专用方位控制网形变检测方案的设计

为检测专用方位控制网是否发生了明显变形 ,本文作者提出了方位边夹角灵敏度的概念 ,并推导了相应的公式。     

Non-destructive estimation of foliar chlorophyll and carotenoid contents: Focus on informative spectral bands

Leaf pigment content provides valuable insight into the productivity, physiological and phenological status of vegetation. Measurement of spectral reflectance offers a fast, nondestructive method for pigment estimation. A number of methods were used previously for estimation of leaf pigment content, however, spectral bands employed varied widely among the models and data used. Our objective was to find informative spectral bands in three types of models, vegetation indices (VI), neural network (NN) and partial least squares (PLS) regression, for estimating leaf chlorophyll (Chl) and carotenoids (Car) contents of three unrelated tree species and to assess the accuracy of the models using a minimal number of bands. The bands selected by PLS, NN and VIs were in close agreement and did not depend on the data used. The results of the uninformative variable elimination PLS approach, where the reliability parameter was used as an indicator of the information contained in the spectral bands, confirmed the bands selected by the VIs, NN, and PLS models. All three types of models were able to accurately estimate Chl content with coefficient of variation below 12% for all three species with VI showing the best performance. NN and PLS using reflectance in four spectral bands were able to estimate accurately Car content with coefficient of variation below 14%. The quantitative framework presented here offers a new way of estimating foliar pigment content not requiring model re-parameterization for different species. The approach was tested using the spectral bands of the future Sentinel-2 satellite and the results of these simulations showed that accurate pigment estimation from satellite would be possible.     

一种稀疏降噪自编码神经网络影像变化检测方法

采用粤西2018—2019年优于0.5 m地理国情监测影像,结合2018年矢量化地表覆盖分类成果,使用稀疏降噪自编码神经网络深度学习方法,首先选取BJ、GF、ZY等各类高分辨率遥感影像训练生成多传感器训练模型;其次利用PCA主成分分析提取样本数据最大特征,实现样本数据白化降维;最后采用tanh函数作为神经元激活函数,选...     

玉米大豆生长中后期遥感辨识的指示性特征研究

玉米和大豆是两种主要的粮食作物,及时准确地监测两者的种植面积对于产量预测和市场价格的制定具有重要的意义。利用遥感技术探究在生长季中后期能有效区分玉米和大豆的指示性特征集,为在不同实验区进行推广应用和提前玉米和大豆种植面积信息发布的时间提供技术支撑。文章以玉米和大豆为研究对象,以黑龙江和安徽省两个典型种植区为实验区,以高分一号影像为数据源,计算多种植被指数特征和两种纹理特征,同时利用特征优选方法评价特征间的相对重要性,并结合随机森林分类算法分析特征个数对精度的影响,得到不同试验区区分两者的最佳特征子集。随后根据不同实验区最佳特征子集的共同点和差异,遴选出对玉米和大豆中后期区分的遥感指示性识别特征集,并设计实验方案验证其有效性和稳定性。实验表明：在玉米和大豆生长中后期存在具有高效辨识两者的遥感特征集,能有效和稳定地增强两者的遥感识别能力;在不同实验区,基于特征优选方法可以选择出区分玉米和大豆的最佳分类特征子集,得到两者最优的识别效果,比仅仅使用原始波段特征的分类精度提升了近10个百分点,总体分类精度能够平均达到97%,Kappa系数0.96,玉米和大豆的单类分类精度平均超过95%;在不同的种植区,利用玉米和大豆的指示性特征集可以得到几乎与优选出的最佳特征子集同样的分类精度和制图效果,且具有稳定性和有效性,较最佳特征集更具推广使用意义。指示性特征集包含6种：植被指数中的比值植被指数（RVI）,差值植被指数（DVI）,转换型植被指数（TVI）,改进型叶绿素吸收比率指数（MCARI）和灰度共生矩阵（GLCM）纹理特征中的二阶矩（the Second Moment）和熵（Entropy）。     

基于改进麻雀搜索算法优化BP神经网络的PM_(2.5)浓度预测北大核心CSCD

针对传统BP神经网络模型收敛速度慢、易陷入局部极值等问题,本文采用分段线性混沌映射(PWLCM)和萤火虫算法(FA)改进麻雀搜索算法(SSA),并优化BP神经网络模型初始权值和阈值,对西安市PM_(2.5)浓度进行预测。通过比较不同模型预测结果的评价指标,并与性能较优的SSA-BP模型对比,ISSA-BP模型预测结果的RMSE、MAPE、MAE分别下降了3.70、3.73、3.34。试验结果表明,改进后的麻雀搜索算法具有高效的全局最优搜索能力,优化后的ISSA-BP神经网络预测稳定性高,精度优于BP、SSA-BP神经网络模型,可用于预测PM_(2.5)浓度。     

基于卷积神经网络的地铁保护区风险源识别北大核心CSCD

城市轨道交通沿线的风险源识别是防止违规作业导致的安全事故的重要手段。为解决传统识别方法效率低、漏检率高、成本大等问题,本文基于无人机地铁保护区巡检系统,采用卷积神经网络对无人机采集的影像数据进行风险源识别。首先介绍无人机影像获取的流程,并在原始影像数据的基础上,通过数据增强的方式制作多角度、多尺度的风险源数据集;然后使用卷积神经网络建立风险源识别模型,对无人机采集影像中的风险源进行自动识别和定位。试验结果表明,多角度、多尺度风险源数据集的建立进一步提升了模型的识别准确率,且比传统方法具有效率高、成本低等优点。     

Semi-automatic delineation of badlands using contrast in vegetation activity: a case study in the lower Chambal valley,India

Population growth worldwide leads to an increasing pressure on the land. Recent studies reported that many areas covered by badlands are decreasing because parts of badlands are being levelled and converted into arable land. It is important to monitor these changes for environmental planning. This paper proposes a remote-sensing-based detection method which allows mapping of badland dynamics based on seasonal vegetation changes in the lower Chambal valley, India. Supervised classification was applied on three Landsat (Thematic Mapper) images, from 3 different seasons; January (winter), April (summer) and October (post-monsoon). Different band selection methods were applied to get the best classification. Validation was done by ground referencing and a GeoEye-1 satellite image. The image from January performed best with overall accuracy of 87% and 0.69 of kappa. This method opens the possibilities of using semi-automatic classification for the Chambal badlands which is so far mapped with manual interpretations only.     

应用图像关联度的图像模糊分类

提出了应用图像关联度的图像模糊分类新方法。该方法在求得每幅图像相对各个类别的关联度的基础上,求得每个类别图像关联度的均值和标准差σ;然后计算每幅图像的关联度与所在类别关联度均值之差Δi,差值Δi≤2σ,该图像应“留“在该类中;否则,该图像应归入与另两个类别的关联度较大的类别。每幅图像都经过这样的检验,实现图像的再划分,直到这一过程稳定为止。实验结果表明,该方法的图像分类质量有一定优势。     

一种网络空间现象同位模式挖掘的新方法

同位模式的挖掘是空间数据挖掘领域关注的热点问题。目前,对于网络空间现象同位模式挖掘的研究较少。本文将欧氏空间已有方法扩展至网络空间,该方法由两个核心步骤组成:①通过对网络进行划分定义同位腜停范ㄍ止叵担虎诙酝止叵到型臣仆贫先范ㄆ涫欠裎荒J健6陨钲谑兄圃煲倒镜耐荒J酵诰蚪辛朔椒ㄋ得鳎谰菁劬醚е械贾虏导鄣娜只贫哉庑┩荒J浇辛硕ㄐ苑治觯ü胍延蟹椒ǖ谋冉弦约巴纾撕募煅檠橹ち吮疚姆椒ǖ挠行浴€     

基于成本最优路径的划界线与实际地形自动匹配

针对当前划界工作中主要采用手工操作实现划界线与实际地形匹配的现状,提出了基于成本最优路径分析实现自动匹配的基本原理,研究了自动匹配的关键技术。首先,基于悬挂特征栅格的反向追踪构造地形特征网络,并借助划界法理对其进行特殊处理。其次,构建成本图层,应用成本最优路径分析实现了划界线与实际地形的自动匹配。最后,通过实验对匹配结果的可靠性进行评价,验证了本文自动匹配方法的有效性和可靠性,并对划界工作提出了若干建议。