本文選題：實時渲染 + 八叉樹　； 參考：《天津師范大學》2017年碩士論文

【摘要】：三維點云數(shù)據(jù)的交互式處理通常依賴點云數(shù)據(jù)三維實時渲染。隨著三維激光掃描技術(shù)的深入應(yīng)用,越來越多的三維點云數(shù)據(jù)被獲取和積累下來。隨著點云數(shù)據(jù)量的不斷增大,大規(guī)模三維激光掃描點云的實時渲染已經(jīng)成為點云數(shù)據(jù)處理的瓶頸問題。本文提出一種支持大規(guī)模點云實時渲染的技術(shù)方法,主要成果如下:根據(jù)用RTK測得的掃描站點和靶標的真實地理坐標與其對應(yīng)的在激光掃描儀獨立坐標系下的相對坐標,使用坐標轉(zhuǎn)換矩陣經(jīng)旋轉(zhuǎn)和平移將點云數(shù)據(jù)全部統(tǒng)一到真實地理坐標系中,然后通過配準完成多站地面三維激光掃描點云的融合。研究了點云數(shù)據(jù)的數(shù)據(jù)特征,采用"分之而治"的思想,使用C++語言實現(xiàn)了對大規(guī)模點云數(shù)據(jù)的分塊,每個分塊包含一定數(shù)量的掃描點,可直接讀入內(nèi)存進行處理。對于每個點云分塊,構(gòu)建kd tree,使用PCL濾波以最近鄰點數(shù)為閾值濾除點云中的孤立點。遍歷經(jīng)過去噪的點云分塊,統(tǒng)計測區(qū)覆蓋的空間范圍。在外存儲器上建立不同細節(jié)層次的八叉樹索引結(jié)構(gòu),對海量點云數(shù)據(jù)進行有效的組織。通過與商業(yè)軟件的對比渲染實驗,驗證了本文方法的可行性。最終的實驗結(jié)果表明,本文提出的方法支持對超過內(nèi)存容量的三維點云數(shù)據(jù)進行實時渲染。該方法有望用于大規(guī)模三維激光掃描點云數(shù)據(jù)的可視化和交互式處理。
[Abstract]:The interactive processing of 3 D point cloud data usually depends on 3 D real time rendering of point cloud data. With the further application of 3D laser scanning technology, more and more 3D point cloud data have been acquired and accumulated. With the increasing of point cloud data, the real-time rendering of large scale 3D laser scanning point cloud has become the bottleneck of point cloud data processing. This paper presents a technical method to support large-scale point cloud real-time rendering. The main results are as follows: according to the real geographical coordinates of scanning stations and targets measured by RTK and their corresponding relative coordinates in the independent coordinate system of laser scanner, The point cloud data are all unified into the real geographical coordinate system by rotation and translation of coordinate transformation matrix, and then the fusion of multi-station 3D laser scanning point cloud is completed by registration. In this paper, the data characteristics of point cloud data are studied, and the idea of "point and rule" is adopted, and the block of large scale point cloud data is realized by C language. Each block contains a certain number of scanning points, which can be read directly into memory for processing. For each point cloud block, kd tree is constructed, and PCL filter is used to filter the outliers in the point cloud with the nearest neighbor points as the threshold. Traversing the de-noised point cloud into blocks, statistics the coverage of the spatial range of the area. The octree index structure with different detail levels is built on the external memory to organize the massive point cloud data effectively. The feasibility of this method is verified by comparison with commercial software. The experimental results show that the proposed method can render 3D point cloud data over memory capacity in real time. This method is expected to be applied to the visualization and interactive processing of large-scale 3D laser scanning point cloud data.
【學位授予單位】：天津師范大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：P225.2

【參考文獻】

相關(guān)期刊論文 前7條

1 李廣云;李明磊;王力;楊凡;王瑞鵬;;地面激光掃描點云數(shù)據(jù)預(yù)處理綜述[J];測繪通報;2015年11期

2 陳馳;王珂;徐文學;彭向陽;麥曉明;楊必勝;;海量車載激光掃描點云數(shù)據(jù)的快速可視化方法[J];武漢大學學報(信息科學版);2015年09期

3 王向玉;謝東輝;汪艷;陳一銘;漆建波;閻廣建;張吳明;;基于地面激光雷達點云數(shù)據(jù)的單木三維重建[J];遙感技術(shù)與應(yīng)用;2015年03期

4 張毅;閆利;楊紅;陳向陽;;地面三維激光掃描的系統(tǒng)誤差模型研究[J];測繪通報;2012年01期

5 張會霞;;基于八叉樹的點云數(shù)據(jù)的組織與可視化[J];太原師范學院學報(自然科學版);2011年03期

6 路明月;何永健;;三維海量點云數(shù)據(jù)的組織與索引方法[J];地球信息科學;2008年02期

7 陳靜,李清泉,李必軍;激光掃描測量系統(tǒng)的應(yīng)用研究[J];測繪工程;2001年01期

，

本文編號：1853024