摘要
行人检测作为智能汽车环境感知的一个重要部分,必须具备很好的精度和实时性。2.5D四线激光雷达能够返回比较充足的障碍物信息,相对于三维激光雷达实时性更好,可以实现高性能的行人检测。针对行人难以检测的问题提出由13个特征构成的特征集,提出基于2.5D激光雷达的行人检测方法。实验结果表明该方法对户外行人检测有很高的精度和鲁棒性。
Pedestrian detection, as an important part of the environment perception of intelligent vehicle, must have good accuracy and real-time performance. 2.5D four-layer laser radar can return plentiful information, has better real-time performance than 3D laser scanner, and can achieve high performance of pedestrian detection. Aiming at the problems that pedestrian detection is difficult, a feature set composed of 13 features and a pedestrian detection method based on 2.5D laser scanner is proposed. The experimental results show that the method of outdoor pedestrian detection has high accuracy and robustness.
作者
宋柱
赵佳
李星星
Song Zhu Zhao Jia Li Xingxing(School of Automobile, Chang'an University, Xi'an City, Shaanxi Province 710064, China)
出处
《农业装备与车辆工程》
2017年第7期19-22,共4页
Agricultural Equipment & Vehicle Engineering
关键词
无人车
2.5D激光雷达
聚类
特征提取
Adaboost分类
行人检测
unmanned vehicle
2.5D laser scanner
, clustering
, feature extraction
Adaboost classification
pedestrian detection