摘要
在复杂场景下开展目标形态识别时,由于三维激光点云数据视角不同,会遗失部分激光点云数据,造成目标形态识别不准确的问题。为有效解决该问题,提出复杂场景下三维激光点云数据融合的目标形态识别方法。利用三维激光扫描设备获取目标三维激光点云数据,提取点云数据关键点,并实现点云配准、融合处理,生成一个更加完整的复杂场景点云数据。再根据完整三维激光点云获取复杂场景点云邻域点集,对其开展主成分分析,联合采样点曲率以及法向夹角双参数提取复杂场景目标的目标边界特征点。最后通过特征点构建链码序列,并对序列点连接完成目标特征线的提取,从而实现目标形态的精准识别。实验结果表明,利用该方法开展复杂场景目标形态识别时,识别精度高、效果好。
When conducting target shape recognition in complex scenes,due to the different perspectives of 3D laser point cloud data,some laser point cloud data may be lost,resulting in inaccurate target shape recognition.To effectively solve this problem,a target morphology recognition method for 3D laser point cloud data fusion in complex scenes is proposed.Using 3D laser scanning equipment to obtain target 3D laser point cloud data,extracting key points from the point cloud data,and achieving point cloud registration and fusion processing to generate a more complete and complex scene point cloud data.Then,based on the complete 3D laser point cloud,a complex scene point cloud neighborhood point set is obtained,and principal component analysis is carried out to extract target boundary feature points of complex scene targets by jointly sampling point curvature and normal angle dual parameters.Finally,a chain code sequence is constructed through feature points,and the target feature lines are extracted by connecting the sequence points,thereby achieving accurate recognition of the target shape.The experimental results show that when using this method for complex scene target shape recognition,the recognition accuracy is high and the effect is good.
作者
李福琳
苏变萍
徐娟娟
LI Fulin;SU Bianping;XU Juanjuan(Guangdong University of Science and Technology Faculty of Computer Science,Dongguan Guangdong 523000,China)
出处
《激光杂志》
北大核心
2025年第8期234-239,共6页
Laser Journal
基金
国家自然科学基金青年基金项目(No.12001418)。
关键词
复杂场景
三维激光点云数据
目标形态识别
特征点提取
轮廓线获取
complex scenes
3D laser point cloud data
target morphology recognition
feature point extraction
obtaining contour lines
