Rapid texture mapping from image sequences for building geometry model

An effective approach,mapping the texture for building model based on the digital photogrammetric theory, is proposed. The easily-ac-quired image sequences from digital video camera on helicopter are used as texture resource, and the correspon-dence between the space edge in build-ing geometry model and its line feature in image sequences is determined semi-automatically. The experimental re-sults in production of three-dimension-al data for car navigation show us an attractive future both in efficiency and effect.     

Rapid texture mapping from image sequences for building geometry model

An effective approach, mapping the texture for building model based on the digital photogrammetric theory, is proposed. The easily-acquired image sequences from digital video camera on helicopter are used as texture resource, and the correspondence between the space edge in building geometry model and its line feature in image sequences is determined semiautomatically. The experimental results in production of three-dimensional data for car navigation show us an attractive future both in efficiency and effect.     

一种快速的无人机影像无缝拼接方法

本文针对传统的低空无人机影像拼接处理速度慢和几何精度低的问题,提出了一种快速的无人机影像无缝拼接方法:对输入的原始影像按一定尺度进行降采样,在降采样的影像上进行SURF特征提取和匹配,利用RANSAC方法估计初始的相对单应矩阵,然后用Levenberg-Marquardt方法精化单应矩阵,计算出初始的绝对单应矩阵后利用稀疏光束法平差估计出精确的单应矩阵,通过降采样影像与原始影像的单应关系传递单应矩阵到原始影像级,最后进行影像合成,形成整体拼接图.实验结果表明该方法可以有效地提高拼接速度,解决拼接错位问题.     

一种三维激光点云中建筑物立面渐进分割方法

针对建筑物立面分割的问题,该文提出了一种三维激光点云中建筑物立面渐进分割方法。建立三维格网索引,通过分析建筑物立面在三维格网内的空间分布特征和二维平面格网内投影的线性分布特征,确定立面种子格网和投影线的拟合点,拟合投影线并基于种子格网约束生长完成每层格网中立面粗分割;使用RANSAC算法对粗分割后的立面点云进行面拟合,实现精细分割,并将各层格网中的立面进行合并,实现建筑物立面的完整分割。实验结果表明,该方法能有效地实现建筑物立面的精细化分割,有助于后续的建筑物精细三维模型重建。     

以卫星影像为底图的无人机图像自动拼接方法

针对无人机影像受成像角度倾斜、光照不均匀等影响,导致常规匹配方法拼接效果较差的问题,该文提出一种基于现有卫星影像的由粗到精无人机图像自动拼接方法。首先,在对图像进行粗匹配阶段,提出了一种基于环状描述子的尺度不变特征变换,用于解决图像间旋转和尺度的差异,较好地克服了SIFT算法描述子维度较高且计算量大的不足。其次,引入分块互信息进行精匹配,减少拼接过程中的累计误差。最后,采用该文提出算法对无人机图像进行拼接,实验结果表明,该文提出的无人机图像拼接算法在满足实时性要求的同时可以达到较高精度。     

Generalization of 3D building texture using image compression and multiple representation data structure

Textures are an essential part of 3D building models and often consume large portions of the data volume, thus making visualization difficult. Therefore, we propose a multi-resolution texture generalization method to compress the textures of 3D building models for dynamic visualization at different scales. It consists of two steps: image compression and texture coloring. In the first step, texture images are compressed using wavelet transformation in both the horizontal and the vertical direction. In the second step, a TextureTree is created to store building texture color for dynamic visualization from different distances. To generate a TextureTree, texture images are iteratively segmented by horizontal and vertical dividing zones, until each section is basically in the same color. Then the texture of each section is represented by their main color and the TextureTree is created based on the color difference between the adjacent sections. In dynamic visualization, the suitable compressed texture images or the TextureTree nodes are selected to generate 3D scenes based on the angle and the distance between the viewpoint and the building surface. The experimental results indicate that wavelet based image compression and the proposed TextureTree can effectively represent the visual features of the textured buildings with much less data.     

基于数字图像的车辆行驶跑偏在线测量方法

保证车辆在正常情况下沿直线行驶是汽车制造的基本要求,汽车制造厂商高度重视车辆行驶跑偏的检测.文章分析了车辆行驶跑偏量测量的一般方法并提出了一种基于数字图像的汽车行驶跑偏检测方法,实现了车辆行驶跑偏量的自动测量.通过实验,验证了该方法的可行性,且具有自动化程度高、效率高、测量精度高以及可靠性强等特点.基于数字图像的车辆行驶跑偏在线测量系统已投入汽车制造公司的实际生产使用.     

建筑物立面重复模式的点云投影直方图分割方法

针对建筑物的门、窗、梁、柱等具有规则几何形状的重复模式,提出基于投影点云特征直方图(PPFH)的边界语义分割方法.该方法首先利用PC A方法确定立面的主方向,然后根据点云分布特点,分别进行正立面、侧立面投影得到点云分布直方图,根据直方图的极值点对重复结构边界进行提取和语义分割,最后利用LS-TLS方法进行规则化及误差分析.通过对某建筑物立面窗户提取的算例表明:该方法直接利用重复结构的几何特性,算法简洁,对边缘部分遮蔽和缺失的点云也适用,通过对50个窗户的统计分析,长度和宽度平均误差约6 m m,可广泛用于建筑高精度BIM建模.     

平行摄影序列影像的核线影像生成方法

针对平行摄影方式较传统航空摄影方式获取的影像在核线排列及采集方面存在较大差异,导致平行摄影序列影像难以生成核线影像的问题,该文在研究平行摄影方式的摄影基线与影像核点位置间几何关系的基础上,推导了平行摄影序列影像核点位置的严密计算公式,并根据相邻平行摄影序列影像重叠区域最大化原则选取核线生成的起始点和终止点,以此确定核线影像的生成区域,最后采用灰度双线性内插法完成核线影像的采集。以勘探洞平行摄影序列影像作为实验对象,首先基于该文算法生成其核线影像,然后采用密集匹配算法建立左、右核线影像像素间对应关系,最后结合前方交会算法生成洞室的三维点云,进而验证了该文核线影像生成算法的正确性。     

一种放电检测中紫外和可见光图像配准方法

为了将紫外图像和可见光图像进行融合让人们及时清晰地观测到设备故障,该文提出了一种基于相机模型和相机间透视变换的图像配准方法和相应的标定装置,能够实现高精度、小计算量的高效紫外图像和可见光图像的对齐。紫外和可见光的信息融合不仅能够探测目标发射的紫外光谱,还可以对目标进行清晰成像、精确定位,生成的多光谱图像改善单一光谱成像信息的局限性。利用多源图像的信息互补,提高图像检测的准确性、及时性,也更适合计算机的后续分析与处理。     

Automatic road extraction using high resolution satellite image based on texture progressive analysis and normalized cut method

In this paper the approach for automatic road extraction for an urban region using structural, spectral and geometric characteristics of roads has been presented. Roads have been extracted based on two levels: Pre-processing and road extraction methods. Initially, the image is pre-processed to improve the tolerance by reducing the clutter (that mostly represents the buildings, parking lots, vegetation regions and other open spaces). The road segments are then extracted using Texture Progressive Analysis (TPA) and Normalized cut algorithm. The TPA technique uses binary segmentation based on three levels of texture statistical evaluation to extract road segments where as, Normalized cut method for road extraction is a graph based method that generates optimal partition of road segments. The performance evaluation (quality measures) for road extraction using TPA and normalized cut method is compared. Thus the experimental result show that normalized cut method is efficient in extracting road segments in urban region from high resolution satellite image.     

一种基于 Contourlet 变换的遥感影像超分辨率重建改进方法

针对原有方法轮廓波(Contourlet)分解高频信息处理的不足,提出用零填充重采样代替双三次插值用于高频信息处理的新方法,该方法可使Contourlet分解后的高频信息重采样过程不引入噪声能量的同时获得理想的插值结果。改进方法利用Contourlet多分辨率分析的特征,以及零填充重采样方法较传统插值方法优越的性能,此两者相结合可改进原有方法的不足,提高单帧影像的超分辨率重建效果。经理论分析与试验验证,改进后方法的超分辨率重建结果无论在定性分析与定量对比上都要优于传统方法。     

一种高分辨率卫星影像自动云检测方法

针对多数已有的卫星影像云检测方法利用云独特的光谱特征、构造指数图像进行云识别的方法,对传感器、环境参数、云层类型等比较敏感的问题,提出了一种改进HIS模型的高分辨率卫星影像自动云检测方法,以有效弥补传统纹理分析方法计算精度低、效率低下的问题.采用两景国产高分二号遥感影像进行试验.结果 表明,该方法错误率低于5％、查全率高于92％、查准率高于95％.     

资源三号卫星前视影像的建筑物高度提取

建筑物高度对建筑物容积率、城市风向以及城市环境等都具有明显的影响。针对太阳入射方向与卫星观测方向在建筑物异侧时,建筑物侧面与阴影在遥感影像上因极其相似而难以区分的问题,该文基于资源三号卫星前视影像,利用基于规则的面向对象特征提取方法提取建筑物侧面及阴影特征。根据卫星成像时的太阳、卫星以及建筑物之间的空间几何关系,构建了建筑物侧面与阴影的长度比例系数,进而估算了建筑物的高度信息。最后以实测高度进行了高度提取的精度评价,验证结果表明,反演的平均精度达到了92.28%,证明了资源三号卫星前视影像在提取建筑物高度方面的良好可行性。     

Evaluation of LiDAR and image segmentation based classification techniques for automatic building footprint extraction for a segment of Atlantic County,New Jersey

Extracting high-quality building footprints is a basic requirement in multiple sectors of town planning, disaster management, 3D visualization, etc. In the current study, we compare three different techniques for acquiring building footprints using (i) LiDAR, (ii) object-oriented classification (OOC) applied on high-resolution aerial photographs and (iii) digital surface models generated from interpolated LiDAR point cloud data. The three outputs were compared with a digitized sample of building polygons quantitatively by computing the errors of commission and omission, and qualitatively using statistical operations. These findings showed that building footprints derived from OOC gave highest regression and correlation values with least commission error. The R² and R values (0.86 and 0.92, respectively) imply that the footprint areas derived by OOC matched more closely with the actual area of buildings, while a low commission error of 24.7% represented a higher number of footprints as correctly classified.     

图像分数维计算模型的改进

分数维作为一种图像纹理特征常用于遥感图像分类,提高图像分数维的测量有助于提高遥感分类精度。本文在介绍目前常用的基于分形布朗运动分数维计算方法的基础上,运用分形的相关理论,指出了现有方法存在的缺陷,提出了改进的分数维计算方法;并用SPOT全色影像中常见的5种地物进行了对比实验,实验结果表明,本文提出的改进方法获得的分数维测量结果比现有的方法更接近理论分析结果,更有利于地物的识别与分类。     

An automated algorithm for mapping building impervious areas from airborne LiDAR point-cloud data for flood hydrology

Buildings, as impervious surfaces, are an important component of total impervious surface areas that drive urban stormwater response to intense rainfall events. Most stormwater models that use percent impervious area (PIA) are spatially lumped models and do not require precise locations of building roofs, as in other applications of building maps, but do require accurate estimates of total impervious areas within the geographic units of observation (e.g. city blocks or sub-watershed units). Two-dimensional mapping of buildings from aerial imagery requires laborious efforts from image analysts or elaborate image analysis techniques using high spatial resolution imagery. Moreover, large uncertainties exist where tall, dense vegetation obscures the structures. Analyzing LiDAR point-cloud data, however, can distinguish buildings from vegetation canopy and facilitate the mapping of buildings. This paper presents a new building extraction approach that is based on and optimized for estimating building impervious areas (BIA) for hydrologic purposes and can be used with standard GIS software to identify building roofs under tall, thick canopy. Accuracy assessment methods are presented that can optimize model performance for modeling BIA within the geographic units of observation for hydrologic applications. The Building Extraction from LiDAR Last Returns (BELLR) model, a 2.5D rule-based GIS model, uses a non-spatial, local vertical difference filter (VDF) on LiDAR point-cloud data to automatically identify and map building footprints. The model includes an absolute difference in elevation (AdE) parameter in the VDF that compares the difference between mean and modal elevations of last-returns in each cell.

The BELLR model is calibrated for an extensive inner-city, highly urbanized small watershed in Columbia, South Carolina, USA that is covered by tall, thick vegetation canopy that obscures many buildings. The calibration of BELLR used a set of building locations compiled by photo-analysts, and validation used independent building reference data. The model is applied to two residential neighborhoods, one of which is a residential area within the primary watershed and the other is a younger suburban neighborhood with a less-well developed tree canopy used as a validation site. Performance results indicate that the BELLR model is highly sensitive to concavity in the lasboundary tool of LAStools® and those settings are highly site specific. The model is also sensitive to cell size and the AdE threshold values. However, properly calibrated the BIA for the two residential sites could be estimated within 1% error for optimized experiments.

To examine results in a hydrologic application, the BELLR estimated BIAs were tested using two different types of hydrologic models to compare BELLR results with results using the National Land Cover Database (NLCD) 2011 Percent Developed Imperviousness data. The BELLR BIA values provide more accurate results than the use of the 2011 NLCD PIA data in both models. The VDF developed in this study to map buildings could be applied to LiDAR point-cloud filtering algorithms for feature extraction in machine learning or mapping other planar surfaces in more broad-based land-cover classifications.     

The assessment of multi-sensor image fusion using wavelet transforms for mapping the Brazilian Savanna

Multi-sensor image fusion using the wavelet approach provides a conceptual framework for the improvement of the spatial resolution with minimal distortion of the spectral content of the source image. This paper assesses whether images with a large ratio of spatial resolution can be fused, and evaluates the potential of using such fused images for mapping the Brazilian Savanna. Three types of wavelet transforms were used to perform the fusion between MODIS and Landsat TM images. Five quality measures were defined to assess the quality of the fused images. The results showed that it was possible to perform the fusion of MODIS and TM images and the pyramidal in Fourier space wavelet transform provided the best quality measures for the fused images. Classification results showed that fused images could be used for mapping the Brazilian Savanna with an accuracy level comparable to the Landsat TM image.     

共面约束的机载LiDAR IMU安置角误差自动检校方法

针对机载LiDAR数据处理中IMU安置角误差检校问题,该文提出一种基于共面约束的自动检校方法。从激光点云中自动提取尖顶房屋顶平面,并建立连接平面关系,基于激光脚点坐标计算公式和共面约束条件,通过平差解算得到IMU安置角误差参数。以Riegl Q780获取的数据进行实验,该方法检校结果与RiProcess软件提供的检校参数非常接近。经过误差改正后,相邻航线获取的点云可以很好地重合在一起。相对于人工选择特征地物进行IMU安置角误差检校,该方法可以大大提高检校的效率和可靠性。     

A novel method for enhancement of radiometric resolution using image fusion

The Resourcesat-I satellite is equipped with different types of sensors with varied characteristics. For the effective utilization of the available multi-sensor, multi-temporal, multi-spectral and multi-radiometric data from these sensors, fusion of digital image data has become a valuable technique. Image fusion enhances the information content and helps in better discrimination of various land cover types. The Resourcesat-1 has equipped with three sensors, AWiFS, LISS-III and LISS-IV, which are having identical spectral resolutions, with different spatial, radiometric and temporal resolutions. The spatial resolutions ratio of the data set for merging are required to be maximum of 1:6, where as the data sets (AWiFS and LISS-III) that are used in the current study are having the ratio of 1:2.5 approximately. A novel merging technique is designed, which retains the multi-spectral response of the input data in the output data. The merged data set provides the higher spatial and radiometric resolutions. In order to evaluate the fusion merits quantitatively, all the data sets are digitally classified and studied the output classes for homogeneity and clear discrimination. A comprehensive comparative study is carried out between the fused image and the LISS-III image based on the contingency matrix and the scatter plots, which demonstrates the strength of fused image for discriminating the object classes at 23.5 m spatial and 10-bit radiometric resolutions. The merged data set gives the improved classification accuracy.