An anchor-free 3D object detection method with camera and liDAR based on H23D R-CNN

Xingyu Li; Jianming Hu; Yuang Zhang; Hantao Liu; Lihui Peng

doi:10.1117/12.2681040

23 May 2023 An anchor-free 3D object detection method with camera and liDAR based on H²3D R-CNN

Xingyu Li, Jianming Hu, Yuang Zhang, Hantao Liu, Lihui Peng

Proceedings Volume 12645, International Conference on Computer, Artificial Intelligence, and Control Engineering (CAICE 2023); 1264558 (2023) https://doi.org/10.1117/12.2681040
Event: International Conference on Computer, Artificial Intelligence, and Control Engineering (CAICE 2023), 2023, Hangzhou, China

Abstract

As an important part of automatic driving technology, the performance of perception determines the safety and availability of automatic driving. Generally, we need specific three-dimensional target information, which is difficult to gain from the camera. As a result, LiDAR plays a significant role in 3D detection due to its advantages of high resolution, immunity and capability to obtain specific location information. In order to overcome the problem that LiDAR often fails to detect objects with similar structures, we propose a method with Camera and LiDAR model based on H²3D RCNN, which can be divided into three sections. Firstly, a Foreground View (FV) map and a Bird Eye’s View (BEV) map are transformed from original point cloud, and the FV features, and image feature will augment the expression ability of features from the BEV maps and point cloud. Then, we introduce a center-based detection head on the BEV maps to generate the heatmaps and box refinement to minimize the redundancy and accelerate the operation of the model. Finally, we present the result of the second stage on the point features to acquire more accurate 3D bounding boxes. To validate our model, we test it on the official KITTI dataset and get a better performance compared to the original method and also get a nearly 5ms speed up.

Citation Download Citation

Xingyu Li, Jianming Hu, Yuang Zhang, Hantao Liu, and Lihui Peng "An anchor-free 3D object detection method with camera and liDAR based on H²3D R-CNN", Proc. SPIE 12645, International Conference on Computer, Artificial Intelligence, and Control Engineering (CAICE 2023), 1264558 (23 May 2023); https://doi.org/10.1117/12.2681040

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available