盐田水体遥感分类方法研究

以连云港台北盐场为研究区，介绍了监督分类法和神经网络分类法及其在盐田水体遥感分类中的具体应用。研究结果表明，用神经网络分类法进行遥感影像自动分类，其分类精度高，显示了其在遥感领域较为广阔的应用前景。     

粗糙集理论进行遥感图像监督分类的样本质量评价的研究(英文)

In the supervised classification process of remotely sensed imagery, the quantity of samples is one of the important factors affecting the accuracy of the image classification as well as the keys used to evaluate the image classification. In general, the samples are acquired on the basis of prior knowledge, experience and higher resolution images. With the same size of samples and the same sampling model, several sets of training sample data can be obtained. In such sets, which set reflects perfect spectral characteristics and ensure the accuracy of the classification can be known only after the accuracy of the classification has been assessed. So, before classification, it would be a meaningful research to measure and assess the quality of samples for guiding and optimizing the consequent classification process. Then, based on the rough set, a new measuring index for the sample quality is proposed. The experiment data is the Landsat TM imagery of the Chinese Yellow River Delta on August 8th, 1999. The experiment compares the Bhattacharrya distance matrices and purity index zl and △x based on rough set theory of 5 sample data and also analyzes its effect on sample quality.     

Grid pattern recognition in road networks using the C4.5 algorithm

Pattern recognition in road networks can be used for different applications, including spatiotemporal data mining, automated map generalization, data matching of different levels of detail, and other important research topics. Grid patterns are a common pattern type. This paper proposes and implements a method for grid pattern recognition based on the idea of mesh classification through a supervised learning process. To train the classifier, training datasets are selected from worldwide city samples with different cultural, historical, and geographical environments. Meshes are subsequently labeled as composing or noncomposing grids by participants in an experiment, and the mesh measures are defined while accounting for the mesh’s individual characteristics and spatial context. The classifier is generated using the C4.5 algorithm. The accuracy of the classifier is evaluated using Kappa statistics and the overall rate of correctness. The average Kappa value is approximately 0.74, which corresponds to a total accuracy of 87.5%. Additionally, the rationality of the classifier is evaluated in an interpretation step. Two other existing grid pattern recognition methods were also tested on the datasets, and comparison results indicate that our approach is effective in identifying grid patterns in road networks.     

Geostatistical incorporation of spatial coordinates into supervised classification of hyperspectral data

 This paper presents a methodology to incorporate both hyperspectral properties and spatial coordinates of pixels in maximum likelihood classification. Indicator kriging of ground data is used to estimate, for each pixel, the prior probabilities of occurrence of classes which are then combined with spectral-based probabilities within a Bayesian framework. In the case study (mapping of in-stream habitats), accounting for spatial coordinates increases the overall producer's accuracy from 85.8% to 93.8%, while the Kappa statistic rises from 0.74 to 0.88. Best results are obtained using only indicator kriging-based probabilities, with a stunning overall accuracy of 97.2%. Significant improvements are observed for environmentally important units, such as pools (Kappa: 0.17 to 0.74) and eddy drop zones (Kappa: 0.65 to 0.87). The lack of benefit of using hyperspectral information in the present study can be explained by the dense network of ground observations and the high spatial continuity of field classification which might be spurious. Received: 12 April 2001 / Accepted: 7 September 2001     

基于遥感技术的城市扩展变化研究

以1975年的MSS影像、1992年的TM影像和2004年的中巴资源卫星影像为数据源,对北京市建成区近30 a的扩展变化进行动态监测,运用计算机监督分类技术提取了北京市建成区30 a来的扩展变化信息,并分析出北京市建成区的变化是以老市区为中心向四周辐射蔓延的规律,为北京市未来城市规划与开发提供参考。     

土地利用/土地覆被变化遥感监测研究

根据1990年和1999年两期TM/ETM 遥感数据,运用遥感分类中的监督分类方法,全面分析乌鲁木齐市的土地利用的数量变化特征和空间变化特征,阐明了区域土地利用变化的区域特点。     

高分辨率遥感影像解译中的机器学习范式

高分辨率遥感影像解译是遥感信息处理领域的研究热点之一,在遥感大数据知识挖掘与智能化分析中起着至关重要的作用,具有重要的民用和军事应用价值。传统的高分辨率遥感影像解译通常采用人工目视解译方式,费时费力且精度低。所以,如何自动、高效地实现高分辨率遥感影像解译是亟待解决的问题。近年来,随着人工智能技术的飞速发展,采用机器学习方法实现高分辨率遥感影像解译已成为主流的研究方向。本文结合高分辨率遥感影像解译的典型任务,如目标检测、场景分类、语义分割、高光谱图像分类等,系统综述了5种代表性的机器学习范式。具体来说,本文分别介绍了不同机器学习范式的定义、常用方法以及代表性应用,包括全监督学习(如支持向量机、K-最近邻、决策树、随机森林、概率图模型)、半监督学习(如纯半监督学习、直推学习、主动学习)、弱监督学习(如多示例学习)、无监督学习(如聚类、主成分分析、稀疏表达)和深度学习(如堆栈自编码机、深度信念网络、卷积神经网络、生成对抗网络)。其次,深入分析五种机器学习范式的优缺点,并总结了它们在遥感影像解译中的典型应用。最后,展望了高分辨率遥感影像解译的机器学习发展方向,如小样本学习、无监督深度学习、强化学习等。     

基于卷积神经网络的ICESat-2光子点云去噪分类

ICESat-2（Ice, Cloud, and land Elevation Satellite-2）是美国NASA（National Aeronautics and Space Administration）在2018年发射的激光测高卫星,其上搭载的激光测高系统ATLAS（Advanced Topographic Laser Altimeter System）采用微脉冲多波束光子计数激光雷达系统,因其低能耗、高探测灵敏度、高重复频率的特性极大改善了沿轨采样密度,但也使获取的数据中包含大量的噪声,如何有效实现光子点云去噪分类成为后续应用的关键,也是当前研究的热点和难点,为此本文提出一种基于卷积神经网络的光子点云去噪和分类算法。首先将光子点云按照沿轨和高程方向划分格网,去除明显的噪声光子,并将每个粗信号光子点栅格化为影像;然后基于少量样本构建的卷积神经网络分类模型实现光子点云精去噪和分类;最后利用机载激光雷达数据进行验证,并与ATL08产品的去噪分类结果进行对比。结果表明,对于裸地和森林区域,卷积神经网络算法均能有效去除噪声光子,特别对于森林区域,可同时实现去噪和分类;其中,裸地区域地表计算的R²和RMSE分别为1.0和0.72 m,森林区域地表和树冠计算的R²分别为1.0和0.70, RMSE分别为1.11 m和4.99 m。本文利用深度学习算法实现光子点云去噪分类,在裸地和森林区域均取得了较好的结果,为后续光子点云数据处理提供了参考。     

Airborne near-real-time monitoring of assembly and parking areas in case of large-scale public events and natural disasters

A critical requirement for an effective and coordinated response by public entities tasked with management, security, and relief during large-scale public events or natural disasters is the availability of current situational information. However, today there is a lack of comprehensive operational systems allowing a near-real-time (NRT) collection, visualization, and provision of situational information for larger areas. In this study a methodological framework is proposed, which allows an NRT extraction and visualization of situational information based on aerial image acquisition. The framework combines digital image analysis using a generic supervised information extraction approach based on statistical modeling with a downstream web-based visualization component realized through an automatic update of web services. Even though being applicable for different scenarios, the workflow will be demonstrated for the specific use-case of a NRT monitoring of open spaces including assembly and parking areas. Compared to other approaches, image analysis results indicate a high robustness and a low demand for computational power sources (7 seconds per image). Due to a high degree of automation, the proposed workflow contributes to a NRT ‘end-to-end’ monitoring system, which was developed within the VABENE (German acronym for ‘traffic management under large-scale public events and disaster conditions’) project covering all parts from the acquisition of raw aerial imagery to the dissemination of information products to end-users.     

基于ENVI 的唐山湾三岛土地利用遥感分类方法的比较分析

对唐山湾三岛2010年10月10 m分辨率的SPOT5多光谱数据源,采用不同分类方法进行识别和判断,并对比不同分类器在遥感影像分类中的应用效果和分类精度。有针对性地探究海岛土地利用遥感分类过程中不同方法的优劣,获取最适于岛陆地区土地利用的遥感分类方法。根据土地利用现状分类标准(GB/T21010-2007)和海岛陆域土地利用类型划分的相关规定,将唐山湾三岛主要分为滩涂、裸地、林地、草地、居住区、内陆水体和潮间带(潮水覆盖区)6类。并分别构建结合人为控制的非监督分类、监督分类和基于专家知识的决策树分类系统,参照更高分辨率影像、先验知识和野外调查资料,评价分类结果与实地调查结果的吻合程度,最终通过总体分类精度和Kappa系数等指标对各分类器精度进行评价和对比分析。