Leiderdorp 4th February 2014
Skeye BV, a Dutch company specialised in aerial photogrammetric surveys using unmanned aircraft and multicopters conducted a survey of a breakwater in IJmuiden. The survey was carried out simultaneously with an underwater bathymetric survey by our sister company Deep BV from Amsterdam. The two datasets were merged to create a detailed and highly accurate 3D model of the breakwater.
The port of Amsterdam is connected to the North Sea by the North Sea Channel. This channel meets the North Sea at the town of IJmuiden. To protect the mouth of the channel two large breakwaters were constructed. The breakwater is regularly monitored by the ministry of infrastructure to see if any of the blocks on the breakwater are displaced by the force of the waves that batter these piers. The normal technology to do this monitoring is using helicopter based LiDAR. To check if the breakwaters decreases in height by material flushing out from underneath, sprit level measurements are carried out yearly. Skeye conducted this survey using an unmanned aerial multicopter equipped with a photo camera.
Drone Operations in The Netherlands
In the Netherlands a commercial operation that involves an unmanned aircraft is forbidden unless an exemption is obtained from the Ministry of Infrastructure. These exemptions are given either per individual project or as a full company exemption. Skeye is at present the first and only commercial operator of drones in the Netherlands that is in the possession of a company exemption. In order to obtain this exemption all aircraft must be tested for airworthiness, be insured for third party liability and registered. In addition the pilots must hold a valid BNUC-S license and operate according to an operations manual that has been audited by the Civil Aviation Authorities in the UK and the Netherlands. However even with an exemption it is not possible to operate in all locations at all times. Prior to any operation other airspace users must be notified and the province requires an additional permission for a take off and landing outside an airfield.
Skeye mobilised a two-man team to the breakwater of IJmuiden in August 2013. The first task was to set out ground control points. These are circular disks that are placed on the breakwater and whose position is accurately measured using RTK GPS equipment. The reason for placing ground control points is that contrary to an aerial survey using a manned aircraft the position and orientation of a UAS is only roughly estimated. The on-board GPS only receives the GPS L1 signal and the internal motion sensor gives the orientation of the UAS to one-degree accuracy. This is sufficient to let the UAS take photographs at pre-determined positions but nowhere near good enough to create an accurate 3D model. The ground control points were surveyed using a 72 channel Novatel Frog GPS that uses the commercially available ’06-GPS’ correction signals. The estimated accuracy of these points is 2 cm in X, Y and 3 cm in Z. In total 20 Ground control points were used over a distance of 1000 meters. On the breakwater 35 marked survey points were identified that are measured yearly using a total station and spirit levels. These check points were also measured using the RTK GPS in order to check the quality of the result of the survey after processing.
Once all the ground control points were established the data acquisition using the UAS could start. The UAS was pre programmed to collect aerial images with at least an 80% forward and side overlap. The UAS used was a Microdrone MD4-1000. This UAS is especially suited for aerial surveys because it can stay airborne for around 45 minutes. The UAS flew three parallel lines over the breakwater at a height of 50 meters above ground level. The camera on board was a 24 MP Sony NEX7. In total 433 images were acquired. The survey had to be carried out at low tide so as to make sure that as much information as possible was collected. This was important since the collected heights would be merged with the data set that was collected at high tide using a multibeam echo sounder operated by the Hydrographic survey company Deep. In this manner an ‘overlap’ was created and which could be used to check each dataset relative to the other.
Prior to and once again immediately after the data acquisition the camera and lens were calibrated to model the imperfections. These values are subsequently used in the data processing. The first step in the data processing is find terrain features that are present in multiple images so as to tie these images together. Modern-day software uses image-matching techniques and for this survey the software was able to identify around 4000 points per image. The next step is to accurately mark the centre of the grounds control points in the aerial images. Based upon this information the exact position and orientation of where the images were taken can be reconstructed. At this point the software also gives an indication of the accuracy obtained. Subsequently the software will try and match as many pixels in adjacent images and calculate the terrain height associated with these pixels. This is not possible for every pixel but in this case an average of 96 points per square meter was obtained. The result is a dense 3D point cloud from which a digital elevation model can be created. All images were then used to create a seamless orthophoto mosaic whereby every pixel is corrected for terrain heights, scaling and corrected for colour. The final orthophoto mosaic was created with a ground resolution (GSD) of 2cm per pixel. After processing the digital terrain model was integrated with the underwater model created by Deep. In the tidal zone it could be observed that the two data sets matched perfectly. The result was a seamless 3D model.
The previously mentioned 35 check points were then compared with the location of these points in the orthophoto and their height from the digital terrain model. From this comparison it could be seen that an average planar accuracy of 14mm in X and 13 mm in Y was obtained. For the elevation the obtained accuracy was 13 mm. To further check these results the accurate spirit levels of the check points were supplied by the Ministry of Infrastructure and this showed that the obtained accuracy was even better than expected. This was a mere 11 mm for the elevation. (see table below)
The approach of using a UAS to create a 3D model showed that this can be achieved with extremely high accuracies. In fact the obtained accuracies are somewhat surprising as these are higher than expected due to the fact that the ground control was measured with 2 to 3 cm accuracy using RTK GPS. One can only guess what the results would have been if the ground control were to be placed and measured using a total station and a spirit level.