Road mark detection based on local point cloud projection and R-FCN

Tianxiong Lv; Huimin Ma

doi:10.1117/12.2524376

6 May 2019 Road mark detection based on local point cloud projection and R-FCN

Tianxiong Lv, Huimin Ma

Proceedings Volume 11069, Tenth International Conference on Graphics and Image Processing (ICGIP 2018); 110693S (2019) https://doi.org/10.1117/12.2524376
Event: Tenth International Conference on Graphic and Image Processing (ICGIP 2018), 2018, Chengdu, China

Abstract

For the road mark in the presence of abrasion, adhesion, and occlusion, it is difficult to accurately locate the road mark by typical point cloud clustering and segmentation method. An automatic road mark detection method based on local point cloud projection and 2D deep learning object detection is proposed. Firstly, the local ground point cloud regions are obtained by prior information; Secondly, we perform orthogonal projection on the point cloud to project the point cloud into two-dimensional image by affine transformation and use R-FCN detection method to detect the road mark in the image. The holes on the image are filtered by local maximum filtering. A line-by-line search strategy based on maximum reflection rate is proposed to refine the detected bounding boxes to improve the detection accuracy; Finally, we implement inverse affine transformation and local coordinate search to restore the road mark coordinates from twodimensional image to 3D point cloud. In the point cloud dataset collected from China highway, our experimental results show that the proposed method can achieve road mark detection under complex scenes such as occlusion, abrasion and adhesion. Comparing with other methods, the detection recall, accuracy and efficiency have been improved greatly.

Citation Download Citation

Tianxiong Lv and Huimin Ma "Road mark detection based on local point cloud projection and R-FCN", Proc. SPIE 11069, Tenth International Conference on Graphics and Image Processing (ICGIP 2018), 110693S (6 May 2019); https://doi.org/10.1117/12.2524376

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