After the flight is complete, all the images captured by the camera are transferred to a powerful desktop computer and imported into our photogrammetry software. We use state-of-the-art Pix4Dmapper Pro photogrammetry software supplied by Pix4D SA.
Pix4D begins its processing by roughly aligning the set of images using the embedded GNSS data; this georeferencing data is provided by the airborne GNSS, which is not survey grade and so contains significant positional errors. However, this is adequate to allow Pix4D to create an initial rough flight path for the images.
In the next step, Pix4D examines sets of adjacent images for ground features occurring in the image overlaps – these features are used to more accurately tie the images together into a coherent and consistent orthophoto. Pix4D uses a very large number of ground features to create a network of ‘tie points’ to link the images; there may be thousands of tie points linking any given pair of adjacent images. At the same time, Pix4D is using relief displacement to calculate the elevation of the tie points.
The screenshot to the right shows Pix4D after generating the tie points for a small section of an agricultural land survey. The area shown covers around 7 ha, and contains more than 9 million tie points.
Once the tie points are established and a coherent composite photograph has been created, Pix4D is able to correct for inaccuracies in the georeferencing of the individual camera positions. This process removes much of the random error that arises from using a non-survey-grade airborne GPS. The greater the number of images acquired, the better are the statistical corrections that can be applied to the image georeferencing (however, ground-surveyed control points will remove all of these georeferencing errors in a later processing step). The screenshot to the right shows a sequence of images together with their camera positions as reported by the airborne GNSS system (blue dots) and the corrected camera positions computed by Pix4D (green dots). The red lines show the virtual light rays connecting a single tie point on the ground to its corresponding images.
With the composite photo created and tie points established, Pix4D applies the accurately-known positions of the Ground Control Points to scale and position the entire dataset correctly, relative to the required geodetic system. It is then ready to create a regularly-spaced 3D point cloud representing the elevation of the surface (the digital surface model), a 3D mesh, a contour map, and an orthophotomosaic of the survey site.
The huge quantity of data gathered during a typical survey flight requires considerable processing power to generate a DSM. It is not practical to carry out the photogrammetric data processing in the field with a laptop, so we use a powerful desktop computer at our office. Processing can take several hours, so we normally run the data overnight to be ready by the following morning.