Orthophoto and DTM

The basics of aerial photogrammetry

How orthomosaic and Digital Terrain Model from UAV are created

Orthomosaic, orthophoto, orthophotograph or orthoimage – cartographic product depicting a selected area (usually on Earth), generated on base of processed aerial photographs, placed in reference to selected coordinate system.

To create an orthomosaic, we need to meet certain conditions of photogrammetry. The tool used in photogrammetry that becomes more and more common is Unmanned Aerial Vehicle (UAV, UAS, so called drones).

The planning of photogrammetric flight

The first step is to plan the photogrammetric mission. The range of UAV must be considered, as well as the sensor (camera) mounted on the drone (mostly the size of matrix and parameters of the lens), required Ground Sampling Distance (GSD) as well as front and side overlap of the photographs.

The range of work determines the selection of the drone: in case of small areas (single aggregate piles or deposits) it is easier to use multirotor drones, as they have practically no restrictions concerning take-off and landing area. In case of bigger assignments, it is better to use fixed wing airplane because of its flight range.

Fenix RTK Colidrone
Unmanned Photogrammetric System Phoenix

The next step is the selection of the camera. The bigger the sensor, the more detailed the mosaic (GSD is smaller), and the shorter the flight. Very important is the selection of the lens. From the mathematics, we can deduct that when we have bigger height differences of the terrain, it is better to use a lens with longer focal length (f>40mm). On a flat terrain, the focal length can be shorter. The choice of sensor and lens is strictly combined with the flight height which further determines the size of the Ground Sampling Distance (i.e. the size of every pixel). At a fixed GSD, the flight level grows in proportion to the length of the lens and the size of the sensor.

In the end, we need to define the longitudinal overlap (the percentage of the overlap of consecutive photos in a row) and lateral overlap (the overlap of the photos between the rows). The overlap depends from the end product. For the orthophoto, 70% longitudinal and 50% lateral overlap is enough. It we want to generate the precise terrain model, it is advisable to augment the overlap even to 80%/80%.

dobór parametrów lotu UAV
Mission planning in the Mission Manager
fotopunkt colidrone UAV
Ground control point

Preparation on the spot

The next step in the process of creating an orthophoto is the preparation of Ground Control Points in the area of our flight. This step consists of creating and measuring of points that will be visible and recognizable on the aerial images. In practice, the control points are painted with white spray paint or characteristic points like manhole covers are used. The measurement of the points is conducted using surveying techniques.

Safety of unmanned flights

With the flight planned and control points measured, we can proceed to the photogrammetric flight itself. At this point let’s leave partly the surveying and let’s look into aviation. Whenever we fly with a drone, we operate in airspace and therefore we need to stick to its rules. It is more complex than road traffic, because we have to consider not only turns but also altitude. The airspace is divided into controlled and uncontrolled zones.

strefy lotnicze nad Polską

In the uncontrolled ones we can fly without additional procedures, we just need to observe the sky. In the controlled zones there are areas, where the flights are permitted after contacting the managing authority, where the flights require special permits and those, where flights are prohibited. Because an unmanned aircraft is the last in the “hierarchy”, before the flight we need to control airspace in the area where we want to fly. If the area is free, we just need to select the place for take-off and landing and fly. The preparation of the aircraft and the flight itself are covered here.

rzut ortogonalny fotogrametria
Elements of external orientation of a photograph (source: J. Bernasik „Wykłady z fotogrametrii i teledetekcji”)

From aerial photos to digital terrain model

After the photogrammetric flight, we receive photos and data from GPS and IMU (inertial measurement unit) which inform us about the location where the photo was taken and the deviation of the camera in three axis. In photogrammetry, this data is called the elements of external orientation of the photograph. The data is then processed using dedicated software*.

The matching – automatic search for points showing the same piece of area on two or more photos is the first step of the digital processing. The programmes like AgiSoft Photoscan, Pix4D, 3D Survey use different algorithms for the same task. To make matching faster and more precise, we can introduce the parameters of camera’s calibration – the so-called elements of internal orientation. These parameters are the focal length of the camera, the coordinates of the main point in the photo, the size of pixel and distortion. These parameters can be found in the camera’s calibration certificate. If we don’t have it, we should introduce approximate values. Furthermore, the process is faster if we have exact elements of external orientation, for example from additional GPS receiver on board of the aircraft. After the matching process is finished, we receive a cloud of matching points. The next step is the aerotriangulation connected with autocalibration of the camera.

matching punkty wiążące ortofotomapa dron
Matching points (own materials)

Aerotriangulation is a process based on the collinearity equation. In simple words, it is an iterative mathematical process (i.e. based on multiple rounds of analysis), in which the elements of external orientation of the photograph are exactly calculated. This process wouldn’t be possible if it were not for the matching points and ground control points. Based on them, the equation with the unknown is solved. Thanks to the matching points, the photos are connected together and thanks to the control points we can locate the future model in a certain coordinate system.

Aerotriangulation relates to the autocalibration of the camera. In this process, the exact elements of internal orientation are determined – i.e. the focal length, the main point of the photo and distortion. Of course, it’s better to use cameras with predefined elements of internal orientation – those which have calibration certificates, in other words – metric cameras. But as long as UAVs use non-metric cameras, the process of autocalibration is necessary. Otherwise, the calibration would need to be conducted after every landing (due to the construction of the elements of the camera).

Knowing the precise location of every photo in space and their mutual orientation, we can pass to the process of constructing the Digital Surface Model (DSM). This technology doesn’t allow the construction of Digital Terrain Model – that is differing between foliage, surface and buildings. The LiDAR technology allows that.

But how is it possible to receive a 3D model from 2D photographs? It is thanks to the stereoscope effect. The same as in 3D movies, thanks to two cameras filming the same thing at slightly different angle (you can see that when you take off the 3D glasses) the viewer has the impression that the object “sticks out” of the screen. The image processing programs catch that “object sticking out” and record its location. The algorithm processes all the photos in that way, creating the Digital Surface Model.

efekt stereosokopowy
Artificial stereoscope effect (source: www)
The influence of height difference on the photo’s scale (source: www)

Orthomosaic

Only now can we come back to the orhophoto. What makes it different to a regular photo is that it is cartometric – we can make length and area calculations with a specific, high accuracy. The aerial photos become orthophotos after the process of orthorectification. This is what the digital terrain model is necessary for. Because a regular photo does not take height differences into consideration, it is subject to scale error. The orthorectification process consists of matching points from the model and the photos. In other words, changing the multiview projection into orthographic projection.

rzut ortogonalny fotogrametria
Orthorectification of an aerial photograph (source www)

At the end, multiple photos are connected into one orthophoto (othomosaic). We need to underline that not every photomap from a drone will be an orthophoto. Only following certain rules will let us receive a complete surveying product.

*The process described on the example of AgiSoft programme.

Kamil Kaczorowski