Abstract Manager

Vehicle manufacturers have been competing to achieve fully autonomous driving. However, many experts believe that this goal cannot be achieved using on-board sensors alone. These experts stress the need to produce precise maps that represent a digital twin of road and its surrounding environment in a sustainable manner. This is enabled by High Definition (HD) Maps. HD maps are digital twins of road infrastructure which provide 3D representations of road features, such as lane markings, road edges, etc. HD maps production is performed by first surveying roads using remote sensing technology. The collected data is then segmented using computer vision and machine learning algorithms to produce a HD map of the features of concern. These features are essential to a vehicles ability to navigate a road and maintain a consistent trajectory. This facilitates several autonomous driving functions including, localisation, perception and navigation. One critical issue that is specific to extracting lane marking information is that the markings are occasionally occluded or degraded due to high traffic volumes. This results in failure to detect lane markings and the existence of significant gaps in the information extracted to produce the HD map. Conventionally, these gaps are manually filled in by quality control staff in a tedious and time-consuming process. To overcome this issue, a novel algorithm is proposed through which Kalman filtering is combined with Bezier curve fitting to automatically refine lane marking information. To evaluate the performance of the algorithm, LiDAR data collected on multiple roads was used, where segmentation was first applied using deep learning technology. The proposed algorithm was able to close all the gaps and recreate missing portions of the extracted lane markings in a highly efficient manner. In seconds per kilometre, the runtime on the test segments ranged from 5.86 to 27.6s/km which is extremely efficient compared to the manual checks that take place today.