motion estimation in unstructured and dynamic environments because
they directly use raw laser data. Exact point correspondence is im-
possible due to sensor limitations, so matching is usually regarded as
an optimization problem where the maximum expected precision is in-
trinsically limited by the working environment and by the rangefinder
performance.
Different optimization-based techniques have been proposed for point
matching methods, such as a direct-descent technique denominated Iterative
Closest Point (ICP) (Chen and Medioni, 1991), (Besl and McKay, 1992),
(Zhang, 1992), Gradient Computation (G C) (Neugebauer, 1997), (Fitzgib-
bon, 2001), (Thrun et al., 2000), Evolutionary Programming (EP) (Agrawal
et al., 1994), Simulated Annealing (SA) (Blais and Levine, 1995), and, re-
cently, a Genetic Algorithm (GA) (Mart´ınez, 2003). Among them, the ICP
procedure is the most popular because of its simplicity and effectiveness, and
many variants have been proposed (Rusinkiewicz and Levoy, 2001). GC is
based on calculating gradients of either probabilistic (Thrun et al., 2000)
or squared-error (Neugebauer, 1997), (Fitzgibbon, 2001) cost functions. Fi-
nally, the EP, SA and GA methods introduce a stochastic component in the
search of either the correspondence of points (Agrawal et al., 1994) or the
whole pose (Blais and Levine, 1995), (Mart´ınez, 2003).
Point correspondences can be computed either in polar (Neugebauer,
1997), (Thrun et al., 2000), (Blais and Levine, 1995), (Mart´ınez, 2003) or
cartesian (Chen and Medioni, 1991), (Besl and McKay, 1992), (Zhang, 1992),
(Fitzgibbon, 2001), (Agrawal et al., 1994) coordinates. The latter strategy
can be accelerated for large amounts of data by using closest point caching or
multi-dimensional binary search trees (Zhang, 1992), which use information
from previous matching tentatives. Moreover, unmatchable points can dis-
rupt the optimization process if included in the cost function, so an adaptive
statistical threshold can be used to discard outliers (Zhang, 1992), (Fitzgib-
bon, 2001).
In this paper, we have analyzed two essentially different point matching
techniques for motion estimation: a fast 2D version of ICP, a local-scope
optimizer which is the most widely used registration algorithm for 3D sur-
faces, and the GA method, a wide-scope technique which constitutes a novel
stochastic approach for scan matching. Then, we propose a hybrid GA-ICP
algorithm that combines the best characteristics of these methods. Com-
parative exp erimental results have been obtained with the tracked vehicle
Auriga-α equipped with a commercial scanner.
The rest of the paper is organized as follows. In the next section, we
enunciate the 2D laser point matching problem, we review the ICP and GA
3
剩余24页未读，继续阅读
白菜888
- 粉丝: 7
- 资源: 74
我的内容管理
展开
最新资源
- 构建Cadence PSpice仿真模型库教程
- VMware 10.0安装指南：步骤详解与网络、文件共享解决方案
- 中国互联网20周年必读：影响行业的100本经典书籍
- SQL Server 2000 Analysis Services的经典MDX查询示例
- VC6.0 MFC操作Excel教程：亲测Win7下的应用与保存技巧
- 使用Python NetworkX处理网络图
- 科技驱动：计算机控制技术的革新与应用
- MF-1型机器人硬件与robobasic编程详解
- ADC性能指标解析：超越位数、SNR和谐波
- 通用示波器改造为逻辑分析仪：0-1字符显示与电路设计
- C++实现TCP控制台客户端
- SOA架构下ESB在卷烟厂的信息整合与决策支持
- 三维人脸识别：技术进展与应用解析
- 单张人脸图像的眼镜边框自动去除方法
- C语言绘制图形：余弦曲线与正弦函数示例
- Matlab 文件操作入门：fopen、fclose、fprintf、fscanf 等函数使用详解
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
