The high-precision generation of 3D building models is a controversial research topic in the field smart cities. However, due to limitations single-source data, existing methods cannot simultaneously balance local accuracy, overall integrity, and data scale model. In this paper, we propose novel model method based on multi-source fusion point cloud mesh with deep learning method. First, A Multi-Source Quality Evaluation Network (MS3DQE-Net) proposed for evaluating quality meshes clouds....

Although 3D point cloud classification neural network models have been widely used, the in-depth interpretation of activation neurons and layers is still a challenge. We propose novel approach, named Relevance Flow, to interpret hidden semantics networks. It delivers class activated in intermediate back-propagation manner, associates with input points visualize each layer. Specially, we reveal that has learned plane-level part-level layers, utilize normal IoU evaluate consistency both...

Semantic segmentation in 3D meshes is the classification of its constituent element(s) into specific classes or categories. Using powerful feature extraction abilities deep neural networks (DNNs), significant results have been obtained semantic various remotely sensed data formats. With increased utilization DNNs to segment data, there commensurate in-depth reviews and surveys summarizing learning-based techniques methodologies that entail these methods. However, most focused on methods...

Unsupervised domain adaptation (UDA) segmentation aims to leverage labeled source data make accurate predictions on unlabeled target data. The key is the network learn domain-invariant representations. In this work, we propose a prototype-guided multitask adversarial (PMAN) achieve this. First, an intensity-aware (IAS-Net) that leverages private intensity information of substantially facilitate feature learning domain. Second, category-level cross-domain alignment strategy introduced flee...

Large-scale Light Detection and Ranging (LiDAR) point clouds provide basic 3D information support for Augmented Reality (AR) in outdoor environments. Especially, matching 2D images across to LiDAR can establish the spatial relationship of space, which is a solution virtual-real registration AR. This paper first provides precise 2D-3D patch-volume dataset, contains paired image patches cloud volumes, by using Mobile Laser Scanning (MLS) data from urban scene. Second, we propose an end-to-end...

Unsupervised domain adaptation (UDA) is a low-cost way to deal with the lack of annotations in new domain. For outdoor point clouds urban transportation scenes, mismatch sampling patterns and transferability difference between classes make cross-domain segmentation extremely difficult. To overcome these challenges, we propose category-level adversarial framework. Firstly, multi-scale conditioned block that facilitates extract critical low-level domain-dependent knowledge reduce gap caused by...

Describing the same scene with different imaging style or rendering image from its 3D model gives us domain images. Different images tend to have a gap and local appearances, which raise main challenge on cross-domain patch matching. In this paper, we propose incorporate AutoEncoder into Siamese network, named as H-Net, of structural shape resembles letter H. The H-Net achieves state-of-the-art performance Furthermore, improved H-Net++. H-Net++ extracts invariant feature descriptors in...

Accurate high-resolution downscaling of surface climate variables (such as temperature) over urban areas has long been a critical yet unresolved research problem in the field and environmental sciences. In this paper, we propose novel physics informed neural network (PINN) based framework: DeepUrbanDownscale (DUD) for temperature estimation. Anchored process-based modeling satellite remote sensing, DUD leverages high-precision 3D point clouds to achieve accurate land (LST) estimation at an...

Background Artificial Neural Network (ANN), as a potential powerful classifier, was explored to assist psychiatric diagnosis of the Composite International Diagnostic Interview (CIDI). Method Both Back-Propagation (BP) and Kohonen networks were developed fit programmed (using 60 cases) classify neurosis, schizophrenia normal people. The cross-tested using another 222 cases. All subjects randomly selected from two mental hospitals in Beijing. Results Compared ICD-10 by psychiatrists, overall...

Semantic segmentation of building facade point clouds has diverse applications. The development semantic methods is inextricably linked to datasets. available datasets suffer from a lack abundant categories and data completeness. To compensate for these shortcomings, we propose new dataset characterized by various relatively complete 3D facades. In addition, most existing focus on fully supervised learning, which relies manually labeling large-scale cloud results in high time labor costs....

Generating 3D point clouds from a single image has attracted full attention researchers in the field of multimedia, remote sensing and computer vision. With recent proliferation deep learning, various models have been proposed for cloud generation. However, they require objects to be captured with absolutely clean backgrounds fixed viewpoints, which highly limits their application real environment. To guide generation, we propose novel network, RealPoint3D, integrate prior shape knowledge...

This paper presents a novel virtual-real registration approach for augmented reality (AR) in large-scale outdoor environments. Essentially, it is pose estimation the mobile camera images (ground images) 3D model recovered by Unmanned Aerial Vehicle (UAV) image sequence via Structure-From-Motion (SFM) technology. The considers to indirectly establish spatial relationship between 2D and space inferring transformation ground UAV rendered images. Specifically, proposed can overcome positioning...

Existing point cloud modeling datasets primarily express the precision by pose or trajectory rather than effect itself. Under this demand, we first independently construct a set of LiDAR system with an optical stage, and then build HPMB dataset based on constructed system, High-Precision, Multi-Beam, real-world dataset. Second, propose evaluation method for object-level to overcome limitation. In addition, existing methods tend generate continuous skeletons global environment, hence lacking...

The contours, one of the most significant human perceptual features, have a impact on point cloud processing. In urban scenes, contour extraction is quite challenging due to enormous number unstructured and irregular points (typically greater than 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">7</sup> points). this paper, we propose Large-scale 3D Contour Extraction Network (LCE-NET) generate contours consistent with perception outdoor...

<title>Abstract</title> Defect detection on metal surfaces is crucial for quality control and precise maintenance in various industrial scenarios. Nevertheless, surface defects demonstrate variations intricacies, making identifying high-precision a formidable task. With the development of artificial intelligence technology, defect technology has ushered huge breakthrough. Compared to manual inspection, methods based computer vision have apparent advantages efficiency accuracy, becoming...