图卷积网络在道路网选取中的应用     

张康  郑静  沈婕  马劲松 《测绘科学》2021,46(2):165-170,177

针对现有制图综合中的道路网自动选取方法不能有效地利用道路网的空间特征问题,该文把道路网抽象为图结构,提出了使用图卷积网络来进行道路网的自动选取,并比较分析了不同的图卷积网络在道路网选取中的适用性。结果表明,图卷积网络可以通过多层卷积来自动提取不同局部范围的空间特征,从而减少空间特征的人工构建,相比传统的多层感知机(MLP)等人工智能选取方法,具有更高的选取精度。对于不同的图卷积网络模型,使用最大池化聚合的GraphSAGE获得了最优的性能。  相似文献   

基于“深度学习”识别模型的玉米农田监测应用系统设计与实现     

阎妍  行鸿彦  刘刚  吴红军  吴慧  戴学飞  余培 《气象科技》2019,47(4):571-580

为了精准判断玉米所处生长阶段,远程实时监测玉米长势,分析生长阶段与田间环境要素间的关系,本文提出深度局部关联神经网络,克服了玉米生长阶段识别中存在的多模态和模糊性问题,在Oxford VGGNet(Visual Geometry Group Net)模型中添加一个新的监督层,即局部关联损失层,提高深层特征的判别能力。基于所提的玉米生长阶段图片识别新算法,拓展环境要素监测功能,设计一套基于深度学习的玉米农田监测系统。系统由玉米农田监测装置和云端服务器组成,监测装置采集玉米图像、气象要素和田间位置数据,通过4G无线发送给云端服务器,云端服务器利用深度局部关联神经网络识别生长阶段,显示结果并存入数据库中。仿真试验表明,深度局部关联神经网络平均识别准确率达到92.53%,较VGGNet的87.21%和LSTM的88.50%,准确率分别提高了5.32%和4.03%。实地测试结果表明,野外环境下系统准确率可达到91.43%,能够稳定地对农田玉米生长情况进行监测,具有重要的应用价值。  相似文献   

A novel residual graph convolution deep learning model for short-term network-based traffic forecasting   总被引：1，自引：0，他引：1  

Yang Zhang  Tao Cheng  Yibin Ren  Kun Xie 《International journal of geographical information science》2020,34(5):969-995

ABSTRACT

Short-term traffic forecasting on large street networks is significant in transportation and urban management, such as real-time route guidance and congestion alleviation. Nevertheless, it is very challenging to obtain high prediction accuracy with reasonable computational cost due to the complex spatial dependency on the traffic network and the time-varying traffic patterns. To address these issues, this paper develops a residual graph convolution long short-term memory (RGC-LSTM) model for spatial-temporal data forecasting considering the network topology. This model integrates a new graph convolution operator for spatial modelling on networks and a residual LSTM structure for temporal modelling considering multiple periodicities. The proposed model has few parameters, low computational complexity, and a fast convergence rate. The framework is evaluated on both the 10-min traffic speed data from Shanghai, China and the 5-min Caltrans Performance Measurement System (PeMS) traffic flow data. Experiments show the advantages of the proposed approach over various state-of-the-art baselines, as well as consistent performance across different datasets.  相似文献   

Simple procedure to simulate karstic aquifers     

Alberto Padilla  Antonio Pulido‐Bosch 《水文研究》2008,22(12):1876-1884

A procedure to simulate karstic aquifers is presented. It is based on a simulation of spring discharge using precipitation and, where necessary, temperature as input data. The karstic aquifer system is considered to be divided into three zones: the surface zone, the unsaturated zone (UZ) and the saturated zone (SZ). Each of these is described by a transfer function that determines the water supplied from the overlying zone. Water loss through evapotranspiration is calculated empirically and subtracted from the total precipitation in order to obtain the effective infiltration into the UZ. The transfer function characterizing the UZ can be expressed as a convolution function. The UZ acts as a buffer, delaying effective infiltration into the SZ. Water discharge from the SZ is described by the recession function of the spring, and this becomes the transfer function that characterizes the emergence of water from the SZ. The model permits the simulation of the influence of pumped abstractions from the system by a simple modification of the transfer functions involved. Copyright © 2007 John Wiley & Sons, Ltd.  相似文献   

基于全卷积网络的高分辨遥感影像目标检测   总被引：2，自引：0，他引：2  

徐逸之  姚晓婧  李祥  周楠  胡媛 《测绘通报》2018,(1):77-82

目标检测是遥感图像分析处理中的研究热点之一,具有十分重要的科研和应用价值。传统遥感影像目标检测方法多使用人工构造的浅层次特征,结合支持向量机、随机森林、Adaboost等分类器进行目标识别,难以充分挖掘和利用影像中的深层特征。近年来,深度学习,特别是卷积神经网络在图像认知方面取得了巨大成功。在目标检测领域,以Faster R-CNN算法为代表的方法取得了突破性进展,检测精度大幅提高,检测速度达到了近实时的性能。但是,Faster R-CNN算法由于使用了感兴趣区域（RoI）池化层,各个RoI计算不共享,因此检测速度依然有待提高。R-FCN基于全卷积网络结构,同时采用位置敏感池化来引入平移变化,抵消全卷积网络造成的平移不变形问题,检测精度和效率都有了很大的提高。本文阐述了R-FCN算法原理,并运用于高分辨遥感影像目标检测分析了不同参数和网络结构对R-FCN检测效果的影响,比较了利用Fast R-CNN、Faster R-CNN和R-FCN 3种算法进行飞机识别的性能。试验结果表明,利用R-FCN进行飞机识别定位可以达到99.3%的准确率和每张图180 ms的检测速度。  相似文献   

基于改进型U-Net网络的高分辨率遥感影像建筑物提取     

吕道双  林娜  冯丽蓉  张小青 《地理空间信息》2021,19(1):30-34

针对传统人工提取方法自动化程度低、过分依赖人工设计的特征,以及现有的深度学习方法中存在的提取精度不高等问题,提出了一种基于改进型U-Net网络的高分辨率遥感影像建筑物提取方法。首先将空洞卷积加入到网络中,利用不同尺度的空洞卷积对来自网络编码部分的结果进行多尺度特征提取;再对提取的特征进行特征融合,并输入到网络的下一层;然后将制作的数据集输入到网络中进行训练;最后利用Softmax得到最终分割结果。在建筑物公开的数据集中进行测试,提取结果的像素精度为96.26%;Iou精度为78.59%、Recall为95.65%,表明该方法具有良好的鲁棒性和精度,能从影像中准确地提取建筑物。  相似文献   

Statistical analysis of friction sleeve length effects on soil classification     

D. R. Saussus  J. D. Frost  J. T. DeJong 《国际地质力学数值与分析法杂志》2004,28(12):1257-1278

The cone penetration test (CPT) provides profiles of the tip resistance, sleeve friction, and pore water pressure encountered while penetrating the subsurface. These parameters are used either directly or indirectly to classify the soil types present and to obtain geotechnical design parameters. However, fundamental discrepancies exist in the manner by which these parameters are measured. This paper describes the results of a study that shows the sleeve friction measurement introduces unnecessary redundancy due to the length of the standard friction sleeve compared to the measurement increment. Further, the high sleeve length to measurement increment ratio results in filtering and smoothing of the friction data, thereby causing the variability of the friction between the soil and the cone sleeve to be underestimated. The importance of understanding the role of the sleeve length on measurements is demonstrated using synthetically generated friction profiles and estimating the profiles that would be measured using sleeves of different lengths. Differences in how the soils are classified as a function of the sleeve length used to obtain each profile are illustrated. Solutions are presented to validate the synthetic sleeve friction profiles, to demonstrate the filtering and smoothing effects of the friction sleeve on the data, and to explain the implications of the sleeve length on soil classification. Copyright © 2004 John Wiley & Sons, Ltd.  相似文献   

BP神经网络图象边缘检测器及其内部信息     

冯天瑾 《中国海洋大学学报(自然科学版)》1994,(Z2)

介绍用神经网络实现图象边缘检测的实验研究结果。一幅数字图象用一个规模不大的ＢＰ网络边缘检测器处理小邻区，再用扫描此图象的方法进行边缘检测。此方法之最大优点是设计简单，网络边缘检测器的性能也令人满意．欠缺是：学习时间长，神经网络隐层中权重包含的信息难于解释清楚。  相似文献   

Dynamic Response Analysis of Risers Subjected to Combined Wave and Current Loading     

Nie Wu  Yu Xiukun Ma Chunyan Li Weiyang Associate Professor  Harbin Shipbuilding Engineering Institute  Harbin Research Staff  Harbin Shipbuilding Engineering Institute  Harbin Professor  Harbin Shipbuilding Engineering Institute  Harbin 《中国海洋工程》1993,(2)

A dynamic response analysis in the frequency domain is presented for risers subjected to combined wave and current loading. Considering the effects of current, a modified wave spectrum is adopted to compute the linearized drag force. An additional drag force convolution term is added to the linearized drag force spectrum, therefore the error is reduced which arises from the truncation of higher order terms in the drag force auto-correlation function. An expression of linearized drag force spectrum is given taking the relative velocity into account. It is found that the additional term is a fold convolution integral. In this paper dynamic responses of risers are investigated, while the influence of floater motion on risers is considered. The results demonstrate that the accuracy of the present method reaches the degree required in time domain analysis.  相似文献   

频率空间域显式偏移     

熊金良  李国发  张立勤 《物探化探计算技术》2004,26(2):109-113

频率空间域显式偏移利用有限长度的褶积算子，通过对不同的频率进行空间褶积来实现偏移成像，且具有简洁、高效、便于扩展等特点。褶积算子一般通过泰勒展开或非线性优化来求取。针对以往方法进行分析的基础上，提出了一种新的褶积延拓算子计算方法，通过对模型资料和实际资料的处理，表明该方法是可行的。  相似文献