The City of Port Adelaide Enfield is working with Mapillary, leveraging computer vision and collaboration to unlock the wealth of spatial data contained within their road assessment video logs.
Roads and footpaths have been surveyed over the last few years, with video logs processed to automatically extract thousands of map features. This open approach allows communities such as OpenStreetMap to both use and validate the data, improving the quality of Port Adelaide’s derived datasets.
City councils and departments of transport around the world collect street-level imagery and/or video logs to monitor assets, conduct safety inspections, and coordinate maintenance activities. These surveys usually sit on internal servers which only a handful of individuals can make the most of. Councils like the City of Port Adelaide Enfield have decided to open up this data, making it available to a much wider audience, but also benefiting from the opportunities cloud processing and computer vision create. This approach is beneficial in a number of ways.
1. An improved viewing experience
Structure from motion (SfM) is used to reconstruct the City of Port Adelaide Enfield in 3D. This not only helps to estimate the latitude and longitude of detected features, it also helps city employees navigate between images more seamlessly. The more images collected by the city and the public, the better this 3D reconstruction.
2. Derived Data
Thousands of map features like traffic signs, utility poles, benches, bike racks, and crosswalks are automatically extracted from the images using computer vision. These datasets are compared with existing datasets that the city has, allowing for missing assets to be added without a cost intensive manual collection process.
3. Crowdsourced validation
Another benefit of having this imagery publicly available is that citizens from within Adelaide or as far as Amtserdam can play a role in validating objects that have been detected. A bench detected in an image can be confirmed or rejected by members of the public, improving the quality of the dataset for Port Adelaide, but also the likelihood of these objects being identified correctly in the future.
All of the imagery and derived data is available to the OpenStreetMap community to use and is compatible with the Open Database License (ODbL). Map features like crosswalks that have been detected and curb cuts can be used to improve pedestrian routing or wheelchair accessibility. OpenStreetMap is an open source, editable map of the world that anyone can contribute to. OpenStreetMap is widely used in applications such as Apple Maps, Bing Maps, Facebook, Strava and others.