摘要
基于模糊模型和时空关联分析,提出一种在激光雷达获取的点云数据中寻找可通行区域的方法。首先使用基于最大熵的模糊预测模型将激光雷达获得的单条扫描线数据聚类分段;然后寻找符合可通行特征的分段作为初始可通行区域;最后利用时空关联模型在多条扫描线数据中进一步优化可通行区域提取结果。算法通过利用不同方向三维到二维的映射避免了三维空间中计算量大的缺陷,同时又较好地保留了三维数据包含的信息,不涉及迭代计算和在线机器学习过程。实验表明,在城市环境中,算法能够较好的提取可通行区域。
An algorithm to detect traversable area from ladar point cloud based on fuzzy model and space-time analysis is proposed. The algorithm splits a single row ladar da ta into different segments by using a maximum entropy theory based fuzz cluster model, and then finds initial traversable segments constrained by traversable f eatures. The algorithm optimizes traversable area according to space-time conne ction between adjacent scan lines. Data in 3-D space were projected into 2-D f rom different directions to avoid time consuming computing in 3-D space. The al gorithm doesn’t involve iterative and on line learning which makes it suitable for urban navigation. Experiments in urban area shows that the algorithm can extrac t traversable area well.
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2010年第12期1702-1707,共6页
Acta Armamentarii
关键词
信息处理技术
激光雷达
点云
可通行区域
模糊聚类
时空关联
information processing
ladar
point cloud
traversable area
fuzz cluster
space-ti me association
