Learnable Resized and Laplacian-Filtered U-Net: Better Road Marking Extraction and Classification on Sparse-Point-Cloud-Derived Imagery

High-definition (HD) maps for autonomous driving rely on data from mobile mapping systems (MMS), but the high cost of MMS sensors has led researchers to explore cheaper alternatives like low-cost LiDAR sensors. While cost effective, these sensors produce sparser point clouds, leading to poor feature...

Full description

Saved in:
Bibliographic Details
Main Authors: Miguel Luis Rivera Lagahit, Xin Liu, Haoyi Xiu, Taehoon Kim, Kyoung-Sook Kim, Masashi Matsuoka
Format: Article
Language:English
Published: MDPI AG 2024-12-01
Series:Remote Sensing
Subjects:
Online Access:https://www.mdpi.com/2072-4292/16/23/4592
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850106443531812864
author Miguel Luis Rivera Lagahit
Xin Liu
Haoyi Xiu
Taehoon Kim
Kyoung-Sook Kim
Masashi Matsuoka
author_facet Miguel Luis Rivera Lagahit
Xin Liu
Haoyi Xiu
Taehoon Kim
Kyoung-Sook Kim
Masashi Matsuoka
author_sort Miguel Luis Rivera Lagahit
collection DOAJ
description High-definition (HD) maps for autonomous driving rely on data from mobile mapping systems (MMS), but the high cost of MMS sensors has led researchers to explore cheaper alternatives like low-cost LiDAR sensors. While cost effective, these sensors produce sparser point clouds, leading to poor feature representation and degraded performance in deep learning techniques, such as convolutional neural networks (CNN), for tasks like road marking extraction and classification, which are essential for HD map generation. Examining common image segmentation workflows and the structure of U-Net, a CNN, reveals a source of performance loss in the succession of resizing operations, which further diminishes the already poorly represented features. Addressing this, we propose improving U-Net’s ability to extract and classify road markings from sparse-point-cloud-derived images by introducing a learnable resizer (LR) at the input stage and learnable resizer blocks (LRBs) throughout the network, thereby mitigating feature and localization degradation from resizing operations in the deep learning framework. Additionally, we incorporate Laplacian filters (LFs) to better manage activations along feature boundaries. Our analysis demonstrates significant improvements, with F1-scores increasing from below 20% to above 75%, showing the effectiveness of our approach in improving road marking extraction and classification from sparse-point-cloud-derived imagery.
format Article
id doaj-art-ee5fc717ee204e9da4310bfcc7029111
institution OA Journals
issn 2072-4292
language English
publishDate 2024-12-01
publisher MDPI AG
record_format Article
series Remote Sensing
spelling doaj-art-ee5fc717ee204e9da4310bfcc70291112025-08-20T02:38:50ZengMDPI AGRemote Sensing2072-42922024-12-011623459210.3390/rs16234592Learnable Resized and Laplacian-Filtered U-Net: Better Road Marking Extraction and Classification on Sparse-Point-Cloud-Derived ImageryMiguel Luis Rivera Lagahit0Xin Liu1Haoyi Xiu2Taehoon Kim3Kyoung-Sook Kim4Masashi Matsuoka5Department of Architecture and Building Engineering, Institute of Science Tokyo, Tokyo 152-8552, JapanArtificial Intelligence Research Center, National Institute of Advanced Industrial Science and Technology, Tokyo 135-0064, JapanArtificial Intelligence Research Center, National Institute of Advanced Industrial Science and Technology, Tokyo 135-0064, JapanArtificial Intelligence Research Center, National Institute of Advanced Industrial Science and Technology, Tokyo 135-0064, JapanArtificial Intelligence Research Center, National Institute of Advanced Industrial Science and Technology, Tokyo 135-0064, JapanDepartment of Architecture and Building Engineering, Institute of Science Tokyo, Tokyo 152-8552, JapanHigh-definition (HD) maps for autonomous driving rely on data from mobile mapping systems (MMS), but the high cost of MMS sensors has led researchers to explore cheaper alternatives like low-cost LiDAR sensors. While cost effective, these sensors produce sparser point clouds, leading to poor feature representation and degraded performance in deep learning techniques, such as convolutional neural networks (CNN), for tasks like road marking extraction and classification, which are essential for HD map generation. Examining common image segmentation workflows and the structure of U-Net, a CNN, reveals a source of performance loss in the succession of resizing operations, which further diminishes the already poorly represented features. Addressing this, we propose improving U-Net’s ability to extract and classify road markings from sparse-point-cloud-derived images by introducing a learnable resizer (LR) at the input stage and learnable resizer blocks (LRBs) throughout the network, thereby mitigating feature and localization degradation from resizing operations in the deep learning framework. Additionally, we incorporate Laplacian filters (LFs) to better manage activations along feature boundaries. Our analysis demonstrates significant improvements, with F1-scores increasing from below 20% to above 75%, showing the effectiveness of our approach in improving road marking extraction and classification from sparse-point-cloud-derived imagery.https://www.mdpi.com/2072-4292/16/23/4592low-cost LiDARroad marking extraction and classificationdeep learning
spellingShingle Miguel Luis Rivera Lagahit
Xin Liu
Haoyi Xiu
Taehoon Kim
Kyoung-Sook Kim
Masashi Matsuoka
Learnable Resized and Laplacian-Filtered U-Net: Better Road Marking Extraction and Classification on Sparse-Point-Cloud-Derived Imagery
Remote Sensing
low-cost LiDAR
road marking extraction and classification
deep learning
title Learnable Resized and Laplacian-Filtered U-Net: Better Road Marking Extraction and Classification on Sparse-Point-Cloud-Derived Imagery
title_full Learnable Resized and Laplacian-Filtered U-Net: Better Road Marking Extraction and Classification on Sparse-Point-Cloud-Derived Imagery
title_fullStr Learnable Resized and Laplacian-Filtered U-Net: Better Road Marking Extraction and Classification on Sparse-Point-Cloud-Derived Imagery
title_full_unstemmed Learnable Resized and Laplacian-Filtered U-Net: Better Road Marking Extraction and Classification on Sparse-Point-Cloud-Derived Imagery
title_short Learnable Resized and Laplacian-Filtered U-Net: Better Road Marking Extraction and Classification on Sparse-Point-Cloud-Derived Imagery
title_sort learnable resized and laplacian filtered u net better road marking extraction and classification on sparse point cloud derived imagery
topic low-cost LiDAR
road marking extraction and classification
deep learning
url https://www.mdpi.com/2072-4292/16/23/4592
work_keys_str_mv AT miguelluisriveralagahit learnableresizedandlaplacianfilteredunetbetterroadmarkingextractionandclassificationonsparsepointcloudderivedimagery
AT xinliu learnableresizedandlaplacianfilteredunetbetterroadmarkingextractionandclassificationonsparsepointcloudderivedimagery
AT haoyixiu learnableresizedandlaplacianfilteredunetbetterroadmarkingextractionandclassificationonsparsepointcloudderivedimagery
AT taehoonkim learnableresizedandlaplacianfilteredunetbetterroadmarkingextractionandclassificationonsparsepointcloudderivedimagery
AT kyoungsookkim learnableresizedandlaplacianfilteredunetbetterroadmarkingextractionandclassificationonsparsepointcloudderivedimagery
AT masashimatsuoka learnableresizedandlaplacianfilteredunetbetterroadmarkingextractionandclassificationonsparsepointcloudderivedimagery