Recently, approaches based on fully convolutional networks (FCN) have achieved state-of-the-art performance in the semantic segmentation of very high resolution (VHR) remotely sensed images. One central issue this method is loss detailed information due to downsampling operations FCN. To solve problem, we introduce maximum fusion strategy that effectively combines from deep layers and shallow layers. Furthermore, letter develops a powerful backend enhance result FCN by leveraging digital...

The iterative closest point (ICP) algorithm is widely used in three-dimensional (3D) cloud registration, and it very stable robust. However, its biggest drawback being easily trapped a local optimal solution, which results the incorrect registration result. Currently, there neither clear effective range to define whether ICP will fall into optimum nor study providing comprehensive evaluation of algorithm. In this paper, we take overlap ratio, angle, distance, noise as influencing factors...

The paper proposes a Dynamic ResBlock Generative Adversarial Network (DRB-GAN) for artistic style transfer. code is modeled as the shared parameters ResBlocks connecting both encoding network and transfer network. In network, class-aware attention mechanism used to attend feature representation generating codes. multiple are designed integrate extracted CNN semantic then feed into spatial window Layer-Instance Normalization (SW-LIN) decoder, which enables high-quality synthetic images with...

This paper presents a novel topologically aware 2.5-D building modeling methodology from airborne laser scanning point clouds. The reconstruction process consists of three main steps: primitive clustering, boundary representation, and geometric modeling. In we propose an enhanced probability density clustering algorithm to cluster the rooftop primitives by taking into account topological consistency among primitives. second step, employ Voronoi subgraph-based seamlessly trace boundaries....

Sharded blockchain offers scalability, decentralization, immutability, and linear improvement, making it a promising solution for addressing the trust problem in large-scale collaborative IoT. However, high proportion of cross-shard transactions can severely limit performance decentralized blockchain. Furthermore, dynamic assemblage characteristic sensing sharded is often ignored. To overcome these limitations, we propose HMMDShard, sharding scheme based on Hidden Markov Model. HMMDShard...

Point cloud registration plays a central role in various applications, such as 3D scene reconstruction, preservation of cultural heritage and deformation monitoring. The point data are usually huge. Processing huge is very time-consuming, so fast accurate method crucial. However, the existing methods still have high computation complexity or low accuracy. To address this issue, we develop for terrestrial clouds. projects clouds onto horizontal plane. Therefore, our processes 2D space,...

Edges in mobile light detection and ranging (lidar) point clouds are important for many applications but usually overlooked. In this letter, we propose a fast edge extraction method lidar. First, an index based on geometric center is introduced then gradients unorganized 3-D defined. By analyzing the ratio between eigenvalues, candidates can be detected. Finally, linking algorithm named graph snapping proposed. The tested extensively experimental results demonstrate that proposed able to...

Power-gating is a promising technique to mitigate the increasing static power of on-chip routers. Clos networks are potentially good targets for power-gating because their path diversity and decoupling between processing elements most While power-gated can perform better than direct such as meshes, significant performance penalty exists when conventional techniques used. In this paper, we propose an effective scheme, called MP3 (Minimal Performance Penalty Power-gating), which able achieve...

In this paper, a novel segmentation and recognition approach to automatically extract street lighting poles from mobile LiDAR data is proposed. First, points on or around the ground are extracted removed through piecewise elevation histogram method. Then, new graph-cut-based method introduced each cluster obtained Euclidean distance clustering algorithm. addition spatial information, pole's shape point's intensity information also considered formulate energy function. Finally,...

Automatic extraction of road curbs from uneven, unorganized, noisy, and massive 3-D point clouds is a challenging task. Existing methods often project onto 2-D planes to extract curbs. However, the projection causes loss information, which degrades performance detection. This paper presents robust, accurate, efficient method mobile LiDAR clouds. Our consists two steps: 1) extracting candidate points based on proposed novel energy function 2) refining using least cost path model. We evaluated...

Building extraction is a fundamental research topic in remote sensing image interpretation. Convolutional neural network (CNN)-based building algorithms have achieved high accuracy but require large account of parameters and calculations, which hinders the practical application these algorithms. To address challenge, we propose lightweight (RSR-Net) for from images. The consists three basic units with only few parameters, uses idea fusion shallow features deep features, proposed by U-Net....

Urban modeling from LiDAR point clouds is an important topic in computer vision, graphics, photogrammetry and remote sensing. 3D city models have found a wide range of applications smart cities, autonomous navigation, urban planning mapping etc. However, existing datasets for mainly focus on common objects such as furniture or cars. Lack building has become major obstacle applying deep learning technology to specific domains modeling. In this paper, we present urban-scale dataset consisting...

The success of transformer networks in the natural language processing and 2D vision domains has encouraged adaptation transformers to 3D computer tasks. However, most existing approaches employ standard backpropagation (SBP). SBP requires storage model activations on a forward pass for use during backward pass, making their memory complexity linearly proportional depth, hence, inefficient. Furthermore, point classic QKV matrix multiplication design which comes with bottleneck. To address...