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...     

激光雷达测量技术在地学中的若干应用

对激光雷达测量技术在全球冰川监测、局部断裂带提取、滑坡监测和稳定性评价以及海岸线提取和海岸侵蚀等方面的应用做了较为全面的综述.作为一种新型的对地观测手段,激光雷达(含星载、机载、车载和地面)的应用已经从传统的测绘扩大到包括文物保护在内的诸多其他应用领域.所综述的激光雷达技术在地学研究中的4个应用方面,是传统地学研究中与全球变化和人居环境最为密切的方向.分析表明,激光雷达技术在这些研究方向中的应用大有作为.      

基于Zernike矩的多时相遥感影像匹配方法研究

针对参与变化检测影像的特点,归纳多时相影像匹配方法应满足的条件,提出了一种基于Zernike矩的匹配策略。在进行多时相影像匹配时,该方法对灰度的线性变化不敏感,而且可以匹配存在较大旋转角度的影像。试验表明,该匹配策略具有较高的匹配正确率。     

Monitoring disasters with a constellation of satellites – type examples from the International Charter ‘Space and Major Disasters’

In the case of a major disaster, information derived from satellite observation is not only highly useful, it may at times be indispensable because of the damage caused by the disaster to ground infrastructure. The International Charter ‘Space and Major Disasters’ (‘the Charter’) has been one of the primary sources of satellite data for the past 11 years to cover events like floods, fires, tsunamis, ocean storms, earthquakes, volcanic eruptions and oil spills. With the growing membership of the Charter, an increasingly large number of sensors are now available, which can be planned with the required temporal frequency and spectral range to cover a disaster event. Some of the type Charter activation cases are reported in this article to demonstrate the innovative use of multi-satellite imagery for disaster response.     

Use of multitemporal remote sensing data for mapping the Alfios River network changes from 1977 to 2000

Alfios is the biggest river of the Peloponnese and the ninth longest river in Greece. It drains an area of almost 2575 km² in Western Peloponnese and discharges at Kiparissiakos Gulf. Due to its extent, the Alfios basin presents complex physiography and geomorphology.

During the last 50 years, major direct human activities have had an important impact on the river network shape and on the valley floor morphology. The most important human activities were the channel straightening caused by meander cut-offs in 1950, the construction of two dams: the Ladonas dam in 1955 and the Flokas dam in 1968 and the illegal in-stream gravel extraction that started in the 1960s and continues until today. As a result, the Alfios riverbed has presented very rapid vertical and horizontal (lateral) erosion and significant changes to its network shape. The drainage network follows a straighter course, the number of meanders has been reduced and the seventh order branch is deeply incised. Besides the morphological impacts to the river network shape there are also serious economic damages caused by human activity. During the winter of 1999, the Flokas dam bridge was closed for a long period due to damages at its foundation caused by the heavy rain and illegal gravel extraction. Transportation between the villages in the area became extremely difficult and time consuming and the cost of the repairs was estimated at 500 000 euros. In this study, there is an effort to map the network shape transformation for the period 1977–2000 using multitemporal and multisensor satellite images. One Landsat MSS image, three Landsat TM images, two Landsat ETM images, and one Terra ASTER image have been orthorectified and processed in order to cover the specific period. Principal Component Analysis (PCA) method and Geographical Information System (GIS) techniques have been applied to map the changes in the Alfios River channel. The drainage network straightening and the cut-off of five big meanders were detected and mapped. These changes occurred between 1986 and 2000.     

最优小波尺度空间的图像边缘检测方法

根据图像和检测算子的特性,以相关性为准则,使用遗传算法对图像小波变换的尺度进行选择,从而构成一种自适应的高斯小波尺度空间.融合该空间下不同尺度检测的图像边缘,使得整幅图像的边缘细节丰富清晰,具有更好的抗噪性能.对测试图像使用Canny算法、单尺度、二进尺度和自适应尺度小波进行边缘检测,验证了该算法在去除噪声和准确定位方面的有效性.     

提高管线探测水平分辨率的技术方法

在平顶山矿工路路面改造工程中,使用GX-2型及RD-4000型地下管线探测仪探测地下管网布设情况。根据地下管线种类、管线布设方位、埋置深度、探测难度大小进行了多次试验,在此基础上确定了电磁探测技术与方法：利用直连法提高被测管线中的交变电流,压制邻近平行管线和地下介质中的异常反映;在多管并存、且间距较小的情况下,应选择“梯度法”对磁场水平分量垂直梯度ΔHx进行观测,以得到最大清晰异常;另外,还应根据具体情况,合理选择诸如压制旁侧管线法、选择发射法、偏移感应法、动源发射法等发射方式,保证目标管线中有较强异常呈现。在管线密集区,应尽可能地降低工作频率,以减小旁侧管线中产生的二次电流及二次磁场;为防止信噪比下降,可适当减小收发距,以提高接收机的灵敏度。     

基于物方影像匹配的DEM聚类变化检测

提出了DEM的变化检测问题,采用了基于密度的聚类分析方法对匹配所得的初始变化区域进行分析,实现了地表高程变化区域的检测.     

面向户外增强现实的地理实体目标检测

本文研究了多种计算机视觉领域中用于图像分类和目标检测的卷积神经网络模型,在一种目标检测模型的基础上进行了修改,并结合了残差神经网络结构,旨在解决户外增强现实中地理实体目标检测的问题。我们根据这一模型开发了一款户外增强现实原型系统,通过将训练好的模型作为目标检测引擎,达到了较好的近实时性和准确性。     

利用热红外温差识别沙尘

沙尘识别是沙尘灾害监测和沙尘气溶胶特性研究的首要工作。利用辐射传输方程进行了沙尘气溶胶的辐射计算 ,对不同沙尘气溶胶光学厚度下的热红外通道温差ΔT(T11μm-T12 μm)的变化进行了分析。理论分析表明 ,利用热红外通道的温度差ΔT进行陆地沙尘识别是可行的。并利用NOAA AVHRR热红外通道的温度差ΔT进行了陆地沙尘识别的实验 ,经与地面气象站实测的沙尘天气现象相比较 ,结果一致。