In 2000, a 3D laser scanner could collect roughly 1,000 points per second with position precision of 6 millimeters. Today, manufacturers offer scanners that can capture more than one million points per second with precision of 1 to 2 millimeters. The thousand-fold increase in measuring speed is a technological advancement that has created opportunities, and challenges, for geospatial professionals. It’s also a driving force behind some important advances in software technology.
Along with the massive data sets produced by today’s faster, more capable scanners, the geospatial industry is experiencing increased demand and rising expectations on how the data is used. With projects ranging from forensics and architectural preservation to mining, construction and a plethora of industrial applications, scanning service providers are under pressure to produce actionable information quickly, and with high accuracy.
Managing Big Data
By operating at high data rates and densities, 3D scanning has delivered large gains in field productivity. But according to many geospatial professionals, the gains in the field are partially offset in the office, where processing the data can be time consuming.
For example, technicians must first combine, or “register,” multiple scans into a single point cloud. To do this manually, they must select targets or common points in two or more scans as the basis for merging the scans. The resulting point cloud harbors valuable data, but often requires further processing to produce deliverable information needed by downstream users. The work to create the deliverables is often performed manually by a technician and desktop computer. It’s a painstaking job that can be slow and tedious.
To streamline the process, software developers today are focused on increasing productivity in office processing by automating key tasks. It begins with scan registration.
Automated Registration and Classification
Automated registration lies at the foundation of efforts to increase productivity in processing geospatial LiDAR and imagery. For example, automated registration in Trimble RealWorks software can automatically identify not only scanning targets, but also common features between multiple scans and use those in the registration computations. As a result, the need for scanning targets is reduced or eliminated, saving time in both the field and office, as well as the expense of purchasing and handling the targets. Freed from the high point density required to accurately capture targets, field crews can adjust scan densities and shorten scanning times.
Automated registration also reduces the need for operator interaction, allowing office technicians to perform other tasks while the software handles the data processing. Even with automated software, the work of processing and registering a day’s worth of scanning data may require a few hours of computation. But the processing is automatic and frees up the human operator for other tasks. I spoke recently with a scanning expert who described his work process on a typical project: “At the end of the day I would come back to my hotel and get the software started on registering the scans. When I got back from dinner, everything would be properly registered. Then I could use the results to plan the next day’s work.”
With the point cloud registration complete, automated routines can also take over some tasks formerly relegated to human technicians. The next step in processing typically involves locating outliers, ghosts and superfluous points, as well as points that make up objects of interest. For example, a technician can manually select a group of points that define an object, classify it and assign it to a layer. On a large or complex scanning project, the process may be repeated hundreds of times.
Thanks to advanced computing software, automated classification is becoming the new standard. Features such as ground surfaces, vegetation, poles and building features can be automatically classified and layered with little human interaction. Likewise, the software can isolate noise in the data and unneeded points to layers that can be switched off. No data is deleted in the process; the technique hides the points without removing them from the dataset, an important consideration in forensics and other applications where transparency and data integrity are fundamental.
Moving to the Model: Automated Modeling
Software manufacturers have greatly enhanced the use of point clouds directly within their software, making it easier to use the data natively for as-built or clash inspection applications. However, for other applications and ultimately the users of the information, the key task is to turn the point cloud information into a 3D CAD model. In addition to classifying planes and ground surfaces, automated modeling software can create 3D entities for use in CAD and design applications. For example, office technicians can use modeling tools in Trimble EdgeWise™ software to extract features such as pipes and structural components (steel I-beams) and create 3D solid vector models based on libraries of standard component geometries and dimensions.
Tools for automated modeling are extending beyond traditional point cloud software. Technicians can use the Scan Explorer Extension for SketchUp software to access Trimble RealWorks point clouds. They can extract edges, walls, corners and points and place them into SketchUp. The information can then be used in creating 3D models and designs that relate precisely to the scanned objects.
The benefits of automated modeling and identification are significant. In scanning an industrial facility with thousands of pipes of different sizes and orientations, automatic identification and modeling can save hundreds of operator hours in time and cost. Once the piping is modeled, technicians can draw from libraries of pre-modeled equipment (pumps, valves and fittings) to create a comprehensive, detailed 3D model for design, building information modeling (BIM), construction and asset management.
As software automation continues to evolve, it can provide specialized functionality. For example, storage tank functionality in Trimble RealWorks can automatically register and process scan data to model the floor, roof and walls (or “shell”) of petroleum storage tanks. The results can be analyzed to determine tank capacity and condition. Visualization tools help reveal bulges or irregularities in the shell that might affect capacity or safety. When performed manually, the work to process and analyze a tank scan to produce an inspection report could take up to a week. By using 3D scanning and automated modeling methods, the job can often be completed in one day.
Software automation extends to imagery as well. For example, scanned point clouds can be supplemented by high dynamic range (HDR) photographs. Automated software can match the photos with scanned data to automatically produce colorized point clouds and highly realistic 3D models. And automated classification of aerial imagery can extract information on structures, vegetation and surfaces.
Smart Clouds, Smarter Software
Like most automated systems, the goal of automated modeling is to reduce the time that humans spend on a process or task. While human judgment remains a vital part of data processing and analysis, automated tools make it faster and easier to provide high-value services and deliverables. Time once spent on registration and modeling can be better used for analysis and to focus on deliverables and quality control. Classification routines can produce “smarter” point clouds that provide more efficiency. Classification can be especially helpful if a project needs additional analysis and modeling once the primary work has been completed.
Geospatial professionals can use scanning and imagery to deliver significant gains in productivity and flexibility for their clients. As software evolves and automated capabilities expand, classification and 3D models will become increasingly common.