针对现有卫星观测任务可调度性预测模型难以建模长时间间隔的观测任务依赖关系的问题,提出一种基于双向门控循环单元(Bidirectional Gated Recurrent Unit,Bi-GRU)的卫星对地观测任务可调度性预测模型.该模型以卫星历史规划方案作为学习样本,能够以较低计算代价较高准确率地预测出对地观测任务集合中可以被响应的子集.该模型首先通过多层全连接感知机神经网络提取任务属性间的关联关系,然后采用多组多层双向门控循环单元组成的循环神经网络提取观测任务与其前驱及后继观测任务序列的潜在时序特征,最后融合各组双向门控循环单元的预测结果,从而利用观测任务之间的正向与反向信息依赖关系提升任务可调度性预测准确度.实验结果表明,与现有主流预测模型相比,本文提出方法在准确率、精确率、召回率和F1分数等指标上分别提升了2.27％、2.36％、3.45％和2.37％.

The accurate estimation of urban link travel time plays a significant role in urban traffic monitoring and su-pervision. Using taxicab GPS trip data, which contains origin and destination locations, travel time, and distances, this paper establishes a model to estimate average short-term urban link travel times. Firstly, the urban road network is di-vided into many segments based on crossings, and the running route of the driver was analyzed using the k-shortest path search algorithm. Then, for each road segment, a polynomial incidence relation model of the travel time in double lanes is proposed; this increases precision and avoids the overfitting of the travel time of the road network caused by insuffi-cient training data. Finally, by minimizing the mean square error between the expected path travel time and the ob-served path travel time as the optimization objective, the travel time of the road network is fitted. The results of experi-ments conducted on New York taxi datasets show that, relative to the traditional single-lane estimation method, the pro-posed model and method more efficiently estimate the travel time of the road segments in urban road networks.

The map tiles technology are widely used in WebGIS solutions of the current mainstream GIS software and the Internet map applications, tile processing service is the key technology to realize the fast and seamless browsing on WebGIS. Traditional algorithms and commercial GIS software cannot meet the demand of fast processing large data size raster image tiling. In view of these shortcomings and problems, this paper summarizes the current parallel tiling technology and then puts forward a new parallel raster image tile method which named"ParaTile". ParaTile adopts the shared disk parallel technology, using multi-process to divide the source raster image. Each process of ParaTile is read and written independently to the area that it is owned based on MPIparallel IO function. Finally, each process saves the output tile results independently according to the TMS(Tile Map Service)or the standard made by Google Tile. Experiments are tested by different level size of the remote sensing image, the results show that no matter in speed or stability ParaTile algo-rithm has a great improvement in the same parallel level compared to ArcGIS software, especially the larger size of data the more advantages are.

现有的数字化地质图,多以MapGIS(wl/wt/wp)数据格式存储,无法为相关单位、个人提供较好的数据服务,给业内数据共享带来了障碍;同时,传统GIS软件也无法满足复杂地质图高效绘制的需求.基于此,提出了一种基于地图制图脚本的地质图发布方法,通过对原始电子地质图信息的解析,自动生成制图脚本,实现高效的Web端地质图成图.

随着人类对地观测卫星系统的发展,卫星遥感数据量爆炸性增长.海量的遥感数据对当今存储系统存储容量、数据可用性以及I/O性能等方面提出了越来越高的要求.针对需要持久化存储与高速交换的数据,该文设计并实现了一种基于共享文件系统和文件编目消息推送的海量卫星数据实时存储与交换方法.实验测试表明,该方法针对海量卫星遥感数据的实时存储与交换取得了很好的效果.

随着移动互联网技术地不断发展,网民针对各种话题发出的博文、评论呈爆炸式增长.该文针对各大网站产生的海量新闻、微博文及其评论信息,将网络中海量的新闻、微博文及其评论转变为文本信息进行分析,识别出其中的消极性信息,对于现代商业企业、政府舆情监控具有较高应用价值.

Since linear transformation is used to obtain baseline signal in intrinsic time scale decomposition (ITD) method, burr and instantaneous frequency distortion will appear in the decomposition results. Therefore, a rational Cubic Hermite interpolation—Local characteristic-scale decomposition (CHLCD) method is presented. In this method, any complex signal can be adaptively decomposed into a sum of several independent rational intrinsic scale components (ISCs), whose instantaneous frequencies have obvious physical meanings. Firstly, the principle of the CHLCD method was analyzed. Then, the detailed steps of CHLCD of signal were given. Finally, a simulation signal was adopted to verify the CHLCD method. Experimental results show that the CHLCD method can effectively decompose signals.

存储海量的小文件给文件系统带来容量和效率上的压力,多种传统处理小文件方法的工作机制都存在各自的缺陷.对于实时存储的文件系统,一种数据库与文件系统相结合的针对海量小文件实时存储的方案能较好地缓解小文件带来的压力.该方案先对小文件进行文件合并用数据库存储小文件的元数据,与单纯使用数据库的方法进行实验对比后验证了这种方案的可行性与有效性.

Existing spatiotemporal database supports spatiotemporal aggregation query over massive moving objects datasets. Due to the large amounts of data and single-thread processing method, the query speed cannot meet the application requirements. On the other hand, the query efficiency is more sensitive to spatial variation then temporal variation. In this paper, we proposed a spatiotemporal aggregation query method using multi-thread parallel technique based on regional divison and implemented it on the server. Concretely, we divided the spatiotemporal domain into several spatiotemporal cubes, computed spatiotemporal aggregation on all cubes using the technique of multi-thread parallel processing, and then integrated the query results. By testing and analyzing on the real datasets, this method has improved the query speed significantly.

对地观测具有任务繁多、约束复杂、资源有限等特点,对观测资源配置进行合理规划具有重要意义.面向多飞艇多载荷对地观测任务,结合经典粒子群算法,研究了满足任务完成度、分辨率要求和负载均衡度等多目标的资源配置规划方法.在经典粒子群算法的基础上利用混沌优化技术进行改进,并设计实现了原型系统.实验验证了该资源配置规划方法的有效性.

According to data transmission scheduling problem of an optical Earth Observation Satellite(EOS) with long time window,a new method is presented.Firstly,the constraints and characteristics of EOS are analyzed.Then,a mathematical model that considers multiform data transmission mode coexistence utilizing the feature of a long time window is proposed.Meanwhile,this paper presents a scheduling algorithm for EOS based on the maximal ratio of gain and loss.Finally,extensive experiments are implemented which verify the correctness and practicability of our approach.

To solve the problem of real time visualization of 3D city model based on CityGML,a rapid data parsing strategy which combined the streamed reading and data binding techniques was proposed together with a multi-threaded parallel data loading strategy.Moreover,efficient rendering was achieved through reorganizing the model into Direct3D Mesh.The experimental results implemented by using C# and Direct3D under the.Net platform show the effects of the 3D visualization which verify the effectiveness of the proposed methods.

针对传统的集中式卫星规划方法所存在的不足，建立了基于MAS(Mulit-Agent System)的分布式规划模型，提出了基于改进的合同网协议的分布式卫星任务规划算法。最后经过仿真实验与分析，验证了该算法的可用性和有效性。

A new method is developed based on ArcGIS graphic element to improve the capabilities of decision support for military plotting system.This method using control points generate unformulated military symbol,using ArcGIS plotting function instead of graphic element draw military symbol dynamically,using attribute identification of graphic element integrate military and attribute data,design and realize a military plotting system based on this solution,and successfully apply on practical project.The results show that this system not only is convenient to plot,but also has decision support capabilities for commanders to provide a powerful means of scientific decision-making.

Considering the applied problems in sharing distributed heterogeneous spatial data,this paper proposes a retrieval method for distributed spatial metadata based on directory services.A framework of distributed spatial metadata directory services is established.Moreover,a hierarchical directory structure based on spatial locations is built to provide logical association and organization foundation for distributed spatial metadata.Then a multi-directory-servers-oriented mode is presented to support directory services implementation at the physical level.Thus,distributed spatial metadata can be retrieved effectively.

Research on multi-satellite data transmission scheduling is of great importance to boosting the application benefits of the earth observation system.A constraint optimization model considering time window adjustment is proposed by analyzing the multi-satellite data transmission problem.Moreover,based on Squeaky-Wheel Optimization(SWO) making long leaps in search space,a scheduling algorithm is presented.This algorithm can quickly obtain the optimal solution.Finally,extensive experiments are implemented to verify correctness and practicability of our algorithm.

Research on the affect of geography factors on the satellite imaging mainly focuses on pre-evaluating the influence to provide auxiliary decision-making information for satellite schedule.Taking illumination,cloud coverage and terrain into account,an influence pre-evaluation model of geography factors on visible light sensor imaging is presented to calculate the solar radiation received by spaceborne visible light sensor and considering the terrain feature.An influence pre-evaluation model of geography factors on SAR imaging is also presented to calculate the geometric distortion.Taking advantage of ArcGIS,both models are conceived and influence pre-evaluating methods are introduced based on the corresponding models.Then an application based on the ArcGIS software is developed.

如何既准确获取监测数据,又减少监测过程对实际网络传输数据的影响是网络研究的一个根本问题.该文提出了层次化网络测量模型,强调数据聚集过程的传输延迟和监测数据流量占用链路的带宽都小于预设数值同时,尽量追求聚集节点数目最小化.求解该模型的问题是NP难的,在节点间路由唯一的情况下,该文设计的近似算法的近似程度为,其中是监测节点的数量.

Aiming at the problem of pulse Doppler radar detection of aircraft wake vortices,we investigate the RCS(Radar Cross Section) and Doppler spectrum of aircraft wake vortices in clear air and rain,and analytically derive the radar equations and their detectability factors,On this basis,the performance of aircraft wake vortex detection with X,C,S and L-band radars is studied along with the detection performance of the conventional radar for stealth aircrafts.The results show that,for the aircraft wake vortices with a scattering rate between-90 dBsm/m and-70 dBsm/m,the detection range may range from 10 km to more than 100 km.When the radar reflectivity factor is 10~40 dBZ,the detection performance of wake vortices in rain is generally better than that of the conventional radar detection,and the detection range can be two times more than the latter's one.

Polarimetric detection models are established for mono-pulse detection of range-extended non-fluctuation targets and Rayleigh fluctuation targets.The corresponding polarimetric detectors are derived,and their detection performance is analytically expressed.Then the relationships between the detection performance and the radar bandwidth,polarization of the target's echoes,and estimation error of the target's radial scale are theoretically analyzed,and the detection algorithms proposed in this paper are compared with the polarimetric detection algorithm using binary integration.The results show that,an optimal bandwidth maximizing the detection probability of a Rayleigh target can be found for a given signal-to-noise ratio.The detection performance difference of deterministically-polarized targets and randomly-polarized targets is remarkable when the radar bandwidth is narrow.When the radar bandwidth is wide,the detection of a randomly-polarized target has a performance loss of about 1.3dB in contrast with the detection of a deterministically-polarized target.For a Rayleigh target,the detection performance is more sensitive to the underestimate than to the overestimate of the target's radial scale.The detection performance of binary integration is generally inferior to the detection performance of radial integration,but the performance difference between them is relatively small.