A smart building energy system usually contains multiple sources such as power grids, autonomous generators, renewable resources, storage devices, and schedulable loads. Storage devices batteries, ice/heat units, water tanks play an important role in reducing cost systems since they can help sufficiently utilize resources time-of-use electricity prices. It is to plan, schedule, coordinate all the together with loads a facilitated by microgrid technology. To consider above problem...

Smart home is an emerging area that opens up a diverse set of downstream applications, such as dynamic pricing and demand response techniques, whose goals are typically to lower power consumption while providing comfortable convenient services. has been extensively studied, shown be beneficial in people's real lives. Although useful, typical smart platforms work at the "servant" level - highly dependent on user inputs with no predictions human demands which, if well addressed, would...

The efficient separation of coal and gangue in the mining process is great significance for improving efficiency reducing environmental pollution. Automatic detection key foundation gangue. In this paper, we proposed a hierarchical framework based on deep learning models. framework, Gaussian pyramid principle first used to construct multi-level training data, leading sets image features with multiple scales. Then, regional proposal networks (CG-RPN) are designed determine candidate regions...

Accurate and efficient extraction of road marking plays an important role in transportation engineering, automotive vision, automatic driving. In this article, we proposed a dense feature pyramid network (DFPN)-based deep learning model, by considering the particularity complexity marking. The DFPN concatenated its shallow channels with so that maps high resolution abundant image details can utilize features. Thus, learn hierarchical detailed designed model was trained end to for instance...

The maintenance of subway tunnels is vital to ensure the safety their daily operation. Issues experienced by shield tunnels, especially water leakages, require rapid and accurate detection diagnosis. Due large number disturbances in conventional algorithms face limitations when extracting discriminative features. To solve this problem, we propose a novel efficient deep learning model for multiscale features leakages based on mobile laser scanning (MLS) point cloud intensity images. A new...

In this paper, a fully-mapped field programmable gate array (FPGA) accelerator is proposed for artificial intelligence (AI)-based analysis of electrocardiogram (ECG). It consists 1-D convolutional neural network (CNN) and heart rate estimator, which constitute complementary dual-function analysis. The design projects each layer the CNN to hardware module on an Intel Cyclone V FPGA, virtual flatten effectively bridge feature extraction layers fully-connected layer. Also, maximizes...

Classification of airborne laser scanning (ALS) point clouds is needed in digital cities and 3-D modeling. To efficiently recognize objects ALS clouds, we propose a novel hierarchical aggregated deep feature representation method, which can adequately employ spatial association multilevel structures discrimination. In our learning model constructed to represent the discriminative each cluster structure by decreasing within-class distance increasing between-class distance. Our method...

Accurate and efficient semantic segmentation of buildings in high spatial resolution (HSR) remote sensing images is the basis for applications such as fine urban management, high-precision mapping, land resource utilization investigation, human settlement suitability evaluation. The current building extraction methods based on deep learning can obtain high-level abstract features images. However, due to limitation convolution kernel size vanishing gradient, some inaccurate, small-volume are...

It is well known that enormous computational power and a mass of memory are needed in deep neural networks. That makes them difficult to apply resource-limited environments. Therefore, many network compression acceleration technologies have come out, among which connection pruning widely applied due its effectiveness convenience. A novel method with full model capacity on multiple sparse structures proposed this paper. We design simple efficient function called Dynamic Processing Unit (DPU)...

Energy storage devices are expected to be more frequently implemented in wind farms near future. In this paper, both pumped hydro and fly wheel systems used assist a farm smooth the power fluctuations. Due significant difference response speeds of two storages types, coordination with types energy is problem. This paper presents methods for problem: two-level hierarchical model predictive control (MPC) method single-level MPC method. method, only one controller coordinates follow grid...

Point clouds of large-scale urban street scenes contain large quantities object categories and rich semantic information. The segmentation is the basis key to subsequent essential applications, such as digital twin engineering city information model. global feature point in can provide long-range context information, which critical high-quality segmentation. However, learning spatial saliency considering class label constraints often ignored representation some deep models. With regard this,...

The home appliance scheduling is a promising energy saving technique that has significant commercial potential. In this demo, Smart Home Energy Management System (SHE) developed on Android platform to schedule users' appliances. SHE monitors the power consumption identify operations of appliances using privacy preserving called Non-Intrusion Load Monitoring (NILM). are integrated with dynamic electric price and environment data mine personal demand preference operation. Finally, generates...

Road markings are one of the most important safety elements in a road network, and they play critical role traffic safety. However, automatic extraction remains technical challenge fields smart city construction driving. This paper presents an image-translation-based method obtaining 3D vectors typical from mobile laser point clouds. First, ground roughness is used as criterion to extract points based on topological relationship adjacent scan lines, feature images surface generated using...

Residential load scheduling is of great significance for the building energy saving. But it difficult current methods to discover and consider relationship between various comfort requirements uncertain demand customers. In this paper, a new method residential based on behavior analysis proposed reduce electrical cost while considering human behavior. Firstly, multi-dimensional feature-based profile identification applied identify operating status appliances from historical data. Then,...

Home energy management attracts more and attention in both academic power system. Comfort requirements of occupants are necessary to be considered, which directly affected by the human behavior. Thus profiling unpredictable comfort is one challenging problems. In this paper, a behavior cognition method proposed predict users' occasional activities estimate requirements, monitoring data load, social application mobile sensors. The load applied study pattern user's operation on various...

Urban street shadows can provide essential information for many applications, such as the assessment and protection of ecology environment, livability evaluation, etc. In this research, we propose an effective rapid method to quantify diurnal spatial changes urban shadows, by taking Beijing city example. method, explore a novel way transferring characteristics semantically segment street-level panoramic images using DeepLabv3+. Based on segmentation results, shading situation is further...

Smart home has recently attracted much research interest owing to its potential in improving the quality of human life. How obtain user's demand is most important and challenging task for appliance optimal scheduling smart home, since it highly related unpredictable behavior. In this paper, a human-centered system proposed identify user behavior, predict their schedule household appliances. Firstly, sensor data from wearable devices are monitored profile full-day Then, appliance-demand...

Due to the high splitting-gain of dense small cells, Ultra-Dense Network (UDN) is regarded as a promising networking technology achieve data rate and low latency in 5G mobile communications. In UDNs, each User Equipment (UE) may receive signals from multiple Base Stations (BSs), which impose severe interference networks turn motivates possibility using Coordinated Multi-Point (CoMP) transmissions further enhance network capacity. CoMP-based Networks, great challenge tradeoff between gain...

The undesired distribution of irregular surface astigmatism (SA) on the freeform has been major concern progressive addition lens (PAL) design. Herein, we proposed a segmented (SFS) construction method, which relies lines curvature to rule segmentation and then eliminates difference between principal curvatures correct SA. Based ray tracing numerical simulation results, SFS-PAL design superior performance in image quality within dynamic field view over conventional PAL. To verify feasibility...

Accurate and automatic detection of road surface element (such as marking or manhole cover) information is the basis key to many applications. To efficiently obtain element, we propose a content-adaptive hierarchical deep learning model detect arbitrary-oriented elements from mobile laser scanning (MLS) point clouds. In model, design densely connected feature integration module (DCFM) connect reorganize maps each stage in backbone network. Besides, prediction (HPM) innovatively use...