For the detection of marine ship objects in radar ima- ges, large-scale networks based on deep learning are difficult to be deployed existing radar-equipped devices.This paper proposes a lightweight convolutional neural network, LiraNet, which combines idea dense connections, residual connections and group convolution, including stem blocks extractor modules.The designed block uses series small convolutions extract input image features, network adopts two-way connection module, further...

Data with time attributions such as price, load, and stock, which directly reflect the variation tendency, are most common type of data character available. However, it is difficult to predict complex volatile time-series data. Further, density cluster methods employ existing train initial radius; however, a certain radius hard be made suitable for continuously generated on-going datasets. Therefore, how select timespan according in way that makes possible support an adaptive updated...

With the rapid development of unmanned combat aerial vehicle (UCAV)-related technologies, UCAVs are playing an increasingly important role in military operations. It has become inevitable trend future air battlefields that complete tasks independently to acquire superiority. In this paper, UCAV maneuver decision problem continuous action space is studied based on deep reinforcement learning strategy optimization method. The platform model was established. Focusing insufficient exploration...

Video action recognition based on skeleton nodes is a highlighted issue in the computer vision field. In real application scenarios, large number of and behavior occlusion problems between individuals seriously affect speed accuracy. Therefore, we proposed lightweight multi-stream feature cross-fusion (L-MSFCF) model to recognize abnormal behaviors such as fighting, vicious kicking, climbing over wall, et al., which could obviously improve node calculation, accuracy occluded prediction...

Reports show that DDoS attacks are ubiquitous on the Internet and may jeopardize networks' stable operation. In order to understand nature of this threat further enable effective control management, a whole picture Internet-wide is necessity. Traditional methods use darknets end. However, with IPv4 address space exhaustion, become hard acquire. paper, we seek detect using live network. particular, focus most prevalent SYN flooding attacks. First, complete attack scenario model introduced...

The node capture attack in wireless sensor networks (WSNs) can be decomposed into three stages: physically of node, redeployment compromised and rejoin the network for various insider attacks. A well accepted belief - that physical is easy to implement its detection difficult has directed majority research effort defense stages two. was recently proved false [4]. discovery made first stage an attractive tactic. This paper proposes a new approach detect at stage. based upon missing...

The node capture attack on wireless sensor networks (WSNs) can be broken into three stages: capture, redeployment, and insider attacks. Existing detection techniques of the are mostly implemented in second stage. Recent discovery proved feasibility quick effective first We propose a hello message based stage scheme, which is faster, easier to execute, more reliable than existing approaches. Two implementations FSD SEFSD provided. completely decentralized. utilizes base station achieve energy...

This paper presents a numerical simulation of the evolution main particles in H2/O2 mixture plasma under different initial temperature, and gives time-variation laws density charged neutral after discharge begins. The results show that active is reduced with time, required time for equilibrium chemical reactions reduces increasing temperature.

Internet of Moving Things are connected to a variety different types sensors form world moving things, including people, animals, vehicles, drones, and boats, etc. As the data collectible things continue increase, anomaly detection has become an increasingly popular mining task. Traditional trajectory outlier algorithms can detect common anomalies effectively, but it is hard generalized anomalies, such as viewable direction gravity magnetic field which be collected by accelerometer,...

This paper presents a novel approach on bootstrapping chord and other ring-based peer-to-peer (P2P) systems over mobile ad hoc networks (MANETs). Only inter-neighbor communication is used to build the ring topology in node ID space of structured P2P systems. Upon this entire could be put into normal operation without lengthy stabilization. RAN protocol suite proposed. It includes three patterns: distributed exhaustive, virtual centralized random. Two exhaustive patterns adapt well...

In this article, we propose a novel method that can measure the similarity of FoV-tagged videos in two dimensions. Recently many researchers have focused on measuring videos. The measurement plays an important role various societal applications, including urban road networks, traffic, and geographic information systems. Our preliminary work introduced Largest Common View Subsequences (LCVS) algorithm for computing However, LCVS requires high computational cost calculating common viewable...

The ultrashort-term wind power prediction (USTWPP) technology assists the grid to arrange spare capacity, which is important optimize investment reasonably. To improve accuracy of USTWPP and requirements, a method with dynamic switching multiple models proposed. For high speed fluctuation samples, speed-power curve (WSPC) fitted in large sample historical data, corrected input WSPC. spatiotemporal attentive network model (STAN) built for low samples. According real-time characteristics...

The integration of a high proportion wind power has brought disorderly impacts on the stability system. Accurate forecasting technology is foundation for achieving dispatchability. To improve system after integration, this paper proposes steady-state deduction method based large-scale cluster forecasting. First, reorganization an improved DBSCAN algorithm designed to fully use spatial correlation resources in small-scale groups. Second, extract temporal evolution characteristics data,...

Based on the nonequilibrium plasma dynamics of air discharge, a dynamic model zero-dimensional is established by combining component density equation, Boltzmann and energy transfer equation. The evolution properties nanosecond pulse discharge (NPD) under different pressures are calculated. results show that pressure has significant impacts NPD products peak values particle number for particles such as O atoms, O3 molecules, N2(A3) molecules in excited states, NO molecules. It increases at...

This Letter presents a novel problem for discovering similar trajectories based on the field of view video data. The is important many societal applications such as grouping moving objects, classifying geo‐images, and identifying interesting trajectory patterns. Prior works consider only either spatial locations or relationship between two line‐segments. However, these approaches show limitation to find objects with common views. In this Letter, authors propose new algorithm that can group...