基于局部信息融合的点云3D目标检测算法

张林杰,柴志雷,王宁

Point cloud 3D object detection algorithm based on local information fusion

Linjie ZHANG,Zhilei CHAI,Ning WANG

表 3 Waymo验证数据集不同算法的检测结果对比

Tab.3 Comparison of detection results from different algorithms on Waymo validation dataset

方法 AP/APH (LEVEL_1) AP/APH (LEVEL_2) d = 0~30 m d = 30~50 m d > 50 m 均值 d = 0~30 m d = 30~50 m d > 50 m 均值 SECOND[20] 88.66/88.18 67.35/66.70 42.89/42.09 70.07/69.52 87.33/86.86 60.92/60.23 32.39/31.77 61.63/61.14 PV-RCNN[30] 91.30/90.56 73.00/72.31 51.35/50.34 74.70/74.09 89.75/89.29 66.32/65.68 39.27/38.46 66.05/65.50 Voxel-RCNN[9] 90.81/90.36 72.43/71.78 50.37/49.47 73.90/73.32 89.50/89.05 65.68/65.08 38.32/37.61 65.10/64.58 本文方法 91.20/90.77 73.28/72.68 52.40/51.45 74.92/74.38 89.91/89.48 66.58/66.02 40.04/39.29 66.19/65.69