UAV (Unmanned Aerial Vehicle). First Greek results.
The recent method of UAV photogrammetry is tested in a real case in Greece with PHOTOMOD photogrammetric suite. The method is very new to Greece and the results of horizontal accuracy of 4cm (=1 pixel on the ground) and 10cm in height is very promising for many applications. Some further processing with PHOTOMOD is also presented.
A new photogrammetric method has a wide spread recently. The method uses photogrammetric procedure using digital photographs taken with ordinary digital cameras from small UAVs and the use of photogrammetric software. This article describes the method and the results of it.
UAV Capture of the photograph
The capture of the photographs is done by helicopters (with one, three or more rotors), flying wings or small unmanned airplanes. In the Greek market one can find such models. These models are usually powered by electric motors, they have a duration of flight about 30 minutes and apart from the camera they carry an electronic auto pilot circuit, a small GNSS receiver and inclinometers. The flight programming is done through an ordinary laptop. They usually fly at 100m to 1000m high and the ground sampling distance (GDS) at these highs is 4cm to 35cm. In a half hour flight they can take from 500000 m2 to 4800000 m2 with 60% overlap and sidelap with a spee of 10m/sec. These values are indicative and they vary for other speed and overlap.
This is done by a photogrammetric software. Notice that photogrammetric processing of small UAVs is very demanding because
1. The photographic camera is not a common camera with big distortions an usually without stable calibration
2. the coordinate and the rotations of the photographs are very rough
3. big rotations are observed from ont take to the next (not only because of the wind)
4. many times the photographs are blurred.
In practice the procedure seems like a toy but it is not. It requires experience and good knowledge of photogrammetry to get decent results.
The procedure was tested in real world with the following parameters.
The region was a typical Greek village at an area of 200000 m2. The photographs were taken with a Canon Ixus 220HS 12mpixel (4000x3000 pixels) camera, on a SwingletCam flying wing from a 100m height. The nominal focal length of the camera is 4.3mm and the pixel (GSD) was 4cm on ground.
From the mission 36 photographs were used.
For pre marked ground control points (GCPs), five A4 paper pages were used and their coordinates were measured in the Greek geodetic network using RTK from the permanent Hellenic Positioning System (HEPOS). We must admit that the size of the targets was big for this GSD.
In the picture a part of a photograph is shown.
For the photogrammetric procedure PHOTOMOD was used. This software has particular tools for UAV
- for reading the projection centers data,
- for automatic tie point measurements (relative orientation)
- for automatic block adjustment,
- for camera calibration
The procedure was as follows.
1. The photographs were read.
2. The camera data (focal length, pixel size etc) were introduced and automatic interior orientation was performed
3. The projection center data (coordinates and rotations) were read and interpolated automatically if missing. The software automatically oriented the photographs in 5 strips as shown in picture 2.
Picture 2. The block of photographs oriented automatically by PHOTOMOD
4. The special routine of automatic UAV triangulation was run and a firs absolute orientation was accomplished.
5. A precise automatic relative orientation was done with maximum permitted vertical parallax of 0.5 pixel and of 1 pixel on triplets 1 in horizontal plane and 2 pixels in height. PHOTOMOD computes all possible stereomodels. The accuracy of relative orientation of stereopairs was of an order of 0.25 pixel RMS.
6. The 5 GCPs were measured. The position of the GCPs is shown in picture 3
Picture 3. The position of the ground control points Η
In PHOTOMOD all kinds of stereoscopic measurments are possible from simple automatic correlation of points to a complete stereo comparator a shown in picture 4
Picture 4. The stereo comparator of PHOTOMOD
Stereoscopic observation in PHOTOMOD is done either through a graphic video card with Nvidia quadro fx chipset with special page flipping glassesκαι on a 120Hz monitor or with simple anaglyph (red -blue) glasses (even on a laptop).
7. A block adjustment was performed next. After the filtration of erroneous points, a new block adjustment was done but this time with auto camera calibration activated.
8. With the new camera parameters a new interior orientation was done for all the photographs and a new block adjustment was performed
9. After one repetition of the previous step the final adjustment was accomplished. the accuracy of the final adjustment is shown in picture 5. The RMS, the absolute mean and the maximum errors are shown.
Picture 5. The accuracy of the block adjustment
From this picture it is obvious that the RMS error of the GCPs is about 1 pixel (0.04m) horizontally, while the relative accuracy of tie points is much smaller.
Vertically as expected, accuracy declines and in this case was 0.10m i.e. 2.5 times the size of the pixel
This accuracy is good for many tasks like surveying, cadastral mapping, ground volyme measurements etc.
If the camera is calibrated the steps of auto calibration are not needed.
After the block aerotriangulation adjustment the photographs are ready for further processing in PHOTOMOD like
1. Automatic or semiautomatic DTM capture with automatic filtration of buildings and trees, capture of brake lines, Calculation and on the fly correction of ΤΙΝ, contour lines computation and DEM computation.
2. Stereoscopic compilation.
3. Orthophoto production.
4. 3d city model construction (with facade pictures stretched)
Some pictures of the several processes are given below:
Picture 6. Digital Terrain Model (DTM) after building and trees filtering
Picture 7. Triangulated Irregular Network (TIN) calculation
Picture 8. Digital Elevation Model (DEM)
Picture 9. Orthophoto of the village
Picture 10. Stereo compilation of details
Picture 11. 3D city model (with no facades since they were not available)