Detection of road markings from LIDAR data

It is not rare to find LIDAR technology (Light Detection and Ranging) in the sensor setup of an autonomous car, since it can offer many features and compensate lackings of cameras. Since LIDAR sensors measure reflectivity, they can be useful in road markings detection methods. Many different techniq...

Descripción completa

Detalles Bibliográficos
Autor: Pérez Moré, Xavier
Tipo de recurso: tesis de maestría
Fecha de publicación:2022
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/376994
Acceso en línea:https://hdl.handle.net/2117/376994
Access Level:acceso abierto
Palabra clave:Remote sensing
Automated vehicles
LIDAR
road markings detection
autonomous driving
Teledetecció
Vehicles autònoms
Àrees temàtiques de la UPC::Enginyeria de la telecomunicació::Radiocomunicació i exploració electromagnètica::Teledetecció
Descripción
Sumario:It is not rare to find LIDAR technology (Light Detection and Ranging) in the sensor setup of an autonomous car, since it can offer many features and compensate lackings of cameras. Since LIDAR sensors measure reflectivity, they can be useful in road markings detection methods. Many different techniques have been explored to take advantage of their higher reflectivity in relation to other road elements. An algorithm for the detection of road markings from LIDAR pointclouds is proposed and tested on the KITTI raw dataset. An augmentation of the pointcloud size using the data from the navigation sensor and a filter to detect road points are applied to ensure a good detection thanks to a thresholding performed based on the reflectivity values. It is proven that indeed detection from LIDAR measurements is possible regardless of the external light conditions. For future work, a deep learning approach is encouraged to classify the detected road markings.