the application of view frustum culling technology and multi-thread dynamic dispatch technology, the rendering and roaming of massive point cloud can realize real time updating according to viewpoint change, which significantly improves the dispatch efficiency of vehicle-mounted laser point cloud data. In the experiment, our method was compared with several popular software of point cloud data processing (e.g. XGRT, Quick Terrain Reader) on the data dispatch. The results show our method has obvious advantage on storage capacity, data access time, memory footprint, rendering frame rate and so on, it indicates our method is effective for the management of vehicle-mounted laser point cloud data and can improve the production efficiency for indoor-work of vehicle-mounted laser point cloud data processing.

Abstract. Space-filling curves (SFCs) have enabled transformers to process massive 3D point clouds with linear complexity by mapping them into 1D sequences. However, standard SFCs rely on fixed, axis-aligned traversal orders, which introduce systematic axial anisotropy and frequently sever the connectivity of oblique geometric structures – a phenomenon we term “locality breaches”. While existing methods attempt to mitigate this by sequentially stacking multiple patterns, they incur a prohibitive linear increase in latency. To resolve this efficiency–accuracy dilemma, we propose GAPS (Group-wise Affine-Perturbed Serialization). Mathematically, we derive a flexible mechanism to generate diverse, bijective scanning paths via affine transformations over the Galois field 𝔽2. This allows for the synthesis of “sheared” and “rotated” traversals that effectively establish neighborhood connectivity along non-axial geometries. Architecturally, GAPS employs a group-wise parallel design that splits feature channels to aggregate multiple-perspective contexts simultaneously, thereby circumventing the computational penalty of serial execution. Extensive experiments on the ScanNet200 benchmark demonstrate that GAPS achieves 36.8 % mean Intersection over Union (mIoU), outperforming the strong baseline, Point Transformer V3 (PTv3), by 1.6 %. Remarkably, GAPS delivers this significant gain with zero parameter overhead and minimal latency overhead, establishing a new state-of-the-art performance among standard train-from-scratch 3D backbones.