The Polytechnic University of Hong Kong has released a multi-modal sensor dataset of the city that they claim is more accurate than previous mappings. It is the first urban mapping that provides high-precision references for point clouds and image processing.

In creating the mapping, the researchers have addressed several weaknesses of modern urban mapping using sensor data, and have developed a multi-modal approach that combines at point of mapping, multiple sensor inputs. The PUHK dataset provides a more accurate basis for the creation of urban high-definition maps, digital twins and smart city planning.

In addition, the researchers have created a number of applications from the dataset, including accurate mapping urban canyons, urban infrastructure management and maintenance, and change direction.

The researchers argue that mobile mapping is the research trend in the mapping field due to its superior time efficiency compared to traditional fixed mapping methods. It has become an important digital base for numerous applications, including high-definition (HD) maps, digital twins, and smart cities.

However, as most mobile mapping datasets are based on portable platforms, such as backpacks and robotics, the researchers argue that this leads to insufficient research on large-scale mobile mapping and autonomous driving.

To address this, the PUHK researchers developed a multimodal sensor dataset from a vehicle-mounted mobile mapping system for comprehensive urban scenes. This has been entitled the MSD-VMMS-HK dataset, and has been placed in the public domain.

The MSD-VMMS-HK dataset has rich, high-precision, and large-scale multimodal sensor information, including high-precision (millimeter-level) light detection and ranging (LiDAR), the panoramic camera, and GNSS/INS.

The dataset features a variety of scenarios in Hong Kong, which is a representative urban area with diverse and comprehensive challenging urban scenes like mountain tunnels, cross-harbour tunnels, urban canyons, mountain and seaside roads.