What can scanning not do?
Author: Dr. Robert Radovanovic, P. Eng., A.L.S., C.L.S.
When people see a live rendering of a 3D point cloud, and start zooming, rotating and orbiting in a virtual space, you can immediately see that they’re thinking of a million things they can do with such a data set. “The old ways of doing things are gone forever!”
Unfortunately, the reality of actually using 3D scan data is a bit different. Many project managers have suffered from ‘point cloud overload syndrome’ – the sinking feeling you get when you’ve requested a laser scan be done of a project and you are left with billions and billions of unusable points as a product.
At CorridorMaps, we’re big believers in the value of using laser scanning to improve project outcomes. However, we’re also realists about how that data needs to be collected and processed, and that scanning isn’t a cure-all.
In particular, the following is a (non-comprehensive) list of things scanning CANNOT do :
- Collect data under all weather conditions (snow, fog, rain)
- Ensure 100% coverage if there are obstructions in the area (cars, fences, vegetation)
- Guarantee data quality without additional QA/QC
- Instantaneously provide 3D CAD models, or ‘quick-fixes’ in the design world without pre-processing or additional tools
- Drive survey data acquisition/processing costs to zero
Surveyors are generally known for their “whatever the weather, we’ll do it” attitude, but the physics of 3D scanning puts a damper on what can be accomplished in certain weather conditions. Since scan collection relies upon a laser pulse successfully bouncing off a target and returning to the instrument, anything that impedes this results in a bad scan day.
In particular, raindrops, snowflakes and fog droplets become millions of tiny reflectors when scanned, resulting in an incredibly noisy data set. Also, although modern scanners often have the ability to collect more than one return per pulse, the mass of reflectors a beam needs to go through on a really rainy day means that there is usually no energy left to actually bounce off of the intended target objects in the scene. Similarly, the laser pulse cannot penetrate through snow, so scanning in snow covered areas can’t pick up ground details.
In both mobile and static scan situations, the scanner can only ‘see’ things in its direct line-of-sight. Although that seems obvious, it’s surprising how often that is forgotten in practice. As a result, scanning will not pick up objects located behind shrubs, hedges or fences – objects in shadow zones must be picked up via conventional survey.
A commonly overlooked issue is that of cars on the road. In the case of moving traffic, cars surrounding a mobile scanner will obstruct features on the roadway. This is one reason why scans should bedone with multiple passes. It is unlikely a car will be in exactly the same place with respect to the scanner on the second pass.
Parked cars unfortunately don’t benefit as much from that mitigation strategy. It’s important to plan a scan collect around times when the number of parked cars on a road is at a minimum, coordinate a parking restriction, or manually verify objects such as catch basins/manholes after the fact.
Guaranteeing data quality
Making sure that scan data is fit-for-purpose is worthy of a blog post on its own, but suffice it to say that single-pass laser scan data without additional quality control / quality assurance (i.e. from conventionally surveying check points) is exceptionally unreliable.
The main problem is that all scan data ‘looks’ good – it’s detailed and dense, and it seems that everything is in the right place. However, without actually checking it against reference data such as an independent scan or a conventional survey, it’s impossible to reliably say how correct it is. There could be datum or georeferencing issues or the trajectory quality during collection could be poor, and without a control set to check against you would not be able to find these issues This is also one of the main reasons why scan data can get a bad wrap, if the data did not go through a quality control / quality assurance process, then these issues are often found during the final design stage – or even worse the construction stage.
Actually using 3D scan data
Billion-point laser scans are undeniably cool to look at and visualize. However, turning them into useful information is still really hard. There are a number of 3rd party software packages that let you create 3D models from point clouds, but the vast majority essentially allow you to virtually survey an area by manually selecting points, drawing lines, and creating surfaces. While it’s certainly faster (and safer) to do this from a computer screen as opposed to physically walking around in real life, it’s not significantly faster – essentially, we are replacing time spent surveying in the field with time surveying from the office.
Increasingly, companies are investing in technologies like machine-learning to automatically pull out features like signs, powerpoles, breaklines and pipes from scan data, but the success rate of feature extraction isn’t perfect – and unfortunately, in surveying, not being 100% correct may as well be 0%.
Aside from using scanning as a survey-replacement tool, some companies are focusing on providing software to create Digitized Realities based on scan data, with the theory being that the true value in the scan data is to create a 3D world of points. Instead of extracting features as CAD elements, the goal is to drop proposed models into the digitized reality and look for clashes, or to render what the proposed model would look like.
People tend to equate time with money, so when they hear that mobile scanners can collect data in the time it takes to drive down the street, they generally assume this will be a ‘cheaper’ way to get the survey data they need. Unfortunately, this isn’t always the case.
As we’ve seen, collecting the scan data itself is only one piece of the puzzle. There is still the need for remedial conventional survey for QA/QC, pickup of hidden features, and there is often the considerable requirement for feature extraction to make the data meaningful. Not to mention the very high capital cost of scan equipment to contend with.
For those reasons, we’ve found that full-cycle costs for laser scanning can be approximately 60% of conventional costs on projects larger than 10 km – shorter than that, and it becomes uneconomic to compete with sending people in the field to do conventional pickup.
Of course, there are specialized cases where the data detail available from laser scanning, or the lack of requirement for traffic control make it appealing for even short projects.
At CorridorMaps, we’ve generally found that the real cost benefits stem from the fact that your costs can be spread over the duration of a project. For example, since raw collection costs are low, we can provide unvalidated scan data at a low price, which is suitable for visualization and initial planning. Once your project needs to move into preliminary design (and now has a budget), we can arrange QA/QC surveys and adjust scan data later. Similarly, feature extraction costs can be triaged by specifying certain features to be pulled out of the scan data up front, and other features later.
At CorridorMaps, we firmly believe in the value of reliable 3D scan data to support projects. However, thanks to lots of experience in delivering products to repeat clients, we’re also very aware of the limitations of this technology. The upside is that whatever your project need is, you can call us up and we’ll be able to help you find the solution that’s right for you.