Automated Feature Extraction Goes Mainstream
Civil Maps is an artificial intelligence (AI) start-up based in Silicon Valley. Over the past couple of years, they have developed industry leading technologies for the autonomous vehicle industry. Their key technologies include automated registration of LiDAR point clouds without the need for ground control points, automated feature extraction to rapidly produce large area maps, and Scout, their low-cost self-contained mobile LiDAR collection unit.
Over the past year, Civil Maps has been commercializing this technology stack for the engineering, environment and construction market and has gained considerable traction, including work on a major Canadian pipeline project. Anuj Gupta, VP of Business Development, recalls, “The bottleneck in map creation starts after LiDAR data collection. Manufacturers like Velodyne provide cost effective and scalable data collection solutions. The engineering, environment and construction industry was struggling with using manual resources for feature extraction from LiDAR data. Since introducing it to the market our automated feature extraction has been well received by the industry and we continue to gain traction through work on capital and operations infrastructure projects such as pipelines and transmission corridors.”
Automated Point Cloud Registration
One of Civil Maps’ key technologies for autonomous vehicles, is their ability to automatically register point clouds. This has been a major leap in the development of the navigation systems for autonomous vehicles because it allows highly accurate point cloud to be generated when the same area is surveyed at different times. But when only a few control points are available, they can further increase the georeferenced accuracy of the point cloud to near survey grade. For engineering mapping purposes this “few centimeter” level of accuracy certainly meets most requirements, particularly for regional mapping purposes.
When LiDAR data of this level of accuracy is available, customers can use Civil Maps Automated Feature Extraction technology to rapidly map the above ground features in an area of interest. Traditional survey teams only map specific areas and often have to return to an area for more collection if a project’s Right-of-Way changes, or the facility footprint changes considerably. This usually entails a period of more landowner interaction, and higher cost and schedule protraction.
Using feature extraction from LiDAR, the entire swath of the LiDAR collect can be processed and all the features mapped. With a larger surveyed area and relevant map with extracted features, project managers can achieve a complete situational awareness of the project at a regional scale. A high regional awareness leads to detailed survey planning, reducing the amount of landowner interaction and impact on the environment, and ultimately is safer because of less field time.
As Joseph Hlady, Head of Sales for Engineering, Environment and Infrastructure explains, “… our technology does not replace the need for traditional field survey, but it helps reduce the features that need to collected by time constrained, often expensive, survey teams. Our customers, some of whom are survey companies, are ecstatic at the fact that they no longer have to send a team to go and collect nominal features like poles, fences and trees. Rather, they can go to the field and focus on buried utilities, complex crossings and structures, and assets that require legal and high precision survey. Our technology is lowering the cost and time for field data collection.”
How Does Feature Extraction Work?
As mentioned above, Civil Maps is an Artificial Intelligence company. They realized that to efficiently create and audit maps at global scale, traditional software tools & manual methods were not scalable. Therefore, they have developed a complex set of supervised and unsupervised classification technologies, using AI to classify and feature code point clouds.
Depending on the size and complexity of the area and features to be mapped, different AI tools are used. Like any mapping projects, there will always be some human interaction to either create and/or edit features. But at Civil Maps, AI is used in every process, even for the most basic mapping tasks in order to reduce human error, and often, to eliminate human input all together.
The scalability and accuracy that AI has brought to feature extraction is revolutionizing the mapping industry. Working with ESRI, Velodyne and other industry partners, Civil Maps is in the process of commercializing this service offering to every industry, which is offered as an online tool where customers can upload their data and select the features they would like to extract from the Civil Maps feature catalogue (which continually grows). Then the customer is contacted to discuss the processing quote and project parameters in detail. And then Civil Maps goes to work to register the client’s data (if required) and conduct the feature extraction.
Who Benefits from this technology?
Recent projects for Civil Maps have included a major pipeline project and road infrastructure mapping to name just a few. But the technology stack and the service delivery model is designed to service all customers, both collectors and users involved in engineering, environment, and construction projects. Clearly, projects or operations and maintenance work involving large geographic areas benefit most from this technology because of the scalability the service brings.
What about Collection?
In their work mapping California for autonomous vehicles, Civil Maps realized that there is just not enough data out there yet. So they developed a low-cost collection platform called Scout, that they can use to ensure they have enough data for their mapping campaign. They were able to develop this platform very quickly because while developing software for the engineering market Civil Maps began to work with Velodyne, the manufacturer of a number of cost effective 3D laser scanning products.
Velodyne had their roots in the autonomous vehicle market beginning with the DARPA challenges some ten years ago. Anuj noted, “Being lower in cost compared to other survey grade LiDAR manufacturers, it is a go-to LiDAR manufacturer for many autonomous vehicle companies, especially because it can still achieve high levels of accuracy.”
The Civil Maps Scout is a low-cost, high-precision geospatial data collection system, with real-time localization and mapping capabilities. It integrates a Velodyne VLP-16 LiDAR sensor, cameras, inertial measurement devices, and wired and wireless communication interfaces into one package. Anuj commented, “We designed this unit for mobile mapping and localization applications at speeds up to 160 kph.”
In addition to the Scout, Civil Maps is also using survey systems with Velodyne HDL-32 LiDAR sensors. They use these sensors to collect data for autonomous navigation and high-level route planning. Anuj explained, “Once the raw sensor data is collected, we use our AI engine to extract features such as road lane markings, road centerline, signals and regulatory signs. We put these and other navigation related features and more into our 3D semantic maps. These semantic maps provide the relationships between the lane centerlines and signs, signals, and lane markings which allow for autonomous navigation.”
LiDAR Goes Full Circle with Civil Maps
With the addition the Scout collection platform Civil Maps is now able to provide the full range of LiDAR services for Infrastructure. Joseph, states that “… although targeted at our autonomous vehicle customers, Scout has really rounded out our service offering. Now, not only can we provide feature extraction services from LiDAR, we can also help our smaller customers, like field inspection companies, collect LiDAR data appropriate to their projects and using our registration technology, we can ensure that the point cloud is ready to be used by anyone.”
Civil Maps is well along the path of position itself as an industry leading technology company for autonomous vehicles and data processing for infrastructure. The company is excited about upcoming technology announcements this summer and fall.