Unstructured Road Extraction in UAV Images based on Lightweight Model

There is no unified planning standard for unstructured roads,and the morphological structures of these roads are complex and varied.It is important to maintain a balance between accuracy and speed for unstructured road extraction models.Unstructured road extraction algorithms based on deep learning have problems such as high model complexity,high computational cost,and the inability to adapt to current edge computing devices.Therefore,it is best to use lightweight network models.Considering the need for lightweight models and the characteristics of unstructured roads with different pattern shapes,such as blocks and strips,a TMB(Triple Multi-Block)feature extraction module is proposed,and the overall structure of the TMBNet network is described.The TMB module was compared with SS-nbt,Non-bottleneck-1D,and other modules via experiments.The feasibility and effectiveness of the TMB module design were proven through experiments and visualizations.The comparison experiment,using multiple convolution kernel categories,proved that the TMB module can improve the segmentation accuracy of the network.The comparison with different semantic segmentation networks demonstrates that the TMBNet network has advantages in terms of unstructured road extraction.

A typhoon-induced storm surge numerical model with GPU acceleration based on an unstructured spherical centroidal Voronoi tessellation grid

Storm surge is often the marine disaster that poses the greatest threat to life and property in coastal areas.Accurate and timely issuance of storm surge warnings to take appropriate countermeasures is an important means to reduce storm surge-related losses.Storm surge numerical models are important for storm surge forecasting.To further improve the performance of the storm surge forecast models,we developed a numerical storm surge forecast model based on an unstructured spherical centroidal Voronoi tessellation(SCVT)grid.The model is based on shallow water equations in vector-invariant form,and is discretized by Arakawa C grid.The SCVT grid can not only better describe the coastline information but also avoid rigid transitions,and it has a better global consistency by generating high-resolution grids in the key areas through transition refinement.In addition,the simulation speed of the model is accelerated by using the openACC-based GPU acceleration technology to meet the timeliness requirements of operational ensemble forecast.It only takes 37 s to simulate a day in the coastal waters of China.The newly developed storm surge model was applied to simulate typhoon-induced storm surges in the coastal waters of China.The hindcast experiments on the selected representative typhoon-induced storm surge processes indicate that the model can reasonably simulate the distribution characteristics of storm surges.The simulated maximum storm surges and their occurrence times are consistent with the observed data at the representative tide gauge stations,and the mean absolute errors are 3.5 cm and 0.6 h respectively,showing high accuracy and application prospects.

A finite volume method for global electromagnetic induction forward modeling on collocated unstructured grids

Global electromagnetic induction provides an efficient way to probe the electrical conductivity in the Earth’s deep interior.Owing to the increasing geomagnetic data especially from high-accuracy geomagnetic satellites,inverting the Earth’s three-dimensional conductivity distribution on a global scale becomes attainable.A key requirement in the global conductivity inversion is to have a forward solver with high-accuracy and efficiency.In this study,a finite volume method for global electromagnetic induction forward modeling is developed based on unstructured grids.Arbitrary polyhedral grids are supported in our algorithms to obtain high geometric adaptability.We employ a cell-centered collocated variable arrangement which allows convenient discretization for complex geometries and straightforward implementation of multigrid technique.To validate the method,we test our code with two synthetic models and compare our finite volume results with an analytical solution and a finite element numerical solution.Good agreements are observed between our solution and other results,indicating acceptable accuracy of the proposed method.

Multitarget Flexible Grasping Detection Method for Robots in Unstructured Environments

In present-day industrial settings,where robot arms performtasks in an unstructured environment,theremay exist numerousobjects of various shapes scattered in randompositions,making it challenging for a robot armtoprecisely attain the ideal pose to grasp the object.To solve this problem,a multistage robotic arm flexible grasp detection method based on deep learning is proposed.This method first improves the Faster RCNN target detection model,which significantly improves the detection ability of the model for multiscale grasped objects in unstructured scenes.Then,a Squeeze-and-Excitation module is introduced to design a multitarget grasping pose generation network based on a deep convolutional neural network to generate a variety of graspable poses for grasped objects.Finally,a multiobjective IOU mixed area attitude evaluation algorithm is constructed to screen out the optimal grasping area of the grasped object and obtain the optimal grasping posture of the robotic arm.The experimental results show that the accuracy of the target detection network improved by the method proposed in this paper reaches 96.6%,the grasping frame accuracy of the grasping pose generation network reaches 94%and the flexible grasping task of the robotic arm in an unstructured scene in a real environment can be efficiently and accurately implemented.

UPWIND SCHEME FOR IDEAL 2-D MHD FLOWS BASED ON UNSTRUCTURED MESH

An upwind scheme based on the unstructured mesh is developed to solve ideal 2-D magnetohydrodynamics （MHD） equations. The inviscid fluxes are approximated by using the modified advection upstream splitting method （AUSM） scheme, and a 5-stage explicit Runge-Kutta scheme is adopted in the time integration. To avoid the influence of the magnetic field divergence created during the simulation, the hyperbolic divergence cleaning method is introduced. The shock-capturing properties of the method are verified by solving the MHD shock-tube problem. Then the 2-D nozzle flow with the magnetic field is numerically simulated on the unstructured mesh. Computational results demonstrate the effects of the magnetic field and agree well with those from references.

Three-dimensional magnetotellurics modeling using edgebased finite-element unstructured meshes

Three-dimensional forward modeling magnetotellurics （MT） problems. We present a is a challenge for geometrically complex new edge-based finite-element algorithm using an unstructured mesh for accurately and efficiently simulating 3D MT responses. The electric field curl-curl equation in the frequency domain was used to deduce the H （curl） variation weak form of the MT forward problem, the Galerkin rule was used to derive a linear finite-element equation on the linear-edge tetrahedroid space, and, finally, a BI-CGSTAB solver was used to estimate the unknown electric fields. A local mesh refinement technique in the neighbor of the measuring MT stations was used to greatly improve the accuracies of the numerical solutions. Four synthetic models validated the powerful performance of our algorithms. We believe that our method will effectively contribute to processing more complex MT studies.

A three-dimensional semi-implicit unstructured grid finite volume ocean model

A new three-dimensional semi-implicit finite-volume ocean model has been developed for simulating the coastal ocean circulation, which is based on the staggered C-unstructured non-orthogonal grid in the hor- izontal direction and z-level grid in the vertical direction. The three-dimensional model is discretized by the semi-implicit finite-volume method, in that the free-surface and the vertical diffusion are semi-implicit, thereby removing stability limitations associated with the surface gravity wave and vertical diffusion terms. The remaining terms in the momentum equations are discretized explicitly by an integral method. The partial cell method is used for resolving topography, which enables the model to better represent irregular topography. The model has been tested against analytical cases for wind and tidal oscillation circulation, and is applied to simulating the tidal flow in the Bohal Sea. The results are in good agreement both with the analytical solutions and measurement results.

3D magnetotelluric inversions with unstructured finite-element and limited-memory quasi-Newton methods

Traditional 3D Magnetotelluric(MT) forward modeling and inversions are mostly based on structured meshes that have limited accuracy when modeling undulating surfaces and arbitrary structures. By contrast, unstructured-grid-based methods can model complex underground structures with high accuracy and overcome the defects of traditional methods, such as the high computational cost for improving model accuracy and the difficulty of inverting with topography. In this paper, we used the limited-memory quasi-Newton(L-BFGS) method with an unstructured finite-element grid to perform 3D MT inversions. This method avoids explicitly calculating Hessian matrices, which greatly reduces the memory requirements. After the first iteration, the approximate inverse Hessian matrix well approximates the true one, and the Newton step(set to 1) can meet the sufficient descent condition. Only one calculation of the objective function and its gradient are needed for each iteration, which greatly improves its computational efficiency. This approach is well-suited for large-scale 3D MT inversions. We have tested our algorithm on data with and without topography, and the results matched the real models well. We can recommend performing inversions based on an unstructured finite-element method and the L-BFGS method for situations with topography and complex underground structures.

Modeling and Defending Passive Worms over Unstructured Peer-to-Peer Networks

Passive worms can passively propagate through embedding themselves into some sharing files, which can result in significant damage to unstructured P2P networks. To study the passive worm behaviors, this paper firstly analyzes and obtains the average delay for all peers in the whole transmitting process due to the limitation of network throughput, and then proposes a mathematical model for the propagation of passive worms over the unstructured P2P networks. The model mainly takes the effect of the network throughput into account, and applies a new healthy files dissemination-based defense strategy according to the file popularity which follows the Zipf distribution. The simulation results show that the propagation of passive worms is mainly governed by the number of hops, initially infected files and uninfected files. The larger the number of hops, the more rapidly the passive worms propagate. If the number of the initially infected files is increased by the attackers, the propagation speed of passive worms increases obviously. A larger size of the uninfected file results in a better attack performance. However, the number of files generated by passive worms is not an important factor governing the propagation of passive worms. The effectiveness of healthy files dissemination strategy is verified. This model can provide a guideline in the control of unstructured P2P networks as well as passive worm defense.

Three-dimensional inversion of controlled-source audio-frequency magnetotelluric data based on unstructured fi nite-element method

We propose a new 3D inversion scheme to invert the near-and transition-zone data of CSAMT with topography accurately.In this new method,the earth was discretized into unstructured tetrahedra to fit the ragged topography and the vector fi nite-element method was adopted to obtain precise responses and good sensitivity.To simulate the attitude and shape of the transmitter,we divided a long-grounded transmitter into dipoles and integrated these dipoles to obtain good responses in the near-and transition-fi eld zones.Next,we designed an L2 norm-based objective functional and applied a standard quasi-Newton method as the optimization method to solve the inverse problem and guarantee steady convergence.We tested our 3D inversion method first on synthetic data and then on a field dataset acquired from select sites near Changbai Mountain,China.In both tests,the new inversion algorithm achieved excellent fitting between the predicted and observed data,even in near-and transition-fi eld zones,and the inversion results agreed well with the true model.These fi ndings reveal that the proposed algorithm is eff ective for 3D inversion of CSAMT data.

Two-dimensional shallow water equations with porosity and their numerical scheme on unstructured grids

In this study, porosity was introduced into two-dimensional shallow water equations to reflect the effects of obstructions, leading to the modification of the expressions for the flux and source terms. An extra porosity source term appears in the momentum equation. The numerical model of the shallow water equations with porosity is presented with the finite volume method on unstructured grids and the modified Roe-type approximate Riemann solver. The source terms of the bed slope and porosity are both decomposed in the characteristic direction so that the numerical scheme can exactly satisfy the conservative property. The present model was tested with a dam break with discontinuous porosity and a flash flood in the Toce River Valley. The results show that the model can simulate the influence of obstructions, and the numerical scheme can maintain the flux balance at the interface with high efficiency and resolution.

A Novel VOF-Type Volume-Tracking Method for Free-Surface Flows Based on Unstructured Triangular Mesh

A novel VOF-type volume-tracking method for two-dimensional free-surface flows based on the unstructured triangular mesh is presented. Owing to the inherent merit of the unstructured triangular mesh in fitting curved boundaries, this method can handle the free-surface problems with complex geometries accurately and directly, without introducing any complicated boundary treatment or artificial diffusion. The method solves the volume transport equation geometrically through the Modified Lagrangian-Eulerian Re-map （MLER） method, which is applied to advective fluid volumes. Moreover, the PLIC method is adopted to give a second-order reconstructed interface approximation. To validate this method, two advection tests were performed for the establishment of the accuracy and convergence rate of the solutions. Numerical results for these complex tests provide convincing evidence for the excellent solution quality and fidelity of the method.

GENERATION AND APPLICATION OF UNSTRUCTURED ADAPTIVE MESHES WITH MOVING BOUNDARIES

This paper presents a method to generate unstructured adaptive meshes with moving boundaries and its application to CFD. Delaunay triangulation criterion in conjunction with the automatic point creation is used to generate 2 D and 3 D unstructured grids. A local grid regeneration method is proposed to cope with moving boundaries. Numerical examples include the interactions of shock waves with movable bodies and the movement of a projectile within a ram accelerator, illustrating an efficient and robust mesh generation method developed.

A Coastal Ocean Model of Semi-Implicit Finite Volume Unstructured Grid

A two-dimensional coastal ocean model based on unstructured C-grid is built, in which the momentum equation is discretized on the faces of each cell, and the continuity equation is discretized on the cell. The model is discretized by semi-implicit finite volume method, in that the free surface is semi-implicit and the bottom friction is implicit, thereby removing stability limitations associated with the surface gravity wave and friction. The remaining terms in the momentum equations are discretized explicitly by integral finite volume method and second-order Adams-Bashforth method. Tidal flow in the polar quadrant with known analytic solution is employed to test the proposed model. Finally, the performance of the present model to simulate tidal flow in a geometrically complex domain is examined by simulation of tidal currents in the Pearl River Estuary.

Random walk search in unstructured P2P

Unstructured P2P has power-law link distribution, and the random walk in power-law networks is analyzed. The analysis results show that the probability that a random walker walks through the high degree nodes is high in the power-law network, and the information on the high degree nodes can be easily found through random walk. Random walk spread and random walk search method （RWSS） is proposed based on the analysis result. Simulation results show that RWSS achieves high success rates at low cost and is robust to high degree node failure.

Hermite WENO-based limiters for high order discontinuous Galerkin method on unstructured grids

A novel class of weighted essentially nonoscillatory （WENO） schemes based on Hermite polynomi- als, termed as HWENO schemes, is developed and applied as limiters for high order discontinuous Galerkin （DG） method on triangular grids. The developed HWENO methodology utilizes high-order derivative information to keep WENO re- construction stencils in the von Neumann neighborhood. A simple and efficient technique is also proposed to enhance the smoothness of the existing stencils, making higher-order scheme stable and simplifying the reconstruction process at the same time. The resulting HWENO-based limiters are as compact as the underlying DG schemes and therefore easy to implement. Numerical results for a wide range of flow conditions demonstrate that for DG schemes of up to fourth order of accuracy, the designed HWENO limiters can simul- taneously obtain uniform high order accuracy and sharp, es- sentially non-oscillatory shock transition.

An efficient unstructured WENO method for supersonic reactive flows

An efficient high-order numerical method for supersonic reactive flows is proposed in this article.The reactive source term and convection term are solved separately by splitting scheme.In the reaction step,an adaptive time-step method is presented,which can improve the efficiency greatly.In the convection step,a third-order accurate weighted essentially non-oscillatory（WENO）method is adopted to reconstruct the solution in the unstructured grids.Numerical results show that our new method can capture the correct propagation speed of the detonation wave exactly even in coarse grids,while high order accuracy can be achieved in the smooth region.In addition,the proposed adaptive splitting method can reduce the computational cost greatly compared with the traditional splitting method.

A Finite Volume Method with Unstructured Triangular Grids for Numerical Modeling of Tidal Current

The finite volume method （FVM） has many advantages in 2-D shallow water numerical simulation. In this study, the finite volume method is used with unstructured triangular grids to simulate the tidal currents. The Roe scheme is applied in the calculation of the intercell numerical flux, and the MUSCL method is introduced to improve its accuracy. The time integral is a two-step scheme of forecast and revision. For the verification of the present method, the Stoker＇s problem is calculated and the result is compared with the mathematically analytic solutions. The comparison indicates that the method is feasible. A sea area of a port is used as an example to test the method established here. The result shows that the present computational method is satisfactory, and it could be applied to the engineering fields.

3D Step-by-step inversion strategy for audio magnetotellurics data based on unstructured mesh

A three-dimensional(3D)step-by-step inversion strategy for audio magnetotellurics(AMT)is investigated in this study.The objective function is minimized by iteratively solving the Gauss-Newton normal equation,and the inversion region is discretized with unstructured tetrahedral elements.The inversion proceeds step-by-step from a coarse mesh to a fine mesh.In the inversion iteration process,a mesh is adaptively optimized according to the spatial gradient information about the model resistivity to fine delineate the boundaries of abnormal bodies.In the early stage of inversion execution,a coarse mesh is used for inversion,and the inversion stability is improved by reducing the number of inversion elements.In addition,mesh refinement is performed in the iterative inversion process.The inversion results obtained from the previous mesh are used as the reference and initial models for the next mesh iterative inversion.The step-by-step inversion strategy can ensure that the inversion is performed in the correct direction,improving the inversion stability and results gradually.Synthetic results show that the step-by-step inversion strategy with a Gauss-Newton method for 3D AMT inversion is stable and reliable,which lays a foundation for further practical 3D AMT data inversion.

Unstructured Grid Immersed Boundary Method for Numerical Simulation of Fluid Structure Interaction

This paper presents an improved unstructured grid immersed boundary method.The advantages of both immersed boundary method and body fitted grids which are generated by unstructured grid technology are used to enhance the computation efficiency of fluid structure interaction in complex domain.The Navier-Stokes equation was discretized spacially with collocated finite volume method and Euler implicit method in time domain.The rigid body motion was simulated by immersed boundary method in which the fluid and rigid body interface interaction was dealt with VOS(volume of solid) method.A new VOS calculation method based on graph was presented in which both immersed boundary points and cross points were collected in arbitrary order to form a graph.The method is verified with flow past oscillating cylinder.