汽车工程

汽车工程

• •    下一篇

基于KD树和欧式聚类的越野环境人员识别研究

范晶晶1 ，王力1，褚文博2 ，罗禹贡3   

  1. 1. 北方工业大学城市道路交通智能控制技术北京重点实验室，北京市，100144； 2. 国汽（北京）智能网联汽车研究院有限公司，北京市，100176 3. 清华大学汽车安全与节能国家重点实验室，北京市，100084
  • 出版日期:2019-03-12 发布日期:2019-03-12
  • 基金资助:
     

Research of person recognition in cross-country environment based on KDTree and euclidean clustering

    

  1.  
  • Online:2019-03-12 Published:2019-03-12
  • Supported by:
     

赞

摘要： 越野环境下人员的识别，是班组伴随自动驾驶车辆的基础要求，其识别问题备受重视，建立完善的越野环境人员识别算法迫在眉睫。本文针对激光雷达点云数据中的行人识别问题，特别是越野环境的特殊问题，提出基于聚类思想的解决方案。在理论分析的基础上，结合人的几何物理特征，设计基于KDTree和欧式聚类的人员识别算法，并在越野环境下履带式车辆平台上进行实验。实验结果表明，所设计的激光雷达人员识别算法能够准确识别激光雷达点云数据中的人员，在越野环境下有良好的识别率。

关键词: 激光雷达, 聚类, 行人识别

Abstract: Recognition of person in cross-country environment is the basic requirement of the group accompanied automatic driving vehicle. The recognition problem has been paid much attention. It is imminent to establish a perfect algorithm for the identification of person in cross-country environment. Aiming at the problem of pedestrian recognition in LIDAR point cloud data, especially for special problems in cross-country environment, a solution based on clustering idea is proposed. On the basis of theoretical analysis, a pedestrian recognition algorithm based on KDTree and Euclidean clustering is designed in combination with the geometric and physical characteristics of pedestrians, and the experiment is carried out on tracked vehicle platform under off-road environment. The experimental results show that the designed pedestrian recognition algorithm of lidar can accurately identify the pedestrians in the LIDAR point cloud data, and have a good recognition rate in the cross-country environment.

Key words: Lidar, Clustering, Pedestrian Recognition

中图分类号: 

  •