3D point cloud object detection algorithm based on Transformer

基于Transformer的3D点云目标检测算法

¹ Shenyang Aircraft Design Research Institute, Shenyang 110035, China
² School of Electronics and Information, Northwestern Polytechnical University, Xi'an 710072, China
³ CSSC Systems Engineering Research Institute, Beijing 100094, China
⁴ No. 1 Military Representative Office of Equipment Department of PLA Airforce in Shenyang, Shenyang 110850, China

Received: 9 January 2023

Abstract

In response to the difficulty in deploying anchor box based methods in 3D object detection due to the increase in spatial dimensions, this paper studies a point cloud object detection algorithm based on set prediction. This article proposes a Transformer based 3D point cloud object detection algorithm, and combines the characteristics of point clouds in autonomous driving scenarios to propose an improved spatial modulation attention and heat map initialization strategy for training acceleration and query initialization, achieving good detection performance in shallow networks. This article compares it with other algorithms on the KITTI dataset, and the results show that our algorithm has reached an advanced level in performance. We also conducted ablation experiments on the main components of the algorithm to verify the contribution of each module to the detection effect.

摘要

针对在三维目标检测中由于空间维度的增加基于锚框的方法难以部署的问题, 研究了基于集合预测的点云目标检测算法。提出一种基于Transformer的3D点云目标检测算法, 并结合自动驾驶场景下的点云特点, 提出了改进空间调制注意力和热图初始化策略进行训练加速和查询初始化, 在浅层网络下取得了良好的检测性能。在KITTI数据集上与其他算法进行比较, 结果表明所提算法在性能上已经达到先进水平, 进一步对算法中的主要组成部分进行了消融实验, 验证了各个模块对检测效果的贡献。