【计算机科学】【2014.09】激光雷达与图像点云比较

本文为美国海军研究生院（作者：Amanda R. Mueller）的硕士论文，共83页。

本文分析了从机载和卫星光电数据中提取三维点云的新技术。这项研究的目的是比较三种类型的点云，以确定图像点云是否可以与激光雷达点云的精度相竞争。图像点云的两种主要类型是通过摄影测量创建的，具有两幅并排的图像，或者使用多视点立体技术在多个图像之间进行特征匹配。两个处理航空图像的软件包IMAGINE Photogrammetry和Agisoft Photoscan Pro被用来创建这样的模型。他们还用亚米分辨率的卫星图像进行了测试，以确定是否可以制作出更大但仍然真实的模型。研究发现，这两个软件包都不具备垂直分析卫星图像的能力，但这两个软件包在应用于航空图像时都是成功的。摄影测量模型包含的点比多视图模型少，但更好地保持了建筑物的形状。虽然摄影测量模型被确定为两个模型中更精确的，但它仍然没有达到与激光雷达数据精度相比较的程度。

This paper analyzes new techniques used toextract 3D point clouds from airborne and satellite electro-optical data. Theobjective of this research was to compare the three types of point clouds todetermine whether image point clouds could compete with the accuracy of LiDARpoint clouds. The two main types of image point clouds are those createdphotogrammetrically, with two side-by-side images, or through feature matchingbetween multiple images using multiview stereo techniques. Two softwarepackages known for handling aerial imagery, IMAGINE Photogrammetry and AgisoftPhotoscan Pro, were used to create such models. They were also tested withsub-meter resolution satellite imagery to determine whether much larger, butstill truthful, models could be produced. It was found that neither softwarepackage is equipped to vertically analyze satellite imagery but both weresuccessful when applied to aerial imagery. The photogrammetry model containedfewer points than the multiview model but maintained building shape better.While the photogrammetry model was determined to be the more accurate of thetwo it still did not compare to the accuracy of the LiDAR data.

1. 引言
2. 项目背景
3. 数据与软件
4. 处理、结果与分析
5. 总结与结论

更多精彩文章请关注公众号：