Multi-class geospatial object detection and geographic image classification based on collection of part detectors

The rapid development of remote sensing technology has facilitated us the acquisition of remote sensing images with higher and higher spatial resolution, but how to automatically understand the image contents is still a big challenge. In this paper, we develop a practical and rotation-invariant framework for multi-class geospatial object detection and geographic image classification based on collection of part detectors (COPD). The COPD is composed of a set of representative and discriminative part detectors, where each part detector is a linear support vector machine (SVM) classifier used for the detection of objects or recurring spatial patterns within a certain range of orientation. Specifically, when performing multi-class geospatial object detection, we learn a set of seed-based part detectors where each part detector corresponds to a particular viewpoint of an object class, so the collection of them provides a solution for rotation-invariant detection of multi-class objects. When performing geographic image classification, we utilize a large number of pre-trained part detectors to discovery distinctive visual parts from images and use them as attributes to represent the images. Comprehensive evaluations on two remote sensing image databases and comparisons with some state-of-the-art approaches demonstrate the effectiveness and superiority of the developed framework.     

通过软硬变化检测识别冬小麦

提出一种软硬变化检测的作物识别方法 SHLUCD(Soft and Hard Land Use/Cover Change Detection Method)。该方法利用多期遥感影像能够有效表达作物的生长物候特征,以达到在离散变化区(即纯净像元区,包括完全转换成作物的突变区域和非作物区域)和连续变化区(即渐变区,混合像元区,是部分转化为作物的区域)准确进行作物的识别。在北京市选择一个研究区,以冬小麦为研究对象,选用2011年10月6日(播种期)和2012年4月16日(拔节期)两期环境减灾1号卫星影像,分别采用硬变化检测方法 HLUCD(Hard Land Use/Cover Change Detection Method)、软变化检测方法 SLUCD(Soft Land Use/Cover Change Detection Method)和SHLUCD进行冬小麦的识别。实验结果表明:在不同尺度窗口下,SHLUCD较传统方法表现出较明显的优势,具有更低的均方根误差RMSE(SHLUCD为[0.14,0.07],HLUCD为[0.15,0.07],SLUCD为[0.16,0.08])和偏差bias(SHLUCD为-0.0008,HLUCD为-0.007,SLUCD为0.014)和更高的决定系数R2(SHLUCD为[0.68,0.86],HLUCD为[0.62,0.86],SLUCD为[0.60,0.86])。针对冬小麦突变区域、冬小麦渐变区域和非冬小麦区域分别进行评价,表明SHLUCD识别精度接近各区最佳的识别方法,进一步验证了SHLUCD的灵活性和适用性。SHLUCD方法在离散变化区能够通过土地覆盖类型状态变化来有效地识别出冬小麦,在连续变化区可识别出土地覆盖的状态变化程度定量表达冬小麦的丰度,是其他作物多时相遥感变化检测的前期实验基础。     

高动态宽带信号码跟踪方法仿真分析

针对高动态场景,单独的码环路很难实现跟踪,由于高动态载波跟踪的算法很成熟,通常应用载波跟踪结果对码环路进行辅助,针对窄体制信号,这种方法可以帮助消除码环的动态误差,但对宽体制信号来说,辅助力度减小。从高动态宽带信号码跟踪误差门限以及跟踪精度入手,分析了单独码跟踪算法的易失锁性,理论和仿真验证应用高动态载波跟踪结果辅助码跟踪算法的有效性,且具有高的跟踪精度。这为导航接收机的跟踪算法提供了理论依据。     

An Overview of Nowcasting Development, Applications, and Services in the Hong Kong Observatory

The Hong Kong Observatory （HKO） has been developing a suite of nowcasting systems to support op- erations of the forecasting center and to provide a variety of nowcasting services for the general public and specialized users. The core system is named the Short-range Warnings of Intense Rainstorm of Localized Systems （SWIRLS）, which is a radar-based nowcasting system mainly for the automatic tracking of the movement of radar echoes and the short-range Quantitative Precipitation Forecast （QPF）. The differential, integral （or variational）, and object-oriented tracking algorithms were developed and integrated into the nowcasting suite. In order to predict severe weather associated with intense thunderstorms, such as high gust, hail, and lightning, SWIRLS was enhanced to SWIRLS-II by introduction of a number of physical models, especially the icing physics as well as the thermodynamics of the atmosphere. SWIRLS-Ⅱ was further enhanced with non-hydrostatic, high resolution numerical models for extending the forecast range up to 6h ahead. Meanwhile, SWIRLS was also modified for providing nowcasting services for aviation community and specialized users. To take into account the rapid development of lightning events, ensemble nowcasting techniques such as time-lagged and weighted average ensemble approaches were also adopted in the nowcasting system. Apart from operational uses in Hong Kong, SWIRLS/SWIRLS-Ⅱ was also exported to other places to participate in several international events such as the WMO/WWRP Forecast Demon- stration Project （FDP） during the Beijing 2008 Olympics Games and the Shanghai Expo 2010. Meanwhile, SWIRLS has also been transferred to various regional meteorological organizations for establishing their nowcasting infrastructure. This paper summarizes the history and the technologies of SWIRLS/SWIRLS-Ⅱ and its variants and the associated nowcasting applications and services provided by the HKO since the mid 1990s.     

基于伪语义相似度模型的中文地址匹配方法

地址匹配中,由于传统相似度模型受字符重叠数影响大,在处理简写、缩写地址要素单元时,错误匹配问题突出;深度学习方法需要大量样本支撑,但庞大的数据量和多样的形式,导致生成样本的成本过高。为解决上述问题,本文首先应用基于条件随机场和双向长短时记忆神经网络的模型,对地址进行分词;然后通过建立一种伪语义相似度,对地址要素进行分级匹配。通过对公安业务中地址数据进行测试,在对缩写、简写等不规范地址描述方面,本文模型能较理想地完成任务,各参考指标均高于0.9。     

结合空间上下文与慢特征分析的多光谱影像变化检测

为了提高多光谱影像变化检测的精度,本文提出了一种结合空间上下文与慢特征分析的方法。首先采用自适应空间上下文提取算法围绕像素构建自适应区域,探索像素周围的上下文信息;然后通过迭代慢特征分析,由相应像素周围的成对自适应区域定量计算成对像素之间的变化强度,增强变化区域与未变区域的可分性;最后生成变化强度图像,采用大津阈值法作二值分类,将变化强度图划分为二值变化检测图。利用Landsat 7卫星ETM+传感器的图像,与4种基于代数的方法及基于变换的方法进行对比试验,结果表明,本文方法在降低漏检方面有所改善,提高了召回率。     

一种基于改进CoHOG的视觉SLAM算法

移动机器人在SLAM的闭环检测环节计算量大、运行时间长、匹配误差大,从而导致闭环检测精度较低。针对该问题,本文在CoHOG闭环检测算法的基础上进行改进,将算法中的HOG描述符改进为GDF-HOG描述符,以增强图像特征表现,提高图像特征提取效率;在匹配环节前添加GDF-HOG全局粗匹配,以减少视觉模板的数量,提高算法的计算效率;在匹配环节后添加感兴趣区域（ROI）位置匹配进行检验,以减少闭环检测的假阳性,提高准确率。将本文闭环检测算法与RatSLAM相结合,在公开数据集与真实环境中进行测试,测试结果表明,本文算法在闭环检测环节的准确率较高,且对环境的适应能力较强。     

空天地一体化智慧农业服务与管理平台

空天地一体化智慧农业服务与管理平台融合北斗定位技术和高分遥感技术,通过实时接收北斗定位数据和定时接收卫星遥感数据,实现农作物的长势监测、病虫害和气象灾害监测、农业旱情监测和灌溉用水管理等应用,从而为农作物估产、病虫害预防、水肥药的变量喷施提供决策。     

航空重力探测目标发现率研究

航空重力测量相比地面测量可快速地完成区域性的重力数据采集。为在发现异常目标体的同时,高效、安全地完成航空重力测量,作者提出了一种新的目标体发现率的评价方式,指导最优化测量参数设计。通过可靠弱异常的判定标准定义目标体可发现的临界条件,确定目标体可发现概率。通过理论模型分析目标体参数(埋深、水平尺寸等)、测网间距和飞行高度对发现率的影响。绘制目标体发现率关于飞行高度、测线距和测点距的三维图像。     

北京山区泥石流灾害预警方法研究

北京山区泥石流灾害较为频繁,总体以暴雨型沟谷泥石流为主,受地形、地貌、地质、降雨以及松散物类型等因素的影响比较明显。在调查分析北京山区泥石流灾害发育特征的基础上,从泥石流的形成和启动条件入手,对泥石流灾害的预警方法进行研究与探讨。