Tunneling through the Trees

Just finished collecting a site for a project with some massive overhead trees.  This wreaks havoc on the GPS signal and validates the importance of having a good Inertial Measuring Unit (IMU) on board to carry the trajectory of the vehicle during the GPS outage.  This section of road is a virtual “tunnel” of trees which makes it difficult to nail the accuracy specification of 0.1-foot with GPS alone!



Elevation Data of Roadway Obscured by Trees

The next graphic shows the same area displayed by Intensity:

Intensity Image of Canopy Road

Here is a 3D Version of the Elevation point cloud:

3D Elevation Point Cloud

And here is an elevation profile of the same area:

Elevation Profile of 3D Point Cloud

Over the next couple of weeks, we will be creating 3D vectors from these point clouds that will be used to determine the geometric characteristics of the roads.  This information will then be used as parts of safety audits related to the following information:

  • Radius of Curvature
  • Horizontal Curve (Cross-Slope)
  • Vertical Curve (Grade)

This data will then be used with crash data to determine if a road needs to be re-aligned based on its geometric characteristics.  For example, a tight curve with a high speed may be contributing to crashes on a segment of road.  This information can be used to understand precisely how a road is constructed and functioning in the real world.

Advertisements

Airborne Accuracy Assessment

We’re moving along with one of our mobile/airborne projects and we just received an independent control assessment and the results are looking great!  We have achieved a project-wide RMSE of about .1112 feet for the airborne portion of this project – which is pretty good for airborne…

SR417 Project Extents

The way we check this is by loading the ground control and LiDAR data into our data viewer and then running a control report against the data.  Basically, we’re intersecting the ground control with the TIN model of the ground class of points.  The Z values are checked against one another and the difference is calculated.  These results are then used to create an RMSE for the project based on the control results.

This is a good way to get an idea of how well the data has been calibrated in terms of absolute accuracy.  Once we get a control report, we typically sort by the worst result and then start examining the control and the surrounding terrain.  The graphic below shows how we can sort the results and then “Go To” the control point in question.

Zoom to Control Point to Examine Local Conditions

We can see how the point is being assessed against the terrain.  Sometimes, there is a blunder in the terrain model and we might be able to edit the terrain to make sure it is the true ground surface.  Elevated objects such as trees can influence the accuracy assessment, but sometimes, it might be as subtle as a gutter drain, as seen in this next graphic.  The profile view of the drain cross-section shows how the terrain is influencing the accuracy assessment.

Control Point Cross-Section Showing Uneven Terrain

The goal in the future will be to collect all control on uniform areas that are not subject to sudden terrain changes.  This will ensure that the TIN correctly models the surface that is being checked for accuracy.  The next graphic shows the actual results for this project…

Control Report for Airborne Data

Our next step will be to check the control against the mobile data.  We should have that in about a week or so…

Cool Mobile / Airborne Project

We have another pretty neat mobile and airborne LiDAR project here in Orlando and the data has just started to materialize.  Over the next couple of weeks, I’ll be posting the data and discussing some of the neat things about it all.  I’ll first start by laying out the project and what we’re trying to do with it and then start discussing the results as we get into the analysis portion of the project.

The project is located in Orlando, FL down near the southern junction of I-4 and SR 417.  We are supporting a resurfacing project that is about 5 miles in length.  The goal of the project is to see if we can save the design engineers time and money by using mobile LiDAR to collect the corridor and give them an Engineering-grade model of the existing paved surface.  They will use this information to design the resurfacing project and hopefully save on materials in the field by using an accurate model of the existing conditions.

Traditionally, this information was collected through the use of Low-Altitude Mapping and Photogrammetry (LAMP) or by surveying cross-sections along the project.  Both of these technologies work and are proven to be accurate, but nothing can beat using a digital terrain model built from millions of points, right?  That is what we’re going to use and we are just getting some preliminary data from our partners in this project, Riegl Corporation.

We have a sneak peek of their new scanners, the VMX-250.  This scanner is pretty amazing and has caused us to re-write our software to handle the large amounts of data it generates.  The graphic below shows an ulfiltered data set of a portion of the project.

417 Project by Drive Line

The colors represent each drive line captured in each direction.  There are a total of 2 drive lines here, each drive line is collecting data from 2 scanners.  As you can see, there is a bunch of “junk” in there, but if you look in 3D mode, you can see that the drive lines are calibrated pretty well.

417 by Drive Line in 3D

All of the data above was collected in 2 passes with 2 scanners which is pretty amazing.  I have all 4 loaded up and the data size is over 10Gb for about a 1-mile portion of the project.  So, as you can tell, there is a ton of data to review, edit and mine for this project!

417 by Intensity and Profile

I’ll end this post with the graphic above showing a cross-section of the road and a view of it by intensity.  We’ll be working with this data over the next few weeks and as we get some results, I’ll post them here!

More Mobile Data

We’ve wrung out the calibration issues with our mobile LiDAR and we have a solid solution that we can hang our hat on. We have achieved RTK-equivalent accuracies for short runs of less than 5 miles and today, we’re heading out to see how well the accuracies hold up for a 30-mile run.

No worries about having this ready for prime time today – most projects we’re dealing with are only 3-5 miles right now. Our goal is to figure out the best way to keep the GPS solution in check for long runs to avoid post-processing and calibration issues on the back-end of the collect. Since most of the required measurements are “relative”, the LiDAR data is good, but we are striving to crack the “absolute” accuracy nut – and that involves a solid calibration of the equipment. We’ve always known this, but actually “doing” that is a different story!

This is a 5-mile run that we collected here locally. The data has been filtered to a bare pavement surface and we have run cross-sections every 5 feet. Each cross-section can be exported to CAD, GIS, Microstation, etc and used to build pavement resurfacing design drawings.

Cross-Slope Calculation for a Section of Alafaya Trail, Orlando, FL

We have been able to make the processing of this data “semi-automated”.  Basically, we have to draw in the breaklines which typically correspond to the pavement stripes.  These define the lanes of travel and then we do a slope calculation (relative measurement) from one breakline to another – effectively calculating the cross-slope percentage for each lane.  We’re also exporting out a tabular format so clients can use that information to verify the values against a pavement design spec.



Slope 1 refers to the left lane (southbound) and Slope 2 refers to the right lane of travel (northbound) and all slopes are percentages.

Mobile LiDAR Update

Here’s some LiDAR data from the City of Charlotte Pilot project.  The images below shows the different data stored in the point cloud.  This information is useful to our team of processors during the data extraction phase.  For example, the image below shows the LiDAR point cloud themed by intensity.  The intensity information can be used to extract different types of vector data such as the road centerline, edge stripes, pavement markings and other special pavement markings.  The intensity values can be used to assess the condition of the markings based on its reflectivity.  Basically, the markings are reflective or they’re not and that is a representation of their condition.

Pavement Marking Condition Assessment

Pavement Markings from LiDAR Point Cloud Intensity

The point cloud can also be used to generate contours.  The image below shows contours that are 0.1-foot.  A trained eye can see the usefulness of this data in that a roadway transition is occurring in this area.

Elevation Data Displaying 0.1-foot Contours

The next image shows a roadway profile from the same area.  This profile information can be used to generate cross-sections at any interval along the roadway.

Elevation Profile

Airborne LiDAR Update

We are operational!

Aircraft – Check!

LiDAR – Check!

Aerial Photography Camera – Check!

Hyperspectral Camera – Check!

Avalon Park LiDAR

LiDAR Data Colored by Elevation

We’re operational and have a ton of data in the can and ready for processing.  Our data sets include samples from residential communities to transmission powerlines to unmentionable clients who have some interesting needs!  One of the biggest hurdles has been developing our own viewing software that we can deliver with these large datasets so that our clients can manage their deliverables.  The goal is to build a piece of software that is lightweight and easy to maintain code-wise, while building tools that clients can use to streamline their business processes.

3"-pixel Aerial Photography

Keep watching here for data samples and updates to our software!

Pavement Camera Update

We’re finishing up on the City of Charlotte’s pilot pavement data collection project.  To date, we have collected Mobile LiDAR, Mobile Video, Ground-Penetrating Radar, Roughness and Rutting data for a 50-mile pilot area.  This was the first go-around for our pavement camera and the results were great.

Downward-Facing Pavement Camera

Pavement Cam

What’s cool about this is that we can now get a great view of the lane of travel and see the low density cracking because we’re basically collecting 2mm pixels.  We’re working on learning how to orthorectify these images so they can be fused with the point cloud to give a real-world representation of the pavement surface that can be viewed in 3d.