Self-organised clustering for road extraction in classified imagery

The extraction of road networks from digital imagery is a fundamental image analysis operation. Common problems encountered in automated road extraction include high sensitivity to typical scene clutter in high-resolution imagery, and inefficiency to meaningfully exploit multispectral imagery (MSI). With a ground sample distance (GSD) of less than 2 m per pixel, roads can be broadly described as elongated regions. We propose an approach of elongated region-based analysis for 2D road extraction from high-resolution imagery, which is suitable for MSI, and is insensitive to conventional edge definition. A self-organising road map (SORM) algorithm is presented, inspired from a specialised variation of Kohonen's self-organising map (SOM) neural network algorithm. A spectrally classified high-resolution image is assumed to be the input for our analysis. Our approach proceeds by performing spatial cluster analysis as a mid-level processing technique. This allows us to improve tolerance to road clutter in high-resolution images, and to minimise the effect on road extraction of common classification errors. This approach is designed in consideration of the emerging trend towards high-resolution multispectral sensors. Preliminary results demonstrate robust road extraction ability due to the non-local approach, when presented with noisy input.     

Thematic mapping from remotely sensed data with neural networks: MLP, RBF and PNN based approaches

 Neural networks are attractive tools for the derivation of thematic maps from remotely sensed data. Most attention has focused on the multilayer perceptron (MLP) network but other network types are available and have different properties that may sometimes be more appropriate for some applications. Here a MLP, radial basis function (RBF) and probabilistic neural network (PNN) were used to classify remotely sensed data of an agricultural site. The accuracy of these classifications ranged from 86.25–91.25%. The accuracy of the PNN classification could be increased through the incorporation of prior probabilities of class membership but the accuracy of each classification could also be degraded by the presence of an untrained class. Post-classification analyses, however, could be used to identify potentially misclassified cases, including those belonging to an untrained class, to increase accuracy. The effect of the post-classification analysis on the accuracy of the classification derived from each of the three network types investigated differed and it is suggested that network type be selected carefully to meet the requirements of the application in-hand. Received: 23 March 2000 / Accepted: 9 July 2000     

The development of computer networks: First results from a microeconomic model

 Computer networks like the Internet are gaining importance in social and economic life. The accelerating pace of the adoption of network technologies for business purposes is a rather recent phenomenon. Many applications are still in the early, sometimes even experimental, phase. Nevertheless, it seems to be certain that networks will change the socioeconomic structures we know today. This is the background for our special interest in the development of networks, in the role of spatial factors influencing the formation of networks, and consequences of networks on spatial structures, and in the role of externalities. This paper discusses a simple economic model – based on a microeconomic calculus – that incorporates the main factors that generate the growth of computer networks. The paper provides analytic results about the generation of computer networks. The paper discusses (1) under what conditions economic factors will initiate the process of network formation, (2) the relationship between individual and social evaluation, and (3) the efficiency of a network that is generated based on economic mechanisms. Received: 5 July 2000 / Accepted: 28 November 2000     

Forecasting Monsoon Precipitation Using Artificial Neural Networks

This paper explores the application of Artificial Intelligent (AI) techniques for climate forecast. It pres ents a study on modelling the monsoon precipitation forecast by means of Artificial Neural Networks (ANNs). Using the historical data of the total amount of summer rainfall over the Delta Area of Yangtze River in China, three ANNs models have been developed to forecast the monsoon precipitation in the corre sponding area one year, five-year, and ten-year forward respectively. Performances of the models have been validated using a ＇new＇ data set that has not been exposed to the models during the processes of model development and test. The experiment results are promising, indicating that the proposed ANNs models have good quality in terms of the accuracy, stability and generalisation ability.     

前馈神经网络隐性单调与泛化能力

应用前馈人工神经网络对广域单调的两组样本进行了模拟反演，引入单调前馈网络的概念对其权值和阈值定解问题和泛化能力进行了较诉研究。表明前馈人工神经网络是一个表达形式简单的复杂系统，其单调特征是隐性的，而且训练网络的成熟性对样本数量和样本内在规律性有一定依赖。强调了前馈人工神经网络的应用效果，指出单调与复合问题还需进一步深入研究。     

River flow mass exponents with fractal channel networks and rainfall

An important problem in hydrologic science is understanding how river flow is influenced by rainfall properties and drainage basin characteristics. In this paper we consider one approach, the use of mass exponents, in examining the relation of river flow to rainfall and the channel network, which provides the primary conduit for transport of water to the outlet in a large basin. Mass exponents, which characterize the power-law behavior of moments as a function of scale, are ideally suited for defining scaling behavior of processes that exhibit a high degree of variability or intermittency. The main result in this paper is an expression relating the mass exponent of flow resulting from an instantaneous burst of rainfall to the mass exponents of spatial rainfall and that of the network width function. Spatial rainfall is modeled as a random multiplicative cascade and the channel network as a recursive replacement tree; these fractal models reproduce certain types of self-similar behavior seen in actual rainfall and networks. It is shown that under these modeling assumptions the scaling behavior of flow mirrors that of rainfall if rainfall is highly variable in space, and on the other hand flow mirrors the structure of the network if rainfall is not so highly variable.     

监测网优化设计的必要性及内容

本文详细分析了监测网优化设计的必要性，并针对监测网的特点，叙述了监测网设计的有关问题。     

用人工神经网络实现地震记录中的废道自动切除

切除废道（即道编辑）的工作是地震信号处理中重要而又工作量巨大的一步.本文提出了一种用人工神经网络实现的自动切除方法,所用的神经网络模型为新奇滤波器模型,方法简单实用,切除效果令人满意.文中介绍了方法的原理并给出了几个实验结果.     

监测网的可区分测度

监测网的设计阶段，变形体的变形特性不是完全清楚的情况下，须考虑监测网的可区分测度．本文详细讨论了监测网的可区分问题，导出了有关的公式，并以算例加以分析和说明，最后论述了监测网设计的可区分准则．     

黄土孔隙性分类判别的人工神经网络方法

运用人工神经网络的一典型模型——“反向传播”神经网络,对洛川黄土孔隙性的实测数据进行了分析,建立了洛川黄土孔隙性预测的计算机智能专家系统。结果表明,神经网络方法性能良好,可望成为黄土孔隙性分类、判别的有效辅助手段。