使用场景流对点云序列的鸟瞰视图表示进行对齐（英文版）

英文标题：Aligning Bird-Eye View Representation of Point Cloud Sequences using Scene Flow中文摘要：研究了利用场景流纠正车辆自运动补偿方法（EMC）中所存在的影子效应， 并提出了一种插件模块，该模块使用场景流来纠正 3D 目标检测的鸟瞰图表示，实验表明，该模块可显著提高大型车辆的平均精度（高达 16％）。英文摘要：Low-resolution point clouds are challenging for object detection methods dueto their sparsity. Densifying the present point cloud by concatenating it withits predecessors is a popular solution to this challenge. Such concatenation ispossible thanks to the removal of ego vehicle motion using its odomet