46 PositionIT – Jan/Feb 2011

Visualisationtechnical

PositionIT – Jan/Feb 2011 47

An introduction to lidar-based aerial surveys (Part 1)by Adriaan Combrink, CK Aerial Surveys

Since the early 1960s laser has often been referred to as “a solution looking for a problem” to solve. Today this still rings true, as lidar-based aerial survey (LBAS) technology continues to provide an increasing number of solutions to survey needs, especially in large-scale projects or in areas that are difficult to access for conventional surveys.

Aerial surveying – and aerial photography in particular – has been around since 1858 when

balloonist Gaspard-Félix Tournachon took photos over Paris. However, for the purpose of this series of articles, it is very import to note the differences between conventional aerial surveying and LBAS:

l Conventional aerial surveying is an indirect, labour-intensive and less precise (for reasons to be explained later) surveying technique which is highly dependent on extensive ground control and point matching for the completion of survey projects.

l Lidar-based aerial surveying is a surveying technique which involves more precise, direct range measurements and is therefore more automated and less dependent on ground control.

In these articles we will focus our attention primarily on LBAS and we will look at

l The basic principles of LBAS

l The typical LBAS workflow

l Examples of survey projects where LBAS can be applied

l LBAS specifications to consider before conducting a survey

The basic principles of LBAS

The main reason why LBAS has only emerged as a surveying technique in recent times, is that the underlying technologies only matured and became available off-the-shelf during the past ten to twenty years. In order to understand the basic principles of LBAS, let us look at the enabling technologies:

l High-power, short-pulse laser. For accurate ranging, i.e. timing of the flight, reflection and return of a laser pulse, a very short pulse is

required; the longer the pulse, the less precise the send and receive timestamps of the laser pulse will be. Furthermore, high-power laser is required to ensure that pulse returns are received from dark surfaces (e.g. coal stockpiles, tar roads) and to penetrate through vegetation to ground level. This ability to penetrate vegetation, enables LBAS to outperform conventional photogrammetric aerial surveying which, through stereoscopic methods, can only determine the heights of the top of vegetation. Typically NdYag crystals are used as the lasing medium, producing the 1064 nm infrared laser pulses. One disadvantage of using near-infrared laser, is that it is absorbed by water and some plant materials with very high levels of chlorophyll.

l Differential GNSS positioning. The Global navigation satellite system (GNSS), and in particular the Global Positioning System (GPS), is used extensively to determine the position of the laser scanner along its trajectory, so that an accurate three-dimensional position and time-stamp can be assigned to each transmitted laser pulse. In order to increase accuracies, differential GNSS positioning is employed –

a technique involving so-called “double-differencing” to cancel the effects of the atmosphere and satellite clock and ephemeris errors experienced by the roving receiver and the base which is set up at a known ground point. (For the sake of completeness, note that a ground base station is not a strict requirement of LBAS; through the technique of precise point positioning the trajectory of the laser scanner can still be determined, although it introduces a slight loss of accuracy and necessitates a waiting period of approximately two weeks until final orbits and clock errors are published for the satellite constellation.)

l Inertial measurements. An inertial measurement unit (IMU) is used to determine the three-dimensional orientation (heading, roll and pitch) of the laser scanner, so that a direction can be assigned to each laser pulse. The IMUs used in LBAS systems are typically of military grade and specification to ensure the highest accuracy of measurements possible.

l High-end computing and data storage. During an LBAS survey, hundreds of thousands of points can be surveyed per second. Combining

Legend

Lidar Light detection and ranging

LBAS Lidar-based aerial survey

Laser Light amplification by the stimulated emission of radiation

GPS Global Positioning System (NavSTAR)

GNSS Global navigation satellite system (comprising GPS, GLONASS, Galileo and others)

IMU Inertial measurement unit

PDOP Positional dilution of precision

technicalVISUALISATION

48 PositionIT – Jan/Feb 2011

these points with aerial photography, which usually forms part of the aerial survey, one can easily record in excess of 100 GB of data per hour. This is the typical current data rate and is set to increase significantly as the technology develops. Secure data storage and high-end computing are required to transform the data to useful survey products, in a format which the end-user can use on a standard desktop or notebook computer.

The diagram in Fig. 1 depicts the basic LBAS workflow principles.

Workflow for a lidar-based aerial survey

With this understanding of the basic principles of lidar-based surveys, we now consider the workflow of a successful aerial survey.

Survey planning

The planning of the survey involves a number of considerations, including:

l Determining the area of interest or route. This is based on the maps or coordinates supplied by the client, which are usually in common industry formats such as Esri shape files, Google Earth files, Microstation DGN files, AutoCAD DWG files, text files with lists of coordinates etc.

l Planning of exactly how the survey area will be covered in order to meet the set specifications, e.g. the flying height and speed, field of view and scan rate.

l Determining suitable time-windows for the survey. In this respect, two main considerations are made: Surveys should be conducted

Fig. 1: An overview of the lidar-based aerial survey workflow principles.

during periods where the satellite constellation at the survey location is optimal, i.e. during a period when eight or more satellites are available for use and the satellites are well distributed to offer good trajectory solutions. A positional dilution of precision (PDOP) of less than 3,0 is ideal. If the lidar survey is combined with aerial photography, the sun angle during the survey should also be considered. With the sun at zenith, there will be bright reflections off flat and reflective surfaces on the imagery, while there will be very long shadows on the imagery if the survey is conducted in the early morning or late afternoon.

l Deciding whether to use pre-marks, post-marks or no ground points as the ground control of the survey. At sites where absolute accuracy is not of importance, the survey is not to be tied to a specific reference network and where ground access to the survey site is impossible, a well-calibrated lidar system can be used for the survey without any ground control points. However, most of the time ground control is used. There is no significant difference between the accuracies attained when using pre-marks compared to using post-marks; however, having the pre-mark coordinates available at the time of data analysis will decrease the processing time, since there is no waiting time between the production of orthophotos and DTMs, and the transformation of data to fit the ground control. Note that the ground control survey does not

only provide a list of coordinates for the ground control points, but also transformation parameters from WGS84/ITRF in which the lidar-based survey is conducted, to the reference frame in which the results are to be presented.

Survey execution

The following steps are involved in executing the aerial survey:

l If required, place pre-marks in locations where they will be visible in the survey data and survey their positions.

l If possible, start a GNSS base station at known coordinates on the ground. This base is to record data during the aerial survey so that its data can be used for the differential positioning of the laser scanner. The recording should be at a high rate – 2 Hz or higher – so that position solutions can be determined more often and only short positional interpolations are required between recorded epochs. Ideally the base should be at or near the survey site, since the spatial correlation of the errors observed at the base and rover (onboard laser scanner) will decrease as a function of distance between them, usually significant beyond 30 km.

l Execute the flight following the guidance of the lidar system’s onboard navigation system, taking weather, safety and all regulatory factors into consideration. The system will automatically record lidar data, positioning information and imagery along its trajectory as per the flight planning.

Data pre-processing

Before the survey data can be analysed, pre-processing is required in order to get all the laser data spatially referenced and prepared for the processing phase:

l The position of the scanner along its trajectory is determined at e.g. half-second intervals, using the GNSS data recorded in-flight. This will be done differentially i.e. relative to the GNSS base or as a stand-alone solution. Typical accuracies obtained relative to the base, are in the order of 4 cm in three dimensions.

l The trajectory positions are then improved using the high-rate (e.g. 500 Hz) acceleration measurements from the IMU. This improves accuracies relative to the base to the

technicalVISUALISATION

50 PositionIT – Jan/Feb 2011

order of 2,5 cm. Furthermore, the additional IMU measurements provide orientation parameters to each epoch.

l Laser scanning data recorded in-flight consist of time-stamps and time-ranges. Using the time-stamps, specific laser pulses are tied to specific points along the trajectory. The time-range is then used to determine coordinates of the point(s) measured by a specific pulse, based on the position and orientation of the laser scanner at the time of pulse transmission. Most systems are able to measure multiple returns from a single transmitted pulse, but most of the time only single returns are received from hard surfaces, e.g. ground, buildings, roads, etc. Multiple points are usually received over vegetation, where the first few returns would be from the vegetation above ground level, and the last recorded return will be from the ground.

This pre-process yields the so-called point cloud. If aerial photography is combined with the LBAS, the development of the imagery from the raw format used by the camera to a common format, which can be used during analysis, also forms part of the pre-processing phase.

Data processing

The raw point cloud needs to undergo a number of processes to transform it into products that are useful to the client.

l Internal calibration: LBAS systems undergo regular external calibrations, a complex process which takes place in a laboratory and by flights over a calibration test site. However, internal calibrations are required for each mission. The heading, roll and pitch misalignments of laser scanners, relative to the system’s reference point, may change slightly between the regular external calibrations. For this purpose, overlapping laser points from various flight lines are compared to each other to determine the optimal misalignment parameters for the project. These are applied to all flight lines and overlapping points are cut so that the resulting point cloud will be both internally consistent i.e. the coordinates of a single position measured from various flight lines match each other and has a consistent density, so that the point density is not significantly higher in areas of overlap than areas without overlap.

l Point classification: Analysis software suites are able to automate the process of point classification. Depending on the requirements of

the client, points can be classified to ground, vegetation, buildings, power lines, roads, etc. Although the classification process is automated, it is not flawless and requires some human intervention to ensure that classification has been done correctly.

For the sake of completeness, and because LBAS is most often accompanied by aerial photography, the data processing workflow for orthophotography is also included here:

l Ground model production: In order to analyse the photography, a ground model is required which is simply generated from a sub-sample of the points classified as ground in the aforementioned step.

l Image projection: Each image has a time stamp assigned to it and can thus be linked back to a specific position and orientation along the system’s trajectory. Using this information, as well as the camera’s focal length and other lens parameters, each pixel from the camera can then be projected onto a position on the ground.

l Internal calibration: As in the case of the laser scanner, the heading, roll and pitch misalignments of the camera are determined by regular calibration flights, while lens distortion parameters are

Fig. 2: A classified point cloud shows vegetation (green), ground (yellow) and buildings (red), as well as the intensity of returned pulses, in three dimensions for a section of Rosebank, Johannesburg.

technicalVISUALISATION

PositionIT – Jan/Feb 2011 51

determined in a laboratory. Since the misalignment parameters may change slightly between external calibrations, pixels are selected on photos which match pixels on overlapping photos. These tie points can then be used to determine an optimal misalignment for the project and each photo is then adjusted so that overlapping pixels from various images match each other spatially.

l Radiometric calibration: Lighting conditions may vary during flight execution, leading to slight variations in brightness, contrast and colour balance between adjacent imagery. Images are radiometrically adjusted so that colours are balanced at the seam line where adjacent images meet.

l Seam line manipulation: The boundary, or seam line, between images is placed so that the most vertically downward looking section of the image is used. However, this could be running through a structure such as a building or tree, which would have a different perspective from adjacent imagery. In order to improve the data aesthetically, seam lines are moved so that whole trees, buildings or other structures are displayed as seen from one image only.

The last step during data processing is the application of a number of corrections to the data, both the imagery and laser data:

l All the aerial survey data are based on WGS84 positions, as received from the GNSS satellites. This implies that the elevations from the LBAS measurements are all ellipsoidal heights. In order to transform these to orthometric heights, a geoid model is used. There are a number of these available, e.g. the global models EGM96 and EGM2008 or the SA2010 model for work conducted within South Africa. Also, more accurate local models can be generated by precise levelling on the ground.

l Furthermore, if the coordinate system to be used for the survey is not based on the WGS84 ellipsoid and datum – such as the local datums used in many locations or the Clarke 1880 ellipsoid and Cape Datum used in South Africa before 1999 – a local Helmert transformation is required to transform the data to the desired projection. The transformation parameters, mostly translation but often also scale and rotation, are determined by the ground control, which should include a number of points around the survey site being measured in both the required coordinate system and WGS84.

l After the application of any required geoid corrections and transformation parameters, the data set is compared to ground control points. This is a quality control step to ensure that

the aerial survey data meet the set requirements. If required, small horizontal and vertical shifts can be applied to match the ground control points, which are considered to be more accurate than aerial survey data.

Fig. 2 shows an example of a classified point cloud derived using the explained workflow. Similarly, Fig. 3 illustrates the high-resolution photography that can be combined with LBAS-derived ground models to generate accurate orthophotos.

Summary

In this article the reader was introduced to lidar-based aerial surveys (LBAS) as a surveying technique and the basic technologies and workflow of LBAS were outlined.

After the data processing workflow, the survey data can be handed over to the client, an external specialist consultant or an in-house expert to produce useful deliverables in order to fulfil the exact purposes of the survey.

In Part 2 of this article we will look at typical LBAS deliverables and applications. We will also consider the specifications of an aerial survey that a client needs to understand prior to commissioning surveys.

Contact Adriaan Combrink, CK Aerial Surveys, Tel 011 949-8905, adriaan@ckas.co.za

Fig. 3: High-resolution aerial photos are projected onto a ground model to generate accurate orthophotos, like this photo taken in August 2008 of Soccer City, Soweto.