Title: Front vehicle detection based on parallel Blob detection using quad-pipeline on FPGA

Authors: Naoto Nojiri; Seben Masaki; Lin Meng; Hiromitsu Shimakawa; Katsuhiro Yamazaki

Addresses: College of Information Science and Engineering, Ritsumeikan University, Nojihigashi 1-1-1, Kusatsu, Shiga, Japan ' Department of Electronic and Computer Engineering, Ritsumeikan University, Nojihigashi 1-1-1, Kusatsu, Shiga, Japan ' Department of Electronic and Computer Engineering, Ritsumeikan University, Nojihigashi 1-1-1, Kusatsu, Shiga, Japan ' College of Information Science and Engineering, Ritsumeikan University, Nojihigashi 1-1-1, Kusatsu, Shiga, Japan ' Department of Electronic and Computer Engineering, Ritsumeikan University, Nojihigashi 1-1-1, Kusatsu, Shiga, Japan

Abstract: Real-time front vehicle detection is one of the important topics in intelligent transport systems. FPGA is an effective hardware device for high performance computing. This paper describes real-time front vehicle detection using quad-pipeline image processing on FPGA for intelligent transport systems. Front vehicle detection is implemented based on tail lights detection and white lines detection. When the two red Blobs of tail lights are detected between the two white line Blobs, the front vehicle is determined. For achieving real-time image processing on FPGA, the image is divided into four parts and the four image processing are realized by four-stage-pipeline. In detail, tail lights are detected using two pipelines, and white lines are detected using the other two pipelines. The experimental results show that the quad-pipeline system can detect a vehicle up ahead in 124.8 μs on Artix7, which is 1.91 times faster than the dual-pipeline one.

Keywords: FPGA; front vehicle detection; quad-pipeline; scalability.

DOI: 10.1504/IJMIC.2019.103981

International Journal of Modelling, Identification and Control, 2019 Vol.33 No.1, pp.20 - 27

Received: 19 Jun 2018
Accepted: 24 Apr 2019
Published online: 04 Dec 2019 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article