LiDAR vs Camera Based Survey Performance

The best way to create a 3D topographic map in steep terrain.

Case Study: Wide Area LiDAR and Camera Based 3D Survey Performance in Steep Terrain

Introduction: AboveGeo Inc. completed a very challenging LiDAR collection project in extreme terrain of an Olympic Ski Resort, then evaluated the capabilities of a competing technology called Structure from Motion in a similar situation. We found that both approached yield comparable results and Structure from Motion can often provide these results at much improved cost.

Preparation: We were tasked with a unique project to provide LiDAR and ortho imagery of a series of ski resorts outside Lake Tahoe, California. This particular project was unique because of its high accuracy and coverage requirements over steep terrain. We immediately rejected the use of traditional fixed-wing aircraft as the approach would not yield the customer required accuracy. We strongly considered using Unmanned Autonomous vehicles with optical cameras and point cloud generating software but the ground control network would require potentially hazardous and time consuming survey work before the job could commence. As the customer wanted the work completed quickly (before the snow), we opted for LiDAR using a Bell 206L helicopter carrying a lightweight integrated LiDAR/optical camera. We were able to gather extremely dense and accurate data of the mountain for use with ski grooming equipment. We then flew a portion of a nearby ski area which already had historic LiDAR data and compared the results of each.

Project Constraints: Our customer would be using the terrain data to provide based heights for snow grooming equipment. Once the snow had covered the mountain, grooming equipment would be able to navigate through areas using a machine guidance interface of which alerts the operator when they are in deep snow or light coverage. Snow grooming equipment is usually operated at night and in white-out conditions. The machine control interface extends the operators knowledge of conditions in a three dimensional view and improves run conditions, improving the effort of snow making and extending the ski season for the resort.
The customer requirements for the wall to wall, three dimensional terrain model were stated at 5 centimeters vertical and 5 centimeters horizontal. This is accuracy requirement, in and of itself, is a challenge. Add the extreme topography of a world class ski resort, and that challenge is multiplied.
Wide area survey grade photogrammetry in extremely steep and challenging terrain can be challenging for any aircraft. The Squaw Valley Ski Resort ranges from 6,000ft above mean sea level (MSL) at the base, to nearly 8,900ft MSL at the top. With nearly 3,000ft of vertical terrain, containing large cliffs and dynamic terrain a fixed wing aircraft flying a constant altitude over the highest terrain would be insufficient to meet the extreme accuracy required.

In order to maintain a high level of accuracy in such varying terrain, several Quality Assurance efforts needed to be followed:
To begin with, acquisition altitudes above ground level (AGL) would need to be as low as possible. Lower acquisition altitude = less ambiguity and error. These degrees of error can come many shapes, but most notably in roll, pitch, and heading errors. In projects with long, straight flight lines, heading will typically be the error that suffers the most. Our Quality Assurance plan included flight lines which are introduced with dynamic movement to keep the Inertial Measurement Unit (IMU) in proper alignment. Traditionally, this is why helicopters and multirotor drones are required for projects that demand greater accuracies. That said, most of the trees were over 250 feet tall. While the use of drones would reduce the distance from ground, we would have issues with treetops and line of site between operator and drone would require time consuming ground station relocations.

Second, the sensor’s field of view (FOV) needs to be narrow to keep angular returns to a minimum. In order to capture adequate ground coverage, LiDAR pulses need to penetrate as vertically as possible. As mentioned, the site contains many Ponderosa Pine which often have a height of 250 feet additionally with the cliffs and obstructions, a high level of overlap and a limited FOV became a requirement. Dynamic terrain and limited FOV necessitates a high number of flight lines to limit shadowing and maintain sufficient coverage on oblique angles of ground.

Next, in order to create a product that has sub 5 centimeter global accuracy, survey correction points (GCPs) need to be strategically placed throughout the survey area. These are placed in areas known to be sufficient for LiDAR correction, but also key areas for calibration and flight line adjustment.

Lastly, a slow acquisition speed is required. Not only did this project have a fairly high density requirement, but more density can also help with data calibration and target identification. In this particular project, the acquisition speed was also the helicopter’s vertical speed (Vx). Vx is the speed for best angle of climb. In challenging and steeply rising terrain, a helicopter needs to climb and descend constantly, while trying to maintain a constant ground speed. Flying a Vx speed allows us to maintain a constant speed, and climb/descent the different parts of the mountain.

Results: Placing the ground control points took a day and flight operations took less than one day, capturing both LiDAR and ortho imagery of the project scope. Ortho imagery was captured simultaneously with LiDAR, sharing the same trajectory information. With precise camera calibration and elevation data produced from LiDAR, the imagery can be tiled and mosaic’d.
Post-production efforts for LiDAR included the trajectory solution, sensor calibration, strip line adjusting, and classification. Post-production for imagery includes the fine tuning of camera calibration, exposure enhancements, and seamline adjustments.

Shown below are the resultant accuracies and specifications. Through diligent QA efforts, the largest outlier in the data set was just over 10 centimeters. Total collection produced over 1.7 billion points, scanning at 600 kHz.
Angular and long range returns were limited, all while keep a constant flight altitude above the ground. This helped produce a consistent dataset, ensuring accuracy and density throughout the project scope.

Similar Project using Structure from Motion

Structure from Motion or SfM is a photogrammetric method for creating three-dimensional models of a feature or topography from overlapping two-dimensional photographs taken from many locations and orientations to reconstruct the photographed scene. In addition to ortho-rectified imagery, SfM produces a dense point cloud dataset that is similar in many ways to that produced by airborne or terrestrial lidar. The advantages of SfM are its relative cost in comparison to LiDAR, as well as its ease of use. The only required equipment is a camera. A computer and software are needed for data processing. Additionally, an aerial platform like a balloon or drone can also be useful for topographic mapping applications.
AboveGeo uses a variety of techniques to gather SfM content creating point clouds for the development of three dimensional maps and vegetative analysis. For future projects similar to Squaw Valley we wanted assurance that SfM could provide acceptable results at a greatly reduced price. For the test we went to the Diamond Peak Ski Resort, also near Lake Tahoe, surveyed several locations on a steep ski run and created a point cloud using an autonomous aircraft equipped with a high resolution camera and using terrain following software to plot its path and altitude above ground level (AGL).

The accuracy of the SfM data collection can be described in terms of absolute accuracy which is the consistency of the data with external data sources and relative accuracy which is the consistency of the dataset with itself.

Absolute accuracy was assessed using ground shots taken with Trimble 5800 RTK GPS system and comparing them to triangulated estimated values from SfM. Fundamental Vertical Accuracy (FVA) reporting designed to meet guidelines presented in the FGDC National Standard for Spatial Data Accuracy. For the Diamond Peak SfM project, FVA was calculated by comparing RTK ground shot point data collected on open, bare earth surfaces to the triangulated surface generated by the SfM points. FVA is a measure of the accuracy of SfM point data in open areas where the SfM system has a high probability of measuring the ground surface and is evaluated at the 95% confidence interval (1.96 * RMSE), as shown in Table 1.

The mean and standard deviation (sigma σ) of divergence of the ground surface model from ground shot points coordinates are also considered during accuracy assessment. These statistics assume the error for x, y and z is normally distributed, and therefore the skew and kurtosis of distributions are also considered when evaluating error statistics. For the Diamond Peak survey, 18 ground survey points were collected in total resulting in an average accuracy of 0.067 meters. This content is combined with orthophotography to provide a three dimensional map useful for a variety of related purposes.

AboveGeo finds that each customer project should be evaluated independently of others and the ultimate objective is to provide a valuable product to fit our customer’s needs. In some cases we may use LiDAR or SfM we also determine the best flight operations be they manned or unmanned vehicles.

Conclusion: New sensor technology has allowed for large area/high density/high accuracy capture to take place in just a short morning. Although the size of the project is nothing uncommon, the resultant density and accuracy is what makes this project unique. What would otherwise take a survey crew months to accomplish, AboveGeo was able to accomplish in two days with an acceptable and cost effective alternative in appropriate situations.
There is no doubt that LiDAR sensor and SfM software technology is revolutionizing aerial data collection. In just a few short years, the cost and quality barrier has been reduced dramatically. This makes LiDAR and SfM techniques approachable for projects with tight budgets and schedules, where it was otherwise time consuming and cost prohibitive. AboveGeo takes advantage of this sensor technology on both UAV and manned-aircraft platforms.

Sample Data: Click on the viewer to the right to view this point cloud in your browser. (Works best on Chrome, Explorer, or Safari)