{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,23]],"date-time":"2026-03-23T14:34:27Z","timestamp":1774276467704,"version":"3.50.1"},"reference-count":64,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2023,3,3]],"date-time":"2023-03-03T00:00:00Z","timestamp":1677801600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Robot. AI"],"abstract":"<jats:p>HD-maps are one of the core components of the self-driving pipeline. Despite the effort of many companies to develop a completely independent vehicle, many state-of-the-art solutions rely on high-definition maps of the environment for localization and navigation. Nevertheless, the creation process of such maps can be complex and error-prone or expensive if performed <jats:italic>via<\/jats:italic> ad-hoc surveys. For this reason, robust automated solutions are required. One fundamental component of an high-definition map is traffic lights. In particular, traffic light detection has been a well-known problem in the autonomous driving field. Still, the focus has always been on the light state, not the features (i.e., shape, orientation, pictogram). This work presents a pipeline for lights HD-map creation designed to provide accurate georeferenced position and description of all traffic lights seen by a camera mounted on a surveying vehicle. Our algorithm considers consecutive detection of the same light and uses Kalman filtering techniques to provide each target\u2019s smoother and more precise position. Our pipeline has been validated for the detection and mapping task using the state-of-the-art dataset DriveU Traffic Light Dataset. The results show that our model is robust even with noisy GPS data. Moreover, for the detection task, we highlight how our model can correctly identify even far-away targets which are not labeled in the original dataset.<\/jats:p>","DOI":"10.3389\/frobt.2023.1065394","type":"journal-article","created":{"date-parts":[[2023,3,3]],"date-time":"2023-03-03T06:00:51Z","timestamp":1677823251000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":5,"title":["Traffic lights detection and tracking for HD map creation"],"prefix":"10.3389","volume":"10","author":[{"given":"Simone","family":"Mentasti","sequence":"first","affiliation":[]},{"given":"Yusuf Can","family":"Simsek","sequence":"additional","affiliation":[]},{"given":"Matteo","family":"Matteucci","sequence":"additional","affiliation":[]}],"member":"1965","published-online":{"date-parts":[[2023,3,3]]},"reference":[{"key":"B1","unstructured":"Altec gps2022"},{"key":"B2","first-page":"851","article-title":"Deep convolutional traffic light recognition for automated driving","author":"Bach","year":"2018"},{"key":"B3","doi-asserted-by":"crossref","DOI":"10.1109\/ICRA.2017.7989163","article-title":"A deep learning approach to traffic lights: Detection, tracking, and classification","author":"Behrendt","year":"2017"},{"key":"B4","first-page":"1370","article-title":"A deep learning approach to traffic lights: Detection, tracking, and classification","author":"Behrendt","year":"2017"},{"key":"B5","article-title":"Modelling temporal context for traffic light recognition using rnns","author":"Bj\u00f6rnsson","year":"2021"},{"key":"B6","article-title":"Yolov4: Optimal speed and accuracy of object detection","author":"Bochkovskiy","year":"2020"},{"key":"B7","doi-asserted-by":"publisher","DOI":"10.4271\/epr2021013","article-title":"Unsettled issues on HD mapping technology for autonomous driving and ADAS","author":"Bock","year":"2021","journal-title":"Tech. Rep"},{"key":"B8","first-page":"6154","article-title":"Cascade r-cnn: Delving into high quality object detection","author":"Cai","year":"2018"},{"key":"B9","first-page":"8748","article-title":"Argoverse: 3d tracking and forecasting with rich maps","author":"Chang","year":"2019"},{"key":"B10","doi-asserted-by":"publisher","first-page":"685","DOI":"10.1080\/01441647.2021.1879310","article-title":"The development of autonomous driving technology: Perspectives from patent citation analysis","volume":"41","author":"Cho","year":"2021","journal-title":"Transp. Rev."},{"key":"B11","doi-asserted-by":"publisher","first-page":"564","DOI":"10.1109\/tpami.2003.1195991","article-title":"Kernel-based object tracking","volume":"25","author":"Comaniciu","year":"2003","journal-title":"IEEE Trans. pattern analysis Mach. Intell."},{"key":"B12","doi-asserted-by":"crossref","DOI":"10.1109\/CVPR.2016.350","article-title":"The cityscapes dataset for semantic urban scene understanding","author":"Cordts","year":"2016"},{"key":"B13","first-page":"764","article-title":"Deformable convolutional networks","author":"Dai","year":"2017"},{"key":"B14","first-page":"358","article-title":"Real time visual traffic lights recognition based on spot light detection and adaptive traffic lights templates","author":"De Charette","year":""},{"key":"B15","first-page":"358","article-title":"Real time visual traffic lights recognition based on spot light detection and adaptive traffic lights templates","author":"De Charette","year":""},{"key":"B16","doi-asserted-by":"publisher","first-page":"76","DOI":"10.1016\/j.cag.2020.09.012","article-title":"Deep traffic light detection by overlaying synthetic context on arbitrary natural images","volume":"94","author":"de Mello","year":"2021","journal-title":"Comput. Graph."},{"key":"B17","first-page":"248","article-title":"Imagenet: A large-scale hierarchical image database","author":"Deng","year":"2009"},{"key":"B18","doi-asserted-by":"publisher","first-page":"13255","DOI":"10.1609\/aaai.v34i08.7033","article-title":"Automatic building and labeling of hd maps with deep learning","volume":"34","author":"Elhousni","year":"2020","journal-title":"Proc. AAAI Conf. Artif. Intell."},{"key":"B19","doi-asserted-by":"publisher","first-page":"303","DOI":"10.1007\/s11263-009-0275-4","article-title":"The pascal visual object classes (voc) challenge","volume":"88","author":"Everingham","year":"2010","journal-title":"Int. J. Comput. Vis."},{"key":"B20","first-page":"5421","article-title":"Traffic light mapping and detection","author":"Fairfield","year":"2011"},{"key":"B21","first-page":"248","article-title":"A deep analysis of the existing datasets for traffic light state recognition","author":"Fern\u00e1ndez","year":"2018"},{"key":"B22","first-page":"3376","article-title":"The driveu traffic light dataset: Introduction and comparison with existing datasets","author":"Fregin","year":""},{"key":"B23","first-page":"3376","article-title":"The driveu traffic light dataset: Introduction and comparison with existing datasets","author":"Fregin","year":""},{"key":"B24","first-page":"750","article-title":"Traffic lights detection and state estimation using hidden markov models","author":"Gomez","year":"2014"},{"key":"B25","first-page":"5356","article-title":"Lvis: A dataset for large vocabulary instance segmentation","author":"Gupta","year":"2019"},{"key":"B26","article-title":"Learning the k in k-means","volume":"16","author":"Hamerly","year":"2003","journal-title":"Adv. neural Inf. Process. Syst."},{"key":"B27","first-page":"597","article-title":"Online tracking by learning discriminative saliency map with convolutional neural network","author":"Hong","year":"2015"},{"key":"B28","doi-asserted-by":"publisher","first-page":"975","DOI":"10.3390\/rs9100975","article-title":"Towards high-definition\u00a03d urban mapping: Road feature-based registration of mobile mapping systems and aerial imagery","volume":"9","author":"Javanmardi","year":"2017","journal-title":"Remote Sens."},{"key":"B29","doi-asserted-by":"publisher","first-page":"1800","DOI":"10.1109\/tits.2015.2509509","article-title":"Vision for looking at traffic lights: Issues, survey, and perspectives","volume":"17","author":"Jensen","year":"2016","journal-title":"IEEE Trans. Intelligent Transp. Syst."},{"key":"B30","first-page":"367","article-title":"Machine learning assisted high-definition map creation","author":"Jiao","year":"2018"},{"key":"B31","volume-title":"Networking vehicles to everything. evolving automotive solutions","author":"Karls","year":"2018"},{"key":"B32","doi-asserted-by":"publisher","first-page":"83","DOI":"10.1002\/nav.3800020109","article-title":"The Hungarian method for the assignment problem","volume":"2","author":"Kuhn","year":"1955","journal-title":"Nav. Res. Logist. Q."},{"key":"B33","first-page":"163","article-title":"Towards fully autonomous driving: Systems and algorithms","author":"Levinson","year":"2011"},{"key":"B34","first-page":"2117","article-title":"Feature pyramid networks for object detection","author":"Lin","year":"2017"},{"key":"B35","doi-asserted-by":"publisher","first-page":"740","DOI":"10.1007\/978-3-319-10602-1_48","article-title":"Microsoft coco: Common objects in context","author":"Lin","year":"2014","journal-title":"Comput. Vis. \u2013 ECCV 2014, Lect. Notes Comput. Sci."},{"key":"B36","doi-asserted-by":"publisher","first-page":"324","DOI":"10.1017\/s0373463319000638","article-title":"High definition map for automated driving: Overview and analysis","volume":"73","author":"Liu","year":"2020","journal-title":"J. Navigation"},{"key":"B37","first-page":"21","article-title":"Ssd: Single shot multibox detector","volume-title":"European conference on computer vision","author":"Liu","year":"2016"},{"key":"B38","first-page":"513","article-title":"Real-time routing with openstreetmap data","author":"Luxen","year":"2011"},{"key":"B39","doi-asserted-by":"publisher","first-page":"47","DOI":"10.1109\/79.543975","article-title":"The expectation-maximization algorithm","volume":"13","author":"Moon","year":"1996","journal-title":"IEEE Signal Process. Mag."},{"key":"B40","first-page":"266","article-title":"Detecting traffic lights by single shot detection","author":"M\u00fcller","year":"2018"},{"key":"B41","first-page":"284","article-title":"Traffic light detection with color and edge information","author":"Omachi","year":"2009"},{"key":"B42","doi-asserted-by":"publisher","first-page":"1465","DOI":"10.3390\/s17071465","article-title":"Automatic detection and classification of pole-like objects for urban cartography using mobile laser scanning data","volume":"17","author":"Ord\u00f3\u00f1ez","year":"2017","journal-title":"Sensors"},{"key":"B43","first-page":"2341","article-title":"Traffic light detection: A learning algorithm and evaluations on challenging dataset","author":"Philipsen","year":"2015"},{"key":"B44","first-page":"1","article-title":"Traffic light recognition using deep learning and prior maps for autonomous cars","author":"Possatti","year":"2019"},{"key":"B45","first-page":"652","article-title":"Pointnet: Deep learning on point sets for 3d classification and segmentation","author":"Qi","year":""},{"key":"B46","article-title":"Pointnet++: Deep hierarchical feature learning on point sets in a metric space","volume":"30","author":"Qi","year":"","journal-title":"Adv. neural Inf. Process. Syst."},{"key":"B47","article-title":"Yolov3: An incremental improvement","author":"Redmon","year":"2018"},{"key":"B48","article-title":"Faster r-cnn: Towards real-time object detection with region proposal networks","volume":"28","author":"Ren","year":"2015","journal-title":"Adv. neural Inf. Process. Syst."},{"key":"B49","first-page":"658","article-title":"Generalized intersection over union: A metric and a loss for bounding box regression","author":"Rezatofighi","year":"2019"},{"key":"B50","first-page":"1","article-title":"Going deeper with convolutions","author":"Szegedy","year":"2015"},{"key":"B51","first-page":"1","article-title":"Tracking both pose and status of a traffic light via an interacting multiple model filter","author":"Trehard","year":"2014"},{"key":"B52","first-page":"13029","article-title":"Scaled-YOLOv4: Scaling cross stage partial network","author":"Wang","year":""},{"key":"B53","first-page":"390","article-title":"Cspnet: A new backbone that can enhance learning capability of cnn","author":"Wang","year":"2020"},{"key":"B54","article-title":"You only learn one representation: Unified network for multiple tasks","author":"Wang","year":""},{"key":"B55","article-title":"3d for free: Crossmodal transfer learning using hd maps","author":"Wilson","year":"2020"},{"key":"B56","unstructured":"Detectron2\n            WuY.\n            KirillovA.\n            MassaF.\n            LoW.-Y.\n            GirshickR.\n          2019"},{"key":"B57","doi-asserted-by":"publisher","first-page":"272","DOI":"10.1016\/j.optlastec.2017.06.015","article-title":"Detection and classification of pole-like road objects from mobile lidar data in motorway environment","volume":"97","author":"Yan","year":"2017","journal-title":"Opt. Laser Technol."},{"key":"B58","doi-asserted-by":"publisher","first-page":"80","DOI":"10.1016\/j.isprsjprs.2013.01.016","article-title":"Semi-automated extraction and delineation of 3d roads of street scene from mobile laser scanning point clouds","volume":"79","author":"Yang","year":"2013","journal-title":"ISPRS J. Photogrammetry Remote Sens."},{"key":"B59","first-page":"146","article-title":"Hdnet: Exploiting hd maps for 3d object detection","author":"Yang","year":"2018"},{"key":"B60","doi-asserted-by":"crossref","DOI":"10.1109\/CVPR42600.2020.00271","article-title":"Bdd100k: A diverse driving dataset for heterogeneous multitask learning","author":"Yu","year":"2020"},{"key":"B61","doi-asserted-by":"publisher","first-page":"1374","DOI":"10.1109\/tgrs.2014.2338915","article-title":"Semiautomated extraction of street light poles from mobile lidar point-clouds","volume":"53","author":"Yu","year":"2014","journal-title":"IEEE Trans. Geoscience Remote Sens."},{"key":"B62","doi-asserted-by":"publisher","first-page":"58443","DOI":"10.1109\/access.2020.2983149","article-title":"A survey of autonomous driving: Common practices and emerging technologies","volume":"8","author":"Yurtsever","year":"2020","journal-title":"IEEE access"},{"key":"B63","doi-asserted-by":"publisher","first-page":"15","DOI":"10.1145\/3231541.3231546","article-title":"High definition maps in urban context","volume":"10","author":"Zang","year":"2018","journal-title":"Sigspatial Spec."},{"key":"B64","first-page":"4490","article-title":"Voxelnet: End-to-end learning for point cloud based 3d object detection","author":"Zhou","year":"2018"}],"container-title":["Frontiers in Robotics and AI"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2023.1065394\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,3,3]],"date-time":"2023-03-03T06:00:57Z","timestamp":1677823257000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2023.1065394\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,3,3]]},"references-count":64,"alternative-id":["10.3389\/frobt.2023.1065394"],"URL":"https:\/\/doi.org\/10.3389\/frobt.2023.1065394","relation":{},"ISSN":["2296-9144"],"issn-type":[{"value":"2296-9144","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,3,3]]},"article-number":"1065394"}}