Semantic-Vector HD Map

This is a pipeline to construct HD Semantic Map and HD Vector Map.

This repository hosts an open-source HDVM (high-definition vector map) generation pipeline designed for autonomous vehicles, especially in intricate urban environments. Traditional HDVM creation methods often operate on a planar assumption, causing inaccuracies in real-world scenarios. Our solution, however, integrates data from GNSS (global navigation satellite system), INS (inertial navigation system), LiDAR, and cameras.

The process starts with the extraction of semantic data from raw images using advanced architectures like Vision Transformer (ViT) and Swin Transformer. We then acquire the absolute 3D data of these semantic objects from 3D LiDAR depth and derive high-precision pose estimates from GNSS real-time kinematic (GNSS-RTK) and an INS navigation system. This semantic data aids in the extraction of vector information, such as lane markings, which forms the HDVM.

A significant feature of this repo is its focus on HDVM accuracy. We’ve examined the impact of two primary error sources: segmentation discrepancies and LiDAR-camera extrinsic parameter deviations. An error propagation scheme is provided to showcase how these sources can affect the HDVM’s precision.

For ease of setup and consistency, a Docker version of the pipeline is available and is the recommended method for deployment.

For details, please refer to our official repository at HDMap.

If you find this code useful, we would appreciate it if you cite our paper.

@article{https://doi.org/10.1049/itr2.12524,
author = {Hu, Runzhi and Bai, Shiyu and Wen, Weisong and Xia, Xin and Hsu, Li-Ta},
title = {Towards high-definition vector map construction based on multi-sensor integration for intelligent vehicles: Systems and error quantification},
journal = {IET Intelligent Transport Systems},
volume = {n/a},
number = {n/a},
pages = {},
keywords = {automated driving and intelligent vehicles, autonomous driving, navigation, sensor fusion},
doi = {https://doi.org/10.1049/itr2.12524},
url = {https://ietresearch.onlinelibrary.wiley.com/doi/abs/10.1049/itr2.12524},
eprint = {https://ietresearch.onlinelibrary.wiley.com/doi/pdf/10.1049/itr2.12524}
}