Please use this identifier to cite or link to this item:
Title: Lane detection algorithm for autonomous vehicles using machine learning
Authors: Tee, Daniel Tean Ming
Keywords: Engineering::Mechanical engineering
Issue Date: 2021
Publisher: Nanyang Technological University
Source: Tee, D. T. M. (2021). Lane detection algorithm for autonomous vehicles using machine learning. Final Year Project (FYP), Nanyang Technological University, Singapore.
Project: C092
Abstract: Lane detection is an important component of any driver assistance or autopilot system. As such, it is pivotal to construct a robust lane detector capable of handling various on road conditions such as bad weather, faded lane markings, strong shadow, varied lighting and so on. Traditional technics involving image processing and model fitting showed a good ability to detect lane from its many distinctive features, however, oftentimes failed to do so in less than optimum conditions. With the advent of machine learning and improvements in computing power, self-learning algorithms have been built to handle this complex task of extracting relevant characteristics and relating information to identify lanes. Nonetheless, such models are often computationally expensive, resulting in long training and prediction time. Also, a lot of training sample is required before models can achieve and outperform traditional methods. This project seeks to improve upon previous machine learning techniques to create a model capable of detecting lanes through semantic segmentation. U-Net model was used as a benchmark and was revamped using VGG16 and MobileNetV2 as an encoder to create a 4-lane detector and classifier. Hough transform and guard zone were then introduced as post-processing operations to refine lines detected by model and allow for lane departure warning function, respectively.
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:MAE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
FYP Tee Tean Ming, Daniel.pdf
  Restricted Access
4.34 MBAdobe PDFView/Open

Page view(s)

Updated on Nov 30, 2021


Updated on Nov 30, 2021

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.