基于激光点云的智能挖掘机目标识别  

Target Recognition for Intelligent Excavator Based on Laser Point Cloud

作　　者：朱建新[1] 沈东羽 吴钪 

机构地区：[1]中南大学机电工程学院,长沙410083 [2]山河智能装备股份有限公司,长沙410100

出　　处：《计算机工程》2017年第1期297-302,共6页Computer Engineering

基　　金：国家"十二五"科技支撑计划项目(2013BAF07B02)

摘　　要：传统智能工程机械的环境目标识别方法为单目或双目视觉识别,识别速度慢、效率低且工况适应能力差。为进一步提升挖掘机的环境目标识别能力,提出一种基于点云聚类特征直方图的目标识别方法。对原始点云进行滤波预处理,通过聚类分离取得单个识别聚类,建立待识别聚类的点云特征直方图,在模型库中采用近邻搜索算法获得k个近邻,并根据其匹配度得到最终识别结果。实验结果表明,该方法针对挖掘机作业工况目标识别有较强的稳健性,能在复杂工况下识别出多个目标且识别率高。Traditional intelligent engineering machinery uses monocular or binocular recognition for environmental target recognition,which is inefficient and has low recognition rate and poor adaptation to the environment.A target recognition method based on point cloud clustering feature histogram is proposed in order to improve the ability of environmental target recognition for the excavator.After processing the raw point cloud data by filtering algorithm,the data is split into several single clusters using clustering algorithm.Point cloud characteristics histogram of be recognited cluster is built.In the model library,k neighbor is got by using neighbor search algorithm,and the final recognition result is got according to matching degree.The experimental results show that the method works with strong robustness for target recognition of intelligent excavator.It can identify multiple targets in complex conditions and has high recognition rate.

关 键 词：智能挖掘机 激光扫描 点云数据 特征直方图 目标识别 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]