Parallelization of lane detection algorithm using OpenCL

• Main Author: Kalay, Harshal A.
• Language: EN
• Published: California State University, Long Beach 2015
• Subjects: Computer science
• Online Access: http://pqdtopen.proquest.com/#viewpdf?dispub=1603755

<p> In this thesis we explore and experiment on using OpenCL for the parallelization of an important computer vision problem of lane detection. Lane detection aims at identifying lane markings on a road and has applications in autonomous vehicles as well as for providing guidance to the drivers. The parallelization is implemented using OpenCL on Graphics Processing Units (GPUs) as well as on multi-core CPU, both these platforms are supported by OpenCL library for parallel programming. </p><p> Our study aims at finding an effective way to parallelize the lane detection using OpenCL through experimentations. Lane detection involves use of image processing algorithms and computer vision techniques, which are both often parallelizable and may benefit greatly by using OpenCL. With our hardware configuration, we are able to achieve eight, four and six times the speedup on multi-core CPU, PCIe based GPU and CPU integrated GPU systems respectively, when compared to sequential C++ program.</p>