An Encoder-Sequencer-Decoder Network for Lane Detection to Facilitate Autonomous Driving

Lane detection in all weather conditions is a pressing necessity for autonomous driving. Accurate lane detection ensures the safe operation of autonomous vehicles, enabling advanced driver assistance systems to effectively track and maintain the vehicle within the lanes. Traditional lane detection t...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:International Conference on Control, Automation and Systems (Online) S. 899 - 904
Hauptverfasser: Hussain, Muhammad Ishfaq, Rafique, Muhammad Aasim, Ko, Yeongmin, Khan, Zafran, Olimov, Farrukh, Naz, Zubia, Kim, Jeongbae, Jeon, Moongu
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: ICROS 17.10.2023
Schlagworte:
Autonomous Driving and Robotics
Convolutional LSTM
Decoding
Encoder and Decoder Network
Feature extraction
Lane detection
Process control
Roads
Semantic segmentation
Streaming media
TuSimple
ISSN:2642-3901
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Lane detection in all weather conditions is a pressing necessity for autonomous driving. Accurate lane detection ensures the safe operation of autonomous vehicles, enabling advanced driver assistance systems to effectively track and maintain the vehicle within the lanes. Traditional lane detection techniques heavily rely on a single image frame captured by the camera, posing limitations. Moreover, these conventional methods demand a constant stream of pristine images for uninterrupted lane detection, resulting in degraded performance when faced with challenges such as low brightness, shadows, occlusions, and deteriorating environmental conditions. Recognizing that continuous sequence patterns on the road represent lanes, our approach leverages a sequential model to process multiple images for lane detection. In this study, we propose a deep neural network model to extract crucial lane information from a sequence of images. Our model adopts a convolutional neural network in an encoder/decoder architecture and incorporates an extended short-term memory model for sequential feature extraction. We evaluate the performance of our proposed model using the TuSimple and CuLane datasets, showcasing its superiority across various lane detection scenarios. Comparative analysis with state-of-the-art lane detection methods further substantiates our model's effectiveness.
ISSN:2642-3901
DOI:10.23919/ICCAS59377.2023.10316884