Several months ago I suggested a column on possible future technical innovations in laser mapping. As we all know, predicting the future is not so easy, but its fun to think about and helps shake out the cobwebs of the daily grind. So in no particular order, here are the top five advancements that Id like to see in the not-so-distant future:
One-pass color. It almost goes without saying that we can expect future generations of technology to continue improving along the cheaper, faster, better route. While there are lots of possibilities, one specific feature that would be immensely useful is to eliminate the double-pass requirement for color scanning that most major products require, and simply acquire both the range and color data together during the initial pass. The reduction in collection time would be appreciated, but this enhancement makes the list is because of the mental shift that would take place: colorization of point clouds would go from optional to de facto. The ramifications could be far-reaching, similar to the shift from black-and-white to color photography decades ago.
Error bars. Fundamentally, a laser scanner is a device for making measurements. Remembering our high school science, valid measurements should always be accompanied by an estimate of their accuracy or uncertainty. For scanners, this is often accomplished by publishing a few numbers in a data sheet, e.g., 3 mm (1) at 20m on an 80% reflective target, but these specifications are insufficient for systematic or quantitative use. It would not be difficult for manufacturers to add error estimates to each and every scanned point, thereby giving practitioners a better understanding of the collected data. Error estimates would also improve the quality and power of processing software. For example, when fitting a surface, algorithms that have access to the uncertainties can weight the points appropriately to obtain more reliable results. And wouldnt it be nice to view a colorized plot of your scan: red showing large uncertainty; yellow, medium; and green, small?
Surface normals. For those not familiar with the term, Surface normals refers to an imaginary vector field running perpendicular (normal) to each point on a surface. Surface normals are critically important, for example, in obtaining realism with computer graphics. Surface normals attached to a point cloud improve their visualization, but also enable new, powerful functionality in processing software. As an example, tools that automatically separate roads from bridges (or pipes from structural steel) can do a better job if the surface normals are available. While some products do compute the surface normals, often times this is both time-consuming and tricky to do from a general point cloud. However, there are solid reasons (too technical to address here) that computing the normals at the hardware level can be done relatively quickly and accurately, especially in comparison with computations done as part of post-processing.
Flash lidar. Flash devices have been around for a while, but are not yet able to compete with scanners in terms of operating range, resolution, sample rate, and accuracy. Scanners, of course, require mechanical components for steering a beam through space. By eliminating these parts, flash devices promise better reliability, lower operating power, and smaller packages. And there are other advantages: theoretically a sensor operating from a single flash of light collects data so fast that motion blur is not an issue. Also, 3D frame-rate video is not inconceivable. But, significant technological hurdles stand in the way; hopefully they will be overcome.
Raster imaging framework. OK, this one is out there a bit, but let me explain. Almost all laser scanners are raster devices, that is, they collect measurements in uniform rows, like pixels in an image. This wonderful organization is often thrown away when the data is imported into processing software. In fact, the common term point cloud implies a somewhat disorganized array of points, in contrast to a similar term, 3D image, which connotes more organized information. Benefits of raster preservation include: smaller file sizes and extremely efficient compression, easier encoding of surface normals and error estimates, and real-time, image-quality visualization. In fact, there are several important software products available on the market that preserve the raster nature of individual scans and provide powerful tools for working with them. However the concept is not universally accepted. As an initial step, Id like to see the development of a robust, raster framework for raw scan information that can be used as a working format for a variety of software products and tools.
My list could go on an on, and surely you have your own list. Of course reality has a funny way of upsetting our best-laid plans, so lets not hold our breaths as we wait and see what the future brings.