Open AccessLateral distance detection model based on convolutional neural network
Author(s) -
Zhang Xiang,
Yang Wei,
Tang Xiaolin,
Wang Yun
Publication year - 2019
Publication title -
iet intelligent transport systems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.579
H-Index - 45
eISSN - 1751-9578
pISSN - 1751-956X
DOI - 10.1049/iet-its.2017.0431
Subject(s) - convolutional neural network , computer science , artificial intelligence , benchmark (surveying) , deep learning , computer vision , intelligent transportation system , artificial neural network , image (mathematics) , lane departure warning system , pattern recognition (psychology) , engineering , civil engineering , geodesy , geography
To evaluate the performance of the advanced driver assistant systems, such as lane departure warning systems (LDWs) and lane keeping assist systems (LKAs), a deep learning model is proposed to estimate the lateral distance between the vehicle and lane boundaries. The training of a deep learning model requires a large number of label images, but the generation of label images is time consuming and boring. Therefore, an improved image quilting algorithm based on a convolutional neural network is proposed. A lot of lane and asphalt pavement images can be synthesised using fewer images of a real road scene. Moreover, an algorithm that aims to automatically generate label images using lane and asphalt pavement images to satisfy the distribution of real scenes is proposed. Experimental results showed that the generated label images can be used to train a deep learning model, and the lateral distance can be estimated with a sub‐centimetre precision, which can provide an effective benchmark for the road test of LDWs, LKAs and other driving assistant systems.