本文为法国勃艮第大学（作者：Quoc Hung TRUONG）的毕业论文，共143页。

通过捕获三维数字数据对真实场景进行建模，已被证明在各种工业和测量应用中既有用又适用。整个场景通常由激光扫描仪捕捉，并由庞大的无组织点云表示，可能还有额外的摄影测量数据。处理这些点云和数据的一个典型挑战在于检测、分类场景中的对象。除了存在噪声、遮挡和丢失数据外，这些任务通常还受到同一数据集内和从一个数据集到另一个数据集的捕获条件的不规则性的阻碍。鉴于潜在问题的复杂性，最新的处理方法试图利用语义知识来识别和分类这些对象。

在本论文中，我们提出一种新的方法，利用智慧的知识管理策略来处理三维点云，以及识别和分类数位化场景中的物件。我们的方法是将语义知识的使用扩展到处理的所有阶段，包括单个数据驱动处理算法的指导。完整的解决方案包含基于三个因素的多阶段迭代概念：建模知识、算法包和分类引擎。本文的目标是选择和指导算法，采用自适应和智能的策略来检测点云中的目标。两个案例的实验证明了我们方法的适用性。研究人员对机场候机区和铁路沿线进行了扫描，在这两种情况下，目标都是检测并识别定义区域内的对象。结果表明，我们的方法在使用不同数据类型时成功地识别了感兴趣的对象。

The modeling of real-world scenes throughcapturing 3D digital data has proven to be both useful and applicable in avariety of industrial and surveying applications. Entire scenes are generallycaptured by laser scanners and represented by large unorganized point cloudspossibly along with additional photogrammetric data. A typical challenge inprocessing such point clouds and data lies in detecting and classifying objectsthat are present in the scene. In addition to the presence of noise, occlusionsand missing data, such tasks are often hindered by the irregularity of thecapturing conditions both within the same dataset and from one data set toanother. Given the complexity of the underlying problems, recent processingapproaches attempt to exploit semantic knowledge for identifying andclassifying objects. In the present thesis, we propose a novel approach thatmakes use of intelligent knowledge management strategies for processing of 3Dpoint clouds as well as identifying and classifying objects in digitizedscenes. Our approach extends the use of semantic knowledge to all stages of theprocessing, including the guidance of the individual data-driven processingalgorithms. The complete solution consists in a multi-stage iterative conceptbased on three factors: the modeled knowledge, the package of algorithms, and aclassification engine. The goal of the present work is to select and guidealgorithms following an adaptive and intelligent strategy for detecting objectsin point clouds. Experiments with two case studies demonstrate theapplicability of our approach. The studies were carried out on scans of thewaiting area of an airport and along the tracks of a railway. In both cases thegoal was to detect and identify objects within a defined area. Results showthat our approach succeeded in identifying the objects of interest while usingvarious data types.

1. 引言

1.1 项目背景与动机

1.2 论文研究范围

1.3 本文贡献

1.4 本文概述

2. 文献回顾

2.1 模型驱动方法

2.2 纯数据驱动方法

2.3 智能数据驱动方法

2.4 引入语义的数据驱动

2.5 基于知识的方法

3. 项目背景

3.1 语义知识

3.2 数字处理

4. 研究方法

4.1 系统概述

4.2 基于知识的工程

4.3 数字处理

4.4 算法选择模块ASM

4.5 集成知识信息到3D处理

4.6 集成知识信息的处理技术

5. 具体实现

5.1 铁路系统中的对象分类

5.2 机场大楼内的目标探测（弗拉波特候机区）

6. 结论与未来工作展望

6.1 结果

6.2 未来工作展望

更多精彩文章请关注公众号：