| Title |
Road Lane Markings Detection and Classification Design with Integration of Front Vehicle Recognition based-on Memory-centric Computing |
| Authors |
유습바에브 보보혼(Bobokhon Yusupbaev) ; 위커(Ke Yu) ; 백장우(Jang Woo Baek) ; 최준림(Jun Rim Choi) |
| DOI |
https://doi.org/10.5573/ieie.2024.61.10.18 |
| Keywords |
Memory-centric computing; Distinguishing road lines; Von Neumann; Computer vision |
| Abstract |
Classifying road lane markings is crucial for road safety and is particularly important in the development of autonomous vehicles. Traditional processor-based methods face significant challenges, such as the memory wall issue and Von Neumann bottlenecks, which lead to inefficiencies in lane detection processes. This paper proposes a memory-centric method for detecting and classifying three types of road lane markings: continuous, broken, and double. Initially, the algorithm was developed in Python using OpenCV, and subsequently, it was converted to RTL with the Xilinx Vitis High-Level Synthesis (HLS) tool. The system was then implemented on a Xilinx Alveo U50 FPGA Accelerator. As a result, an algorithm that integrates lane classification and front vehicle detection based on memory-centric computing is implemented with 70.72% lower power consumption and an improved speed of 9.7ms per frame compared with the existing CPU processing algorithm. |