DATA OPTIMIZATION FOR 3D MODELING AND ANALYSIS OF A FORTRESS ARCHITECTURE

Fast automatic three-dimensional road model reconstruction based on mobile laser scanning system

Recently, terrestrial laser scanners have been significantly improved in terms of accuracy, measurement distance, measurement speed, and resolution. They enable us to capture dense 3D point clouds of large-scale objects and fields, such as factories, engineering plants, large equipment, and transport ships. In addition, the mobile mapping system, which is a vehicle equipped with laser scanners and GPSs, can be used for capturing large-scale point clouds from a wide range of roads, buildings, and roadside objects. Large-scale point clouds are useful in a variety of applications, such as renovation and maintenance of facilities, engineering simulation, asset management, and 3D mapping. To realize these applications, new techniques must be developed for processing large-scale point clouds. So far, point processing has been studied mainly for relatively small objects in the field of computer-aided design and computer graphics. However, in recent years, the application areas of point clouds are not limited to conventional domains, but also include manufacturing, civil engineering, construction, transportation, forestry, and so on. This is because the state-of-the-art laser scanner can be used to represent large objects or fields as dense point clouds. We believe that discussing new techniques and applications related to large-scale point clouds beyond the boundaries of traditional academic fields is very important.This special issue addresses the latest research advances in large-scale point cloud processing. This covers a wide area of point processing, including shape reconstruction, geometry processing, object recognition, registration, visualization, and applications. The papers will help readers explore and share their knowledge and experience in technologies and development techniques.All papers were refereed through careful peer reviews. We would like to express our sincere appreciation to the authors for their submissions and to the reviewers for their invaluable efforts for ensuring the success of this special issue.

Mobile vision-based autonomous vehicles use video frames from multiple angles to construct a 3D model of their environment. In this paper, we present a post-processing adaptive noise suppression technique to enhance the quality of the computed 3D model. Our near real-time reconstruction algorithm uses each pair of frames to compute the disparities of tracked feature points to translate the distance a feature has traveled within the frame in pixels into real world depth values. As a result these tracked feature points are plotted to form a dense and colorful point cloud. Due to the inevitable small vibrations in the camera and the mismatches within the feature tracking algorithm, the point cloud model contains a significant amount of misplaced points appearing as noise. The proposed noise suppression technique utilizes the spatial information of each point to unify points of similar texture and color into objects while simultaneously removing noise dissociated with any nearby objects. The noise filter combines all the points of similar depth into 2D layers throughout the point cloud model. By applying erosion and dilation techniques we are able to eliminate the unwanted floating points while retaining points of larger objects. To reverse the compression process, we transform the 2D layer back into the 3D model allowing points to return to their original position without the attached noise components. We evaluate the resulting noiseless point cloud by utilizing an unmanned ground vehicle to perform obstacle avoidance tasks. The contribution of the noise suppression technique is measured by evaluating the accuracy of the 3D reconstruction.

The development and management of green open spaces are essential in overcoming environmental problems such as air pollution and urban warming. 3D modeling and biomass calculation are the example efforts in managing green open spaces. In this study, 3D modeling was carried out on point clouds data acquired by the UAV photogrammetry and UAV LiDAR methods. 3D modeling is done explicitly using the point clouds fitting method. This study uses three fitting methods: the spherical fitting method, the ellipsoid fitting method, and the spherical harmonics fitting method. The spherical harmonics fitting method provides the best results and produces an R2 value between 0.324 to 0.945. In this study, Above-Ground Biomass (AGB) calculations were also carried out from the modeling results using three methods with UAV LiDAR and Photogrammetry data. AGB calculation using UAV LiDAR data gives better results than using photogrammetric data. AGB calculation using UAV LiDAR data gives an accuracy of 78% of the field validation results. However, for visualization purposes with a not-too-wide area, a 3D model of photogrammetric data using the spherical harmonics method can be used.

The 3D modeling of urban buildings has become a common research area in various disciplines such as photogrammetry and computer vision, with different applications such as intelligent city management, navigation of self-driving cars and architecture, just to name a few. The objective of this study is to produce a 3D model of the external facade of the buildings with the required precision, accuracy and level of detail according to the user’s requirements, while minimizing time and cost. This research focuses on the production of 3D models for blocks of residential buildings in Tehran, Iran. The Insta 360 One X2 spherical camera is selected to capture the data due to its low cost and 360 × 180° field of view. The camera specifications have facilitated more efficient data collection in terms of both time and cost. The proposed modeling method is based on extracting lines of external walls through the utilization of the point cloud density concept. Initially, photogrammetric point clouds are generated in with a reconstruction precision of 0.24 m from spherical camera images. In the next step, the 3D point cloud is projected into a 2D point cloud by setting the height component to zero. The 2D point cloud is then rotated based on the direction angle determined by the Hough transform so that the perpendicular walls are parallel to the axes of the coordinate system. Next, a 2D point cloud density analysis is performed by voxelizing the point cloud and counting the number of points in each voxel in both the horizontal and vertical directions. By determining the peaks in the density plot, the lines of the external vertical and horizontal walls are extracted. To extract the diagonal external walls, the density analysis is performed in the direction of the first principal component. Finally, by determining the height of each wall in the point cloud, a 3D model is created at the level of detail one. The resulting model has a precision of 0.32 m compared to real sizes, and the 2D plan has a precision of 0.31 m compared to the ground truth map. The use of the spherical camera and point cloud density analysis makes this method efficient and cost-effective, making it a promising approach for future urban modeling projects.

The variation in point cloud density is driven by many different factors. This variation is expected to affect the quality of the information extracted from the point clouds, however, the extent to which these variations impact the ability to accurately extract and assess geometric features of highways from point cloud data are unknown. This paper investigates the impacts of point density reduction on the extraction and assessment of four critical geometric features. The density of light detection and ranging (LiDAR) data was first reduced and the different features were extracted at varying levels of point density and on a selection of different highway segments in Alberta, Canada. The information obtained at lower point density was then compared to what was obtained at 100% point density. It was found that clearance assessments and sight distance assessments had low sensitivity to reductions in point density (i.e., reducing the point density to as low as 10% of the original data (30 ppm2 on the pavement surface) yielded results comparable to what was obtained at 100% density (300 ppm2) In contrast, for cross section slope estimation and curve attribute estimation higher sensitivity to point density was observed. These findings are critical for transportation agencies considering the adoption of LiDAR technology to manage elements of their infrastructure and for researchers developing data processing tools and algorithms for the semantic segmentation of transportation features from remotely sensed point clouds.

Abstract. The process of building a digital city is the construction of a three-dimensional city model. The construction of 3 d data model is the process of multi-source spatial information collection and fusion. The current means of spatial information collection are mainly divided into three-dimensional laser technology and UAV photogrammetry technology. The three-dimensional laser technology is mainly based on ground laser, which has high acquisition accuracy for the bottom part of the building, but it is insufficient for the information collection at the top of the building. UAV photogrammetry has a large area and high efficiency spatial information collection method. At the same time, a large three-dimensional city model of the building can also be constructed. However, due to practical problems such as flight height and ground occlusion, the model constructed by UAV photogrammetry has a perfect top and poor bottom quality. The two methods are combined to build a more perfect point cloud model. Aiming at the above problems, a three-dimensional modeling method of buildings under multi-source data fusion is proposed. Firstly, the point cloud data and image data obtained by 3D laser scanner and UAV are preprocessed, and the two data forms are transformed into high-precision point cloud data. Then two sets of point cloud data are fused to generate the whole point cloud data. Finally, based on the patch generated by the whole point cloud, the point cloud model is reverse modeled. Get the corresponding building entity model, in order to get a more perfect 3D city model.

Abstract. Photorealistic 3D models play an important role in various applications related to smart cities. Texture mapping plays a crucial role as the last step of 3D reconstruction. The texture of the model directly reflects the visualization and realism of the model. In complex city scenes, objects often have varying degrees of occlusion from one another. Traditional methods to circumvent occlusion by image or model data are not very effective. In this paper, we propose an automatic texture mapping method for 3D models to circumvent occlusion through 3D spatial through-view relationships between images, models, and point clouds. First of all, calculate the density and voxel resolution of the point clouds, the point clouds here can be either dense matching point clouds or laser point clouds. Voxelization is based on point clouds, voxels generated by voxelization can reflect the spatial location of the point clouds. Secondly, to determine whether there is an occlusion of voxels between the image and the model vertices based on the spatial through-view relationship between the image, the model, and the voxels, and filter out the unobstructed images. Finally, by optimizing the data items of the Markov random field, the shading function is added to the original data items for evaluating the shading condition, based on which the most suitable image is selected for texture mapping of a triangular surface of the model. Loop computation to complete the unobstructed texture mapping of the entire model. In this paper, the public test datasets released by the International Society for Photogrammetry and Remote Sensing (ISPRS) and several urban area datasets collected by ourselves are incorporated to test the performance of the proposed method. Both qualitative and quantitative cross-sectional comparisons with existing texture mapping algorithms are conducted and presented. Experimental results indicate that our method can effectively circumvent occlusion for automated texture mapping, and generate 3D models with fewer obstructed textures, which largely improves the visual quality and realism of the models.

In reverse engineering, the point cloud data processing is the key, it will directly affect the quality and efficiency of the subsequent reconstruction of the surface. However, the amount of data obtained by three-dimensional laser scanning is often larger and redundant, which will increase the difficulty of the surface reconstruction, therefore, the point cloud data is needed to streamline. Depth analysis of the efficiency and accuracy of curvature grid method, a secondary streamlined algorithm is proposed, which take use of non-uniform grid streamline law based on octree to streamline to the point cloud, and re-use curvature streamline method based on B-spline surface fitting to streamline it. This method, whose streamline ratio is relatively high, can effectively retain the feature point cloud data, providing a great convenience for the follow-up of reverse engineering.

Airborne laser scanning (LiDAR) point clouds over large forested areas can be processed to segment individual trees and subsequently extract tree-level information. Existing segmentation procedures typically detect more than 90% of overstory trees, yet they barely detect 60% of understory trees because of the occlusion effect of higher canopy layers. Although understory trees provide limited financial value, they are an essential component of ecosystem functioning by offering habitat for numerous wildlife species and influencing stand development. Here we model the occlusion effect in terms of point density. We estimate the fractions of points representing different canopy layers (one overstory and multiple understory) and also pinpoint the required density for reasonable tree segmentation (where accuracy plateaus). We show that at a density of ~170 pt/m² understory trees can likely be segmented as accurately as overstory trees. Given the advancements of LiDAR sensor technology, point clouds will affordably reach this required density. Using modern computational approaches for big data, the denser point clouds can efficiently be processed to ultimately allow accurate remote quantification of forest resources. The methodology can also be adopted for other similar remote sensing or advanced imaging applications such as geological subsurface modelling or biomedical tissue analysis.

Point cloud up-sampling network with multi-level spatial local feature aggregation

The generation of dense and accurate point clouds from airborne light detection and ranging (LiDAR) waveform data is crucial to forest inventory. This work proposes a deconvolution method with: 1) an automatic stopping criterion to differentiate near-adjacent targets and 2) an iterative false subwaveform removal algorithm to remove outliers caused by noise. Synthetic waveforms with different overlap rates were processed using the proposed method, the Gaussian decomposition (GD) method, and the Richardson Lucy (RL) deconvolution method. Results showed that: 1) the number of subwaveforms detected by the proposed method is 9% higher than that of the RL and 20% higher than that of the GD when the overlap rates are larger than 0.6 and 2) the proposed method is of the smallest ground and peak distance errors. Results of the indoor experiment also show that the proposed method is superior in finding near targets meanwhile leading to small ground and peak distance error. Furthermore, the proposed method was tested by airborne waveforms from the Dagujia forest farm. The point cloud density acquired by the proposed method is 3% and 35% higher than that by the RL and GD method. Fewer outliers are produced by the proposed method. The number of individual trees extracted from the proposed point clouds is 22%, 51%, and 57% greater than those extracted from the RL, the GD, and the reference point clouds using the canopy height model-based method. Best individual tree extraction result is produced by the proposed method, especially for an area with small trees.

Abstract. Photo-realistic 3D model of object cannot be created by using conventional Total Station method. The aim of this study is to generate 3D building model using UAV’s dense point clouds. In this research, the ability of remotely detected information for the construction of 3D building models is found to be as specific as UAV datasets, Tacheometry Survey and Traverse Survey. Pix4D and MeshLab software were used in this study to perform image processing of 3D building. The building was built up upon using the information needed from capturing UAV images. This procedure requires the height of building to be measured through conventional method such that of UAV and the height is then compared with a 3D building model. At the end of the process, the comparison of height of building, also known as RMSE, was conducted. The images of building were then processed in Photogrammetry which is the Pix4D software. Using Pix4D, this study proceeded with several processing stages such as camera calibration, photo uploading, building of dense cloud, mesh generation, and model texturing. Those processes were done automatically by using Pix4D tools. High quality of photo-realistic 3D model of building was then successfully constructed. The accuracy assessment was done by comparing the measurement of 3D model of building. Direct measuring of building by using total station had taken place to compare the measurements between 3D model details with actual measurements. The measurement of 3D model of building was done in Pix4D software. The RMSE achieved in this study is 0.015 m which is under reasonable accuracy limit. This study proves that UAV point cloud is capable to produce a very fine 3D model.

Aiming at the problems of low accuracy and low efficiency of traditional point cloud target classification methods, this paper designs a new classification method based on improved random forest algorithm. Bagging is combined with random subspace to form a subset of feature training at random, so that the generalisation ability of random forest algorithm can be increased while the data processing speed can be accelerated to avoid overfitting phenomenon. On the basis of extracting geometric features of coloured point clouds, the optimal feature subset for classification is determined, and then the dense matching point clouds are classified using the improved random forest algorithm. Experimental results show that the classification error rate of this method is less than 1%, the average classification process takes only 83.995 s, and the VIM value is all over 0.1, indicating that this method can effectively improve the classification effect of dense matching point cloud targets.

Masked autoencoding has become a successful pretraining paradigm for Transformer models for text, images, and, recently, point clouds. Raw automotive datasets are suitable candidates for self-supervised pre-training as they gener-ally are cheap to collect compared to annotations for tasks like 3D object detection (OD). However, the development of masked autoencoders for point clouds has focused solely on synthetic and indoor data. Consequently, existing meth-ods have tailored their representations and models toward small and dense point clouds with homogeneous point den-sities. In this work, we study masked autoencoding for point clouds in an automotive setting, which are sparse and for which the point density can vary drastically among ob-jects in the same scene. To this end, we propose Voxel-MAE, a simple masked autoencoding pre-training scheme designed for voxel representations. We pre-train the back-bone of a Transformer-based 3D object detector to reconstruct masked voxels and to distinguish between empty and non-empty voxels. Our method improves the 3D OD performance by 1.75 mAP points and 1.05 NDS on the challenging nuScenes dataset. Further, we show that by pre-training with Voxel-MAE, we require only 40% of the annotated data to outperform a randomly initialized equivalent. Code is available at https://github.com/georghess/voxel-mae.