Automatic extraction of buildings from remote sensing imagery plays a significant role in many applications, such as urban planning and monitoring changes to land cover. Various building segmentation methods have been proposed for visible remote sensing images, especially state-of-the-art methods based on convolutional neural networks (CNNs). However, high-accuracy building segmentation from high-resolution remote sensing imagery is still a challenging task due to the potentially complex texture of buildings in general and image background. Repeated pooling and striding operations used in CNNs reduce feature resolution causing a loss of detailed information. To address this issue, we propose a light-weight deep learning model integrating spatial pyramid pooling with an encoder-decoder structure. The proposed model takes advantage of a spatial pyramid pooling module to capture and aggregate multi-scale contextual information and of the ability of encoder-decoder networks to restore losses of information. The proposed model is evaluated on two publicly available datasets; the Massachusetts roads and buildings dataset and the INRIA Aerial Image Labeling Dataset. The experimental results on these datasets show qualitative and quantitative improvement against established image segmentation models, including SegNet, FCN, U-Net, Tiramisu, and FRRN. For instance, compared to the standard U-Net, the overall accuracy gain is 1.0% (0.913 vs. 0.904) and 3.6% (0.909 vs. 0.877) with a maximal increase of 3.6% in model-training time on these two datasets. These results demonstrate that the proposed model has the potential to deliver automatic building segmentation from high-resolution remote sensing images at an accuracy that makes it a useful tool for practical application scenarios.INDEX TERMS Deep learning, high-resolution remote sensing imagery, building extraction, fully convolutional networks, encoder-decoder.

We use a finite element (FEM) formulation of the level set method to model geological fluid flow problems involving interface propagation. Interface problems are ubiquitous in geophysics. Here we focus on a Rayleigh-Taylor instability, namely mantel plumes evolution, and the growth of lava domes. Both problems require the accurate description of the propagation of an interface between heavy and light materials (plume) or between high viscous lava and low viscous air (lava dome), respectively. The implementation of the models is based on Escript which is a Python module for the solution of partial differential equations (PDEs) using spatial discretization techniques such as FEM. It is designed to describe numerical models in the language of PDEs while using computational components implemented in C and C++ to achieve high performance for time-intensive, numerical calculations. A critical step in the solution geological flow problems is the solution of the velocitypressure problem. We describe how the Escript module can be used for a high-level implementation of an efficient variant of the well-known Uzawa scheme (Arrow et al., 1958). We begin with a brief outline of the Escript modules and then present illustrations of its usage for the numerical solutions of the problems mentioned above.

[1] Dynamic simulations of homogeneous, heterogeneous and bimaterial fault rupture using modified slip-weakening frictional laws with static restrengthening are presented giving rise to both crack-like and pulse-like rupture. We demonstrate that pulse-like rupture is possible by making a modification of classical slip-weakening friction to include static restrengthening. We employ various slip-weakening frictional laws to examine their effect on the resulting earthquake rupture speed, size and mode. More complex rupture characteristics were produced with more strongly slip-weakening frictional laws, and the degree of slip-weakening had to be finely tuned to reproduce realistic earthquake rupture characteristics. Rupture propagation on a fault is controlled by the constitutive properties of the fault. We provide benchmark tests of our method against other reported solutions in the literature. We demonstrate the applicability of our elastoplastic fault model for modeling dynamic rupture and wave propagation in fault systems, and the rich array of dynamic properties produced by our elastoplastic finite element fault model. These are governed by a number of model parameters including: the spatial heterogeneity and material contrast across the fault, the fault strength, and not least of all the frictional law employed. Asymmetric bilateral fault rupture was produced for the bimaterial case, where the degree of material contrast influenced the rupture speed in the different propagation directions.

Automatic building extraction based on high-resolution aerial images has important applications in urban planning and environmental management. In recent years advances and performance improvements have been achieved in building extraction through the use of deep learning methods. However, the design of existing models focuses attention to improve accuracy through an overflowing number of parameters and complex structure design, resulting in large computational costs during the learning phase and low inference speed. To address these issues, we propose a new, efficient end-to-end model, called ARC-Net. The model includes residual blocks with asymmetric convolution (RBAC) to reduce the computational cost and to shrink the model size. In addition, dilated convolutions and multi-scale pyramid pooling modules are utilized to enlarge the receptive field and to enhance accuracy. We verify the performance and efficiency of the proposed ARC-Net on the INRIA Aerial Image Labeling dataset and WHU building dataset. Compared to available deep learning models, the proposed ARC-Net demonstrates better segmentation performance with less computational costs. This indicates that the proposed ARC-Net is both effective and efficient in automatic building extraction from high-resolution aerial images.

No abstract

No abstract

Density is a key driver of tectonic processes, but it is a difficult property to define well in the lithosphere because the gravity method is non-unique, and because converting to density from seismic velocity models, themselves non-unique, is also highly uncertain. Here we use a new approach to define the lithospheric density field of Australia, covering from 100 • E to 165 • E, from 5 • N to 55 • S and from the crust surface to 300 km depth. A reference model was derived primarily from the recently released Australian Seismological Reference Model, and refined further using additional models of sedimentary basin thickness and crustal thickness. A novel form of finite-element method based deterministic gravity inversion was applied in geodetic coordinates, implemented within the open-source escript modelling environment. Three spatial resolutions were modelled: half-, quarter-and eighth-degree in latitude and longitude, with vertical resolutions of 5, 2.5 and 1.25 km, respectively. Parameter sweeps for the key inversion regularization parameters show that parameter selection is not scale dependent. The sweep results also show that finer resolutions are more sensitive to the uppermost crust, but less sensitive to the mid-to lower-crust and uppermost mantle than lower resolutions. All resolutions show similar sensitivity below about 100 km depth. The final density model shows that Australia's lithospheric density field is strongly layered but also has large lateral density contrasts at all depths. Within the continental crust, the structure of the middle and lower crust differs significantly from the crystalline upper crust, suggesting that the tectonic processes or events preserved in the deep crust differ from those preserved in the shallower crust. The lithospheric mantle structure is not extensively modified from the reference model, but the results reinforce the systematic difference between the density of the oceanic and continental domains, and help identify subdivisions within each. The lithospheric static pressure field was resolved in 3D from the gravity and density fields. The pressure field model also highlights the fundamental difference between the oceanic and continental domains, with the former possessing lower pressure through most of the model. Overall pressure variability is large in the upper crust (60 MPa) but reduces significantly by −30 km elevation (20-30 MPa). By −50 km elevation, thick lower-crust generates further disequilibria (25-35 MPa) that are not compensated until −125 km elevation (10-20 MPa). Beneath −125 km elevation higher pressure is observed in the continental domain, extending to the base of the model. This indicates a lithosphere that is to a large degree isostatically compensated near the base of the felsic-intermediate continental crust, and again near the theoretical base of mature oceanic lithosphere.

SUMMARY Numerical simulations of long‐term crustal deformation reveal the important role that damage healing (i.e. fault‐zone strengthening) plays in the structural evolution of strike‐slip fault systems. We explore the sensitivity of simulated fault zone structure and evolution patterns to reasonable variations in the healing‐rate parameters in a continuum damage rheology model. Healing effectiveness, defined herein as a function of the healing rate parameters, describes the post‐seismic healing process in terms of the characteristic inter‐seismic damage level expected along fault segments in our simulations. Healing effectiveness is shown to control the spatial extent of damage zones and the long‐term geometrical complexity of strike‐slip fault systems in our 3‐D simulations. Specifically, simulations with highly effective healing form interseismically shallow fault cores bracketed by wide zones of off‐fault damage. Ineffective healing yields deeper fault cores that persist throughout the interseismic interval, and narrower zones of off‐fault damage. Furthermore, highly effective healing leads to a rapid evolution of an initially segmented fault system to a simpler through‐going fault, while ineffective healing along a segmented fault preserves complexities such as stepovers and fault jogs. Healing effectiveness and its role in fault evolution in our model may be generalized to describe how heat, fluid‐flow and stress conditions (that contribute to fault‐zone healing) affect fault‐zone structure and fault system evolution patterns.