摘要
针对城市道路场景复杂多变,三维激光雷达数据量较大的问题,研究将地面点云转化成多个小扇形区域进行并行处理,结合全景分割与单层光束模型进行区域检测,并采用顺序融合方式将色彩模式图像与点云图像特征进行融合。实验表明研究方法对1000帧激光点云的平均精确率为93.85%,平均召回率为95.73%。该方法对于4种道路的边界点检测精确率、召回率、F1分值最高为0.952、0.910、0.921。研究对研发更加智能且安全的车辆环境感知技术具有重要参考价值。
In view of the complex and changeable urban road scene and the large amount of three-dimensional lidar data,the transformation of ground point clouds into several small sector regions for parallel processing is studied,combined with panoramic segmentation and single-layer beam model for region detection,and adopts sequential fusion to fuse color pattern images with point cloud image features.The results show that the average accuracy and recall rate of 1000 frame laser point cloud are 93.85%and 95.73%respectively.The highest boundary point detection accuracy,recall rate and F1 score of the 4 roads were 0.952,0.910 and 0.921.The research has important reference value for developing more intelligent and safe vehicle environment sensing technology.
作者
韩丹
HAN Dan(Shaanxi Polytechnic Institute,Xianyang Shaanxi 712000,China)
出处
《自动化与仪器仪表》
2024年第2期33-37,共5页
Automation & Instrumentation
基金
陕西工业职业技术学院科研基金资助项目(2023YKYB-016)。
关键词
激光雷达
智能
汽车
环境感知
RGB
laser radar
intelligence
cars
environment perception
RGB