传统的水深测量方法多通过舰载声纳实地探测的方法,灵活性较差且水深资料更新周期长,并且在某些海域,船只往往难以靠近从而无法完成测量.本文使用七连屿海域附近的WorldView-2多光谱遥感影像构建了基于梯度提升决策树(Gradient Boosting Decision Tree,GBDT)算法的水深反演模型,并利用单波束与人工测量相结合的水深数据,与传统的单波段模型、双波段模型以及BP神经网络水深反演模型的水深数据进行了水深反演精度对比.结果表明,在0～20 m深海域,GBDT模型反演精度高于其他模型,且更符合实际水深,其检验点的R2为0.9664,RMSE为0.94 m,MAE为0.75 m,RME为19％.

为探究不同遥感水深反演机器学习模型的差异,以WorldView-2高分辨率多光谱影像与实测水深数据为数据源,应用BP神经网络模型、随机森林模型、梯度提升决策树模型及支持向量机模型开展水深反演实验,对4种水深反演模型的精度进行比较与评价.实验结果表明:机器学习模型反演水深,具有一定精度,平均相对误差(MRE)可优于20％.4种模型中,同为集成学习模型的随机森林模型与梯度提升决策树模型在两个实验区域,反演水深的RMSE值、MRE值与R2值明显优于BP神经网络模型和支持向量机模型,具有更好的浅水水深反演效果和适用性.

传统船载水深测量受船只吃水影响,难以在浅水区域开展,遥感水深反演作为传统方法的有益补充,其重要性日益凸显.以GF-1多光谱影像为数据源,以船载声呐实测水深点作为训练样本和检测样本,以相关系数、均方根误差和平均绝对误差作为评价指标,首次将网格搜索+XGBoost模型应用于启东恒大威尼斯浅海区域水深反演.实验表明,网格搜索+XGBoost模型水深反演的相关性系数达到0.820,均方根误差0.247 m,平均绝对误差0.134 m,与GBDT模型和波段比值模型相比,其水深反演精度更高,且易于实现.该研究方法和成果为快速获取大范围浅海水深提供了借鉴,为相关水上勘察和海洋资源开发提供了技术思路.

为了解长江口的水质状况,现场测量叶绿素a浓度,结合高光谱遥感影像,运用波段比值模型、一阶微分模型和水体叶绿素a提取指数(Water Chlorophyll-a Index,WCI)对整个研究区域叶绿素a浓度进行反演推算,并进行空间分布评价;利用实测数据和遥感影像的关系建立反演模型,并结合相关系数、均方根误差和平均相对误差分析和评价反演效果.结果显示,波段比值模型和叶绿素a浓度的相关性达到0.91,均方根误差为1.79,平均相对误差为9.09％;一阶微分模型的相关性为0.95,均方根误差为2.21,平均相对误差为15.31％;WCI模型的相关性高达0.98,均方根误差为1.44,平均相对误差为6.20％.利用WCI模型对整个研究区域的叶绿素a浓度进行模拟,可见研究区域的中间部分叶绿素a含量较低,从中间到两边逐渐增大,南部出现最大值,造成此差异的原因可能是因为北接居民生活区,南邻上海青草沙水库,并且附近存在植被.研究表明,WCI模型的反演效果优于波段比值模型和一阶微分模型,是一种计算简单、精度较高的方法,可以有效地提取水体叶绿素a的浓度,未来可广泛应用于水体环境质量监测.

为提高多波束数据处理的准确性和简便性,研究了多波束水深点云处理的策略和关键算法,构建了kd-tree(k-dimensional tree)结构加速点云查找,针对大尺度噪点设计了半径滤波和统计滤波算法,针对小尺度噪点提出了双边滤波算法,并对以上算法的适用性进行了探讨分析.实验结果表明,当设置适合的滤波阈值,半径滤波和统计滤波能高效去除大尺度噪点,且保留地形特征,总误差分别为2.25％和2.76％.在大尺度噪点去除的基础上,改进的双边滤波可以实现地形平滑的同时,保持水深点云的有效数据量.研究成果为多波束点云数据的自动化处理提供了解决方案,对多波束测量工作的效果和效率提升做了有益尝试.

利用遥感技术快速提取海岸线是一种重要的技术手段,针对传统分水岭算法在高分辨率多光谱卫星数据处理中存在的过分割和抗干扰能力差的问题,本文提出了一种基于扩展极值变换标记分水岭的算法.首先通过形态学重建、扩展极值变换等方法建立前景和背景标记,初步抑制灰度极小值和极大值区域,然后依据这两类标记对梯度图像进行修正,进而进行分水岭变换,提取岛屿水边线.以南海典型海岛为研究区域,利用2017年GF-2卫星数据对本文提出的方法进行验证和精度评价,研究结果表明:改进的分水岭算法对GF-2数据的人工岸线提取质量在1个像素(4 m)之内高于90％,沙质岸线提取质量在1.5个像素(6 m)之内高于90％,可以用于高分辨率多光谱影像的分割和海岛水边线的提取.

船载移动激光雷达扫描系统由脉冲激光源、测距探测接收器、系统控制处理器和光机扫描器组成,并集成定位定姿系统(Position and Orientation System,POS).本文以自主研发的船载激光扫描仪为研究对象,阐述了扫描系统的软硬件组成与工作流程,并对实验结果进行了精度评定.实验表明,本文设计的船载移动三维激光扫描系统能有效准确地获取目标物的三维点云数据,未来可在海岸带海岛礁测绘、港口码头测量、近岸工程变形监测中发挥重大作用.

Inversion of shallow seawater depth using satellite multi-spectral data is an important measure of water depth measurement.The existing water depth inversion method is to establish an inversion model of unified mathematical parameters in the study area, without considering the problem of spatial non-stationarity caused by changes in sea floor sediment and water quality.In this study, the Geographically Weighted Regression (GWR) model was used to estimate the regression parameters in space.For the influence of the bandwidth of the GWR model on the inversion accuracy, a Cross Validation (CV) method was used to determine the best bandwidth, taking the sea areas of Woody Island and Ganquan Island in the South China Sea as experimental areas, the feasibility and accuracy of the GWR model were verified based on WordView-2multi-spectral data.As a result of the experiment, the accuracy of the GWR model in the study area of Woody Island was improved by 36.05％compared with the linear regression model, and in the ranges of 0-5m, 5-10 m, 10-15 m, and 15-20 m, the precision was increased by 49.46％, 39.97％, 12.36％ and 49.68％respectively.The precision of GWR model in the study area of Ganquan Island was improved by 8.08％.In the ranges of 0-5m, 5-10 m, 10-15 m, and 15-20 m, compared with the linear regression model, the precision was improved by 12.05％, 16.23％, 4.49％and 12.23％respectively, indicating that the GWR model has a better water depth inversion performance.

针对"海图学"的课程特点,制订了"海图学课程设计"的教学内容及实作方案.通过学生应用课堂所学的理论知识、IHO相关国际标准、专业软件等,培养他们创造性思维和实践动手能力.从多年的实际教学效果看,达到了课程建设的预期目标.

According to the principle of remote sensing water depth inversion ,WorldView-2 multi-spectral satellite data and chart depth data of Longwan Port in Hainan Island were used .The water depths in 0-2 m , 2-5 m ,5-10 m ,10-15 m and 15-20 m were partitively treated ,tidal correction ,the correlation analysis and regression analysis of chart water depth and the corresponding image band reflectivity values were also developed ,and a piecewise linear model was built for shallow water bathymetry with which the actual calculation of shallow water depth and precision analysis were carried out .The results show that the depth accuracy of the partitioned linear regression model is higher than that of the non-partitioned model in different water depth ranges .In the partition model ,the multi-band model has the highest inversion accuracy of 0-5 m , and the double-band ratio model has the highest inversion accuracy of 5-20 m ,but inversion of the depth in the most shallow yet to be improved .The water depth extracted by this method is similar to that of chart water depth data ,which can meet the requirements of marine scientific research on large-scale shallow underwater detection requirements .

讨论了差分码偏差DCB在非差数据中的存在形式,以及在非差定位中的改正.采用实测数据,详细研究了DCB(C1-P1)和DCB(P1-P2)对非差定位和解算参数的影响.结果表明,DCB(P1-P2)对单频单点定位的影响比较显著,必须进行相应的改正;DCB(C1-P1)对非差解算参数的影响包括坐标和接收机钟差两个方面,对坐标的影响来自于DCB(C1-P1)的卫星硬件部分,对接收机钟差的影响来自于DCB(C1-P1)的接收机硬件部分.分析DCB(C1-P1)对模糊度参数的影响,结果表明,DCB(C1-P1)改正和不改正时,得到的模糊度参数不一致.当采用无电离层延迟C1/P2、L1/L2和P1/P2、L1/L2分别进行精密单点定位数据处理时,对应的模糊度参数也有差异,差异值等于卫星DCB(C1-P1)的倍数.

The basic performance,composition,advantages and application of C3D bathymetry side-scan sonar system are introduced in this paper.An underwater probing system based on C3D is designed and applied in surveying.The C3D data processing tools and methods are introduced and its strengths,weaknesses,and application prospects are summarized.The tests proved the C3D is a superior performance sonar probing system to meet the needs of a variety of underwater engineering.

由于测深侧扫声呐系统(C3D)和双频声学识别声呐(DIDSON)在浑浊的水体中具有极高的图像分辨率,近年已被广泛应用于渔业生产中.本文针对人工鱼礁建设的规划设计阶段、施工验收阶段的水下声呐探测进行了相关研究,解决了投放区水下地形探测和人工鱼礁水下调查及稳定性探测的一系列问题,实践证明高分辨率水下声呐数据为人工鱼礁的设计、投放、监测等工作提供了十分有效的支持,提高了鱼礁的建设效率.

Abstract LiDAR data processing is an urgent problem to be solved in LiDAR technique application area. The LiDAR data processing module on OpenGL was developed in this study,and the key techniques include LiDAR data reading methods of LAS format, presentation of LiDAR points and improved hierarchical moving curved fitting filtering method of LiDAR points. The LiDAR points filtering experiment was done to validate the LiDAR data processing module, and the filtered results were used to confirm the efficiency of this module in presenting and editing LiDAR points.

环境小卫星HJ是环境与灾害监测预报小卫星星座的组成部分,对提高我国环境监测和防灾减灾能力具有重要意义.针对环境小卫星红外传感器HJ-IRS的性能特点,以MODIS的V4算法为基础,构建了基于HJ-IRS的火点自动探测方法.利用同时相同区域的HJ-CCD火烧迹地影像和MODIS的火点产品数据,对HJ-IRS火点探测方法进行了验证和评价.结果表明,基于火烧迹地的平均实际偏差为11％,与HJ-CCD的火点数量相近率为94％,基于MODIS火点的平均参考偏差为13％,与MODIS的火点数量相近率为89％,表现出良好的可靠性和稳定性.有利于提高环境小卫星HJ森林火灾的自动监测和快速评估能力.

In order to reconstruct 3D building models in a digital city,an Alpha Shapes algorithm has been developed in this study to extract the building boundaries of flat-roofed buildings.The 3D building models are then automatically reconstructed based on the boundaries extracted and the average elevation of buildings.In addition,the clustering method based on normal vector and relevant regularization strategies has been developed to reconstruct the 3D building models of various kinds of common non-flat-roofed buildings.The case studies demonstrate that these algorithms and methods are of high precision,effective,concise and stable,and self-adaptive.They are highly suitable for the automatic construction of 3D building models based on LIDAR data.

The digital elevation data of Shuttle Radar Topography Mission(SRTM)is an important foundation data for geospatial modeling.However the void data of SRTM caused by the deficiencies of radar produce great inconvenience for its application.An approach and process flow of SRTM void data recovery is constructed using raster to vector conversion and mask processing technology based on spatial interpolation in ArcGIS software.To ensure the validity of interpolation results,the most important step of this process flow is deleting points of void data.The interpolation accuracy and operations efficiency of Inverse Distance to a Power(IDW),Kriging,Nearest Neighbor(NN)and Spline are compared and analyzed based on a variety of test samples under different conditions.16 test samples of 4×4 pixel window size distribute on different landforms regions,while 12 test samples of different pixel window size from 1×1 to 12×12 locate on the same landforms region.The hollows of SRTM void data are simulated artificially through fill some negative value to the SRTM data in the areas of test samples.With this way to obtain the true value used to estimate interpolation accuracy.The mean of interpolation error is in decreasing order is IDW,NN,Krging and Spline,whose value is 1.32%,1.13%,1.07% and 0.90% respectively.Their standard deviation value are 0.90%、0.81%、0.62% and 75% respectively.Results show that Spline interpolation method is the optimal one applicable to recover void data of SRTM.When the artificial simulation window size of empty area is 4×4 pixels,the mean and standard deviation of interpolation error is 0.90% and 0.75% respectively.With the levels of undulating topography increase,the mean and standard deviation of interpolation error will increase accordingly.On the other hand,interpolation accuracy will descend with the accretion of window size.When the window size of empty area increase form 4×4 pixels to 12×12 pixels,the mean of interpolation error increased from 0.92% to 3.84%,standard deviation increased from 0.75 % to 3.26%.6 × 6 pixels is the largest window size of empty area which can restrict the interpolation error within the data precision of SRTM.Experimental results show that the approach proposed in this paper is a good way to recover SRTM void data only using ArcGIS software.

The building boundary extraction and normalization are the key approach for LIDAR data processing and building 3D modeling.In this paper,Alpha is first applied on the LIDAR data to extract the building boundary.In addition,an enhanced boundary simplifying algorithm,i.e.Pipe and two other developed normalization Regularization and Adjustment are used to improve the extracted boundary.Finally,the normalized building boundary is generated perfectly with these algorithms.A limited accumulated points S has an alpha shape in polygon.This polygon is determined by S and α.We can imagine that a circle with an α radius is rolling around the S.When α value is big enough,the circle will not fall into the area of accumulated points.The rolling track will form the boundary of these discrete points(for example LIDAR data).Contrarily,when the α value is very small(α→ 0),every point might be the boundary.When the alpha value is approaching infinity(α→∞),alpha shape will be the convex hull.When the S contains evenly distributed points and α value approaching optimum value,the alpha shape can extract the inner and outer boundary of convex and concave polygon.The boundary obtained from Alpha Shape Algorithm above is rough which can be defined as raw boundary.In this paper,an enhanced simplifying algorithm,i.e.Pipe is developed to simplify the raw outline which usually is composed of zigzag shape.Pipe Algorithm retrieves the polygon inflexion points based on the changes of angle direction.These inflexion points are retained while the intermediate points are eliminated.The remained inflexion points will establish a basic framework of the polygon.At the same time,two other developed normalization algorithms,Circumcircle Regularization Algorithm Cluster and Adjustment Algorithm are used to improve the extracted polygon framework.Now,the two normalization algorithms can be applied for four sided and multi-sided polygon(greater than four sides and the number of side must be even number).Compared with other algorithms,Alpha Shapes algorithm can effectively and stably process LIDAR points-cloud data with high precision.At the same time,it can keep fine features of any building shape boundary while filter the footprints of non-building.The experiment results show that these algorithms are excellent in building boundary(convex concave polygons) extraction and normalization.The error between the extracted building boundary and the actual outline is usually less than 0.5 m.

Based on the physical concept of heat energy of preignition, a new fire susceptibility index (FSI) is used to estimate the forest fire risk. FSI combines remotely sensed data with meteorological data and has physical basis. FSIs are computed in Peninsular Malaysia for nine days before fire and validated with hotspot data. Results show that FSI increases as the day gets closer to the fire day. It suggests that FSI can be a good estimator of fire risk. FSI retains the flexibility to be localized to a vegetation type or ecoregions for improved performance..

Land covers in urban areas tend to change drastically over a short period of time due to rapid urbanization.Therefore,it is very important to quickly obtain the distributions and area of urban land-use.Remote sensing images are ideally used to monitor current land cover changes thanks to their rapid up-date capability.In this paper,the method of extracting urban land-use from Landsat TM data is discussed.Firstly,the mechanism of remote sensing for urban land-use is analyzed.Secondly,the differences of urban land-use and other land-use type are discussed.Third,the structure feature among spectrum of urban land-use and the other land-use types is analyzed.Vegetation and water can be extracted by NDVI and NDWI with suitable threshold.Urban land-use and bareness areas can be distinguished by the formula such as TM3-TM20 and TM4-TM30.This method is much faster and better accurate.The experiment results show that this method is very simple and effective to the semi-automatic extraction of the urban land-use.