摘要
针对二维激光雷达在检测行人位置时,因缺少高度信息导致对行人特征识别不准确,易产生误判的问题,提出一种基于二维激光雷达与深度相机联合检测的行人识别方法。使用支持向量机(SVM)筛选出激光雷达点云数据中属于行人腿部曲线的点云段并以此确定行人位置;视觉图像中引入人体识别算法框选行人,利用人体外边框中心点处图像坐标和深度值解算行人位置。将雷达与相机获得的行人位置信息加权融合得出行人实际位置。实验结果表明,所提方法在继承激光雷达测量精度的同时,极大地减小了误判率,证明该方法的有效性。
In response to the inaccuracy of pedestrian feature recognition and the potential for misjudgment due to the lack of height information in detecting pedestrian positions using 2D LiDAR,a method for pedestrian identification based on the joint detection of 2D LiDAR and depth camera is proposed in this paper.Firstly,a Support Vector Machine(SVM)is used to filter out point cloud segments belonging to the pedestrian's leg curve in the LiDAR point cloud data,and determine the pedestrian's position.Additionally,a human body recognition algorithm is applied to the visual images to delineate pedestrians,utilizing the image coordinates of the center point of the outer body box and depth values to calculate pedestrian positions.Finally,the actual pedestrian position is derived through weighted fusion of the pedestrian position information obtained from both LiDAR and the camera.Th experimental results demonstrate that the proposed method greatly reduces the false judgement rate while inheriting the measurement accuracy of Li-DAR,which proves the effectiveness of the method.
作者
李立凡
曹鹏彬
杜兵
沈琼霞
LI Li-fan;CAO Peng-bin;DU Bing;SHEN Qiong-xia(School of Mechanical&Electrical Engineering,Wuhan Institute of Technology,Wuhan 430205,China;Wuhan Fonsview Technologies Co.,Ltd.,Wuhan 430000,China;Fiberhome Communication Technology Co.,Ltd,Wuhan 430205,China)
出处
《激光与红外》
CAS
CSCD
北大核心
2024年第10期1547-1553,共7页
Laser & Infrared
基金
武汉市重点研发计划项目(No.2022012202015069)资助。
关键词
激光雷达
点云处理
联合标定
行人检测
LiDAR
point cloud processing
joint calibration
pedestrian detection
