Semantic segmentation of high spatial resolution images with deep neural networks

Availability of reliable delineation of urban lands is fundamental to applications such as infrastructure management and urban planning. An accurate semantic segmentation approach can assign each pixel of remotely sensed imagery a reliable ground object class. In this paper, we propose an end-to-end deep learning architecture to perform the pixel-level understanding of high spatial resolution remote sensing images. Both local and global contextual information are considered. The local contexts are learned by the deep residual net, and the multi-scale global contexts are extracted by a pyramid pooling module. These contextual features are concatenated to predict labels for each pixel. In addition, multiple additional losses are proposed to enhance our deep learning network to optimize multi-level features from different resolution images simultaneously. Two public datasets, including Vaihingen and Potsdam datasets, are used to assess the performance of the proposed deep neural network. Comparison with the results from the published state-of-the-art algorithms demonstrates the effectiveness of our approach.     

一种改进包模型的遥感图像检索方法

目的 提出了一种改进的包模型图像检索方法,使用本底图像构建视觉词汇字典。在建库过程中,通过近似KD-Tree搜索本底图像局部不变特征与入库图像特征间的对应匹配关系,并记录该关系;图像检索时,搜索检索图像与本底图像特征的对应关系,进而计算与入库图像的关系。该方法仅保存本底图像的局部不变特征,可实现感兴趣区域的图像检索。实验结果表明,该方法在提高图像检索准确性的同时,可以减少所需存储局部不变特征的数量。     

基于神经网络的遥感影像分类方法研究

本文在研究BP神经网络和模糊理论的基础上,提出了传统BP算法的一种改进方法和基于模糊系统的神经网络遥感影像分类方法。通过试验表明:基于模糊技术的神经网络分类方法要优于BP神经网络方法,取得了令人满意的效果。     

爆炸荷载作用下裂隙岩体内损伤范围的观测研究

通过电法CT成像技术,探测微量炸药爆炸前后裂隙岩体深孔之间电阻率的变化,查明探测切面内的岩体损伤裂隙发育形态和空间展布。通过实验可知：微量炸药爆炸后在裂隙岩体的炮孔深部可以形成一定范围、一定数量和一定宽度的裂隙,其中裂隙形成的区域主要集中在炮孔口深的2/3处,当炮孔深度3 m,装药量为330 g时,在眼深1.5～2.5 m处裂隙较为发育,炮孔间距为1.2 m～1.5 m时,可在岩体内部形成贯通裂隙。     

Spatial structure of central places in Jilin Central Urban Agglomeration,Jilin Province,China

Using the radius of gyration from fractal theory, this paper describes the calculation of fractal dimensions for the four tiers of central places in the Jilin Central Urban Agglomeration(JCUA), Jilin Province, China and the structural characteristics of each tier: 1) the 1st tier central place, Changchun Proper(not including Shuangyang District), provides the most service functions and has the most stable primate position; 2) the 2nd tier central places, Jilin Proper, Siping Proper, Liaoyuan Proper and Songyuan Proper have unclear statuses and do not provide certain functions; 3) the 3rd tier central places comprise 23 county-level cities, counties and urban districts(including Shuangyang District of Changchun), exhibiting a dense spatial structure that agrees with theory; 4) the 4th tier contains the largest number of central places(248 designated towns), but they are loosely distributed. In this study, a spatial image of the JCUA was created, based on vectorized data of the urban settlement distribution, which was then modified and abstracted to create a hexagonal network covering the JCUA. Compared to the traditional central place model, the modified spatial image conforms to the K = 3 principle. In reality, however, the growth of some 3rd tier central places should be increased with the cities being upgraded to the 2nd tier so as to overcome that tier′s functional deficiency. The loose distribution in the 4th tier should also be changed. This apparent anomaly can be explained by the fact that the classic hexagon model used to describe the way market areas layout does not exist in the real world. However, this should not be viewed as an obstacle to using central place theory. If its assumptions are properly applied, it can still assist research into the spatial structure of regions.     

一种基于多源遥感影像的亚端元光谱提取方法

提出一种利用多源影像来提取低分辨影像亚端元光谱的方法,该方法利用高分辨分类得到的组分与低分辨的光谱之间的线性混合关系,建立了反演出低分辨影像的亚端元光谱的线性模型,并提出一种快速鲁棒反演亚端元光谱的提取方法。利用IKONOS和ETM数据验证了本文提出方法,实验结果表明了该方法的有效性。     

利用SketchUp制作三维地形景观的探讨

在对高分辨率影像进行几何纠正、影像拼接以及制作正射影像图的基础上,制作了某实验区的三维地形景观,并利用SketchUp进行地物的精细建模同时通过ArcGIS 9.3进行三维显示和三维系统的二次开发。实验结果表明：利用高分辨率影像制作的三维地形具有地形逼真、清晰度高、信息量丰富等特点;完成了ArcGIS与三维建模软件的数据交换以及所建立模型的查询功能。     

策勒绿洲植被覆盖动态变化遥感研究

使用经严格配准的2个时相的TM和HRV图像数据为基础底料编制规一化植被指数图（NDVI），根据策勒绿洲特有的生态环境特征，研究植被指数灰度级与盖度级的对应函数关系，将其转换成植被盖度图。最终利用植被盖度图像所提供的各盖度级的数量和空间分配状况来评价图像所包括时段，即1990年至1998年期间植被环境质量的变化，为该区域生态环境综合治理提供新的研究方法。     

基于奇异值分解的遥感图像鲁棒水印算法探讨

针对遥感图像的安全保护问题,文章提出了一种基于DCT-SVD和最近区间量化的鲁棒水印算法,将遥感图像进行分块,选择视觉掩蔽效应好的纹理子块作为嵌入区域,根据混合变换后的系数在各自所属区间上的位置,将系数进行最近区间量化来嵌入水印信息;利用密钥即可实现水印的提取,不需要原始遥感图像参与。实验结果表明,该算法对加噪、滤波、旋转、缩放、随机剪切、JPEG压缩等攻击均具有较强的鲁棒性,可有效保护遥感图像的安全。     

结合双树复小波变换的四阶偏微分影像去噪方法

针对低阶偏微分方程去噪模型通常会使图像的灰度趋于分段常量而产生阶梯效应的问题,结合小波多尺度分解在图像处理中的应用,提出一种结合双树复小波变换(DT-CWT)的四阶偏微分方程(PDE)影像去噪模型。首先采用DT-CWT对噪声影像进行多尺度分解,保留分解后的低频分量不变,其他层复高频分量采用四阶PDE去噪模型去噪,然后重构相应层的高、低频分量,得到最终去噪后影像。对不同噪声强度下的ZY-3卫星影像去噪实验的结果表明,采用本文方法去除遥感影像噪声相比现有方法得到的去噪结果信噪比平均提高了1~2 d B,提高了去噪影像的结构相似度,在有效去除影像高斯噪声的同时能够较好地保留图像细节信息。