Estimation of crowd density from UAVs images based on corner detection procedures and clustering analysis

With rapid developments in platforms and sensors technology in terms of digital cameras and video recordings, crowd monitoring has taken a considerable attentions in many disciplines such as psychology, sociology, engineering, and computer vision. This is due to the fact that, monitoring of the crowd is necessary to enhance safety and controllable movements to minimize the risk particularly in highly crowded incidents (e.g. sports). One of the platforms that have been extensively employed in crowd monitoring is unmanned aerial vehicles (UAVs), because UAVs have the capability to acquiring fast, low costs, high-resolution and real-time images over crowd areas. In addition, geo-referenced images can also be provided through integration of on-board positioning sensors (e.g. GPS/IMU) with vision sensors (digital cameras and laser scanner). In this paper, a new testing procedure based on feature from accelerated segment test (FAST) algorithms is introduced to detect the crowd features from UAV images taken from different camera orientations and positions. The proposed test started with converting a circle of 16 pixels surrounding the center pixel into a vector and sorting it in ascending/descending order. A single pixel which takes the ranking number 9 (for FAST-9) or 12 (for FAST-12) was then compared with the center pixel. Accuracy assessment in terms of completeness and correctness was used to assess the performance of the new testing procedure before and after filtering the crowd features. The results show that the proposed algorithms are able to extract crowd features from different UAV images. Overall, the values of Completeness range from 55 to 70 % whereas the range of correctness values was 91 to 94 %.     

Classification of glacial lakes using integrated approach of DFPS technique and gradient analysis using Sentinel 2A data

It is important to identify and locate glacial lakes for assessing any potential hazard. This study presents a combination of semi-automatic method Double-Window Flexible Pace Search (DFPS) and edge detection technique to identify glacial lakes using Sentinel 2A satellite data. Initially, Normalized Difference Water Index (NDWI) has been used to identify water and non-water areas, while DFPS and Edge detection technique has been used to identify an optimum threshold value to distinguish between water and shadow areas. The optimal threshold from DFPS process is 0.21, while threshold value of gradient magnitude using edge detection process is 0.318. The number of glacial lakes identified using the above algorithm is in close agreement with previously published results on glacial lakes in Gangotri glacier using different techniques. Thus, a combination of DFPS and edge detection process has successfully segregated glacial lakes from other features present in Gangotri glacier.     

An improved building detection approach using L-band POLSAR two-dimensional time-frequency decomposition over oriented built-up areas

The objective of this study is to efficiently extract detailed information about various man-made targets in oriented built-up areas using polarimetric synthetic aperture radar (POLSAR) images. This paper develops an improved approach for building detection by utilizing Two-Dimensional Time-Frequency (2-D TF) decomposition. This method performs outstandingly in distinguishing between man-made and natural targets based on the isotropic behaviors, frequency-sensitive responses, and scattering mechanisms of objects. The proposed method can preserve the spatial resolution and exploit the advantages of TF decomposition; specifically, the exact outlines of buildings can be effectively located, and more types of features (e.g., flat roofs, roads, and walls that are oblique to the radar illumination) can be distinguished from forests in complex built-up areas by 2-D TF decomposition. The coarser-resolution subaperture images that are produced in the azimuth direction, which correspond to different looking angles, are beneficial for detecting man-made structures with main scattering centers oriented at oblique angles with respect to the radar illumination. In the range direction, the obtained subaperture images, which correspond to various observation frequencies, can be helpful in distinguishing flat roofs and roads from forests. This method was successfully implemented to analyze both NASA/JPL L-band AIRSAR and L-band EMISAR data sets. The building detection results of the proposed method exhibit a significant improvement over those of other methods and reach an overall accuracy over 80%, with approximately 20% higher than the accuracies of K-means clustering and the entropy/alpha-Wishart classifier and approximately 10% higher than the accuracy of the support vector machine method. Moreover, building details can be precisely detected, obliquely oriented buildings can be identified, and the distinction between buildings and forests is significantly improved, as both visually and statistically indicated. This method is highly adaptable and has substantial application value.     

基于ETOPO1数据的全球及重点海域水深统计与分析

由于不同的水深探测手段及海深反演方法适用于不同深度的海域,且全球海域水深分布不均,因此在探测及反演海深之前,需对目标海域的水深分布有一定认识。本文采用ETOPO1模型对全球及部分海域进行统计分析,结果表明,全球及大西洋、印度洋、太平洋区域的水深大多处于3 000—10 000 m,其中3 000—5 000 m水深海域占比较大,因此,在这些海域主要采用重力数据结合船测数据进行反演,同时应注意避开海岸带、岛屿较多、海脊存在的区域;对于水深小于1 000 m的海域,可以根据外部环境,采用多波束船测、水深遥感、机载激光测深等方式进行探测。通过分析及统计全球和重点海域的水深可以为海深探测及海底地形反演方法的选取提供参考。     

深度卷积神经网络支持下的遥感影像飞机检测

针对YOLOv3算法对小目标检测较差及出现较多漏检的问题,本文提出了一种优化的YOLOv3算法.首先使用K-means算法计算出与数据集相适用的锚框;其次将扩张卷积引入到YOLOv3网络,用来增强网络高层的感受野,改善小目标的检测效果;然后使用深度可分离卷积取代YOLOv3网络残差模块中的普通卷积,可减少计算量,从而得到一种新型卷积神经网络结构;最后在数据集上进行对比试验.结果表明,优化的YOLOv3算法能够检测出更多目标,降低漏检率,相比于YOLOv3算法,其召回率提高11.86%,F1-score提高2.99%.     

基于GPS新型L5信号的地表雪深反演研究

利用GPS多路径反射信号测量地表雪深具有全天候和高时空分辨率的特点,因此其可作为一种代替气象站监测雪深的新手段。然而,先前大多数研究仅使用了GPS L1和L2C波段信噪比数据探测积雪深度。为验证新型的L5信号在雪深反演方面的优越性,本文阐述了GPS-R技术反演雪深的原理,利用Lomb-Scargle周期图法所处理的受积雪表层影响的信噪比数据计算了频谱振幅强度,通过获取频谱特征值与天线高度的关系求解雪深值,最后分别与L1反演结果和实测雪深数据进行了对比。试验结果表明：与现有的GPS-R测量雪深结果相比,利用新型的L5反射信号反演地表雪深的精度更佳;采用GPS-R技术探测雪深对把握测站区域内的雪深变化情况和淡水资源储量具有重要价值。     

GF-2支持下的干旱区稀疏植被区植被覆盖度估算

现有像元二分模型MODIS植被覆盖度模型因其形式简单、适用性较强的特点被广泛应用于区域植被覆盖度（FVC）的估算。然而,研究表明在沙漠和低植被覆盖的西部干旱区,从250 m的影像上很难精准地获取NDVI_veg（全植被覆盖植被指数）和NDVI_soil（全裸土区植被指数）参数。利用常用的直方图累计法获取模型所需参数NDVI_veg和NDVI_soil,估算结果存在普遍高估现象。为此,本文首先引入同期获取的GF-2号卫星数据,从GF-2号影像上提取植被覆盖像元;然后,利用Pixel Aggregate方法重采样至250 m分辨率,获取250 m空间分辨率下纯植被和纯裸土像元;最后,将纯植被和纯裸土像元各自空间位置相对应的MODIS NDVI数据最大值作为模型所需NDVI_veg和NDVI_soil参数,实现研究区内植被覆盖度的估算。试验通过与线性回归法、多项式回归法和直方图累计像元二分模型法估算结果进行精度对比,结果表明：利用GF-2影像辅助的像元二分模型,精准地获取了低植被覆盖区NDVI_veg和NDVI_soil模型参数,提高了干旱区植被覆盖度的估算精度,并有效地抑制了受稀疏植被影响NDVI在干旱区普遍偏高问题导致的FVC高估的现象。     

建筑物侧面轮廓线约束的高分辨率遥感影像建筑物高度估算方法

针对现有利用阴影长度法提取建筑物高度时存在的阴影间相互遮挡问题,提出了一种基于建筑物侧面轮廓线进行建筑物高度估算的新方法。首先,利用RPC模型计算建筑物像点位移的方向与卫星成像角度,再将遥感影像进行旋转,使建筑物像点位移沿水平方向;然后,利用Canny算法进行轮廓检测,并构建一定长度的矩形形态学结构元素,对轮廓图像进行形态学开运算,以提取侧面轮廓线,再利用Hough变换与建筑物角点约束,对所提取的轮廓线进一步筛选;最后,根据卫星侧视成像时建筑物高度与像点位移的几何关系进行建筑物的高度估算。利用实际的高分辨率卫星影像对本文方法进行了验证,并与阴影法估算建筑物高度进行了对比。试验结果证明,利用建筑物侧面轮廓线进行建筑物高度估算平均误差可以达到0.7 m,且实际精度优于使用阴影法进行建筑物高度估算。     

利用改进型MODIS火点探测算法实现河北省秸秆焚烧火点识别

为准确地了解河北省秸秆焚烧火点的空间分布,为秸秆焚烧监测的实现、禁烧工作的开展、环境质量改善提供支持.基于MODIS L1B数据、MODIS标准火点产品MOD14、全国秸秆焚烧火点日报数据为基础,采用改进型MODIS火灾探测算法,并通过IDL语言实现,得到秸秆焚烧火点空间分布信息,并进行空间与定量精度分析.研究表明:火点大部分位于河北省南部的一些地区,其中尤以邢台、石家庄、邯郸火点数量最为突出;该算法运算速度快,获取的秸秆焚烧火点数据具有一定检测精度和可靠性,对秸秆焚烧的监测具有一定的实用价值.     

高分辨率遥感影像配准与变化检测一体化处理

为了克服高分辨率遥感影像配准与变化检测作为单独环节处理的局限,该文提出了一种基于变分理论的配准与变化检测一体化处理方法。该方法将配准误差作为一种光谱变化决策因子,变化信息以权值的方式迭代反馈给变分配准模型的解算过程。为了更准确地检测建筑物这个特定目标的真实变化,该文采用多尺度最大形态学轮廓建筑物检测指数的差异作为另外一个决策因子。最后将配准误差反映的变化和建筑物检测指数的差异这两个决策因子在D-S证据理论框架下建立概率模型进行融合处理,进而得到建筑物的变化检测结果。该文选取WorldView-2数据进行实验,实验结果表明,一体化处理思路可以有效地解决单独处理的局限,从根本上解决配准误差对变化检测结果的影响以及由于变化而使配准精度降低的问题,进而提高配准和变化检测的质量。