- Turn off automatic portrait and landscape alignment in the camera. This will no longer change the so-called main point position and you will get better accuracy.
- Switch off the automatic camera shake correction in the camera. This prevents an exact calibration of the camera during the calculation or the correct application of the calibration data if the camera is already calibrated.
- It is best to use fixed focal length lenses. These are always faster than zoom lenses and will give you better images, especially in low light conditions.
- If you use zoom lenses: Set the lens to either full stop wide angle, or full stop zoom and do not change this zoom setting anymore if possible. The full wide angle setting is normally always good to use.
Before taking pictures
- Set the aperture, exposure, and ISO number correctly: The aperture should be as high as possible so that you have well-lit, sharp and noise-free pictures, with freehand photography the exposure time should not be longer than 1/60s, otherwise the pictures can blur.
- Focus once before taking the first picture and turn autofocus off. This allows all images to be calculated with the same mathematical camera model and improves accuracy.
While taking pictures
- If possible, do not zoom or focus: this will allow all images to be calculated with the same camera model, otherwise ELCOVISION 10 will have to calibrate each image separately, which will reduce accuracy.
- Never take 2 pictures from exactly the same position. Always take at least half a step to the side between 2 images: images from the same point of view cannot produce exact points because of the flat ray intersection.
- Photograph the object with 75% overlap between 2 images: With a flight planning software for drones it is best to set a 75% overlap between the images and a 60% - 75% overlap of the stripes.
- If possible, photograph especially large objects in 2 passes: The first time you take a picture of it, the second time you take a picture of all the details you are interested in. This prevents you from accidentally forgetting parts of the object: If details are also photographed in the 1st pass, you will get bogged down and often have holes in the photo block.
All current digital cameras are equipped with very good automatic systems, which take excellent pictures under almost all lighting conditions. Nevertheless, you should have some basic knowledge about the correct exposure of images:
The selective metering (spot metering) measures the light conditions in the middle of the image area; the large field integral metering (multi spot metering) measures the light conditions in the whole image area.
Selective metering of the exposure makes sense if the area of an object that is of interest for evaluation has very different lighting conditions to the surroundings.
If there are not too great differences in the illumination conditions in a certain situation and if the entire situation is of interest for the evaluation, the integral measurement must be used for the exposure measurement. Thus, the entire object area is imaged in the best possible way.
It is important that an over-illuminated, i.e. too bright image can no longer be saved:
from an underexposed, too dark image a lot of information can be gained:
Since close-up photogrammetry normally works with fixed lenses or lenses focused at certain distances, the sharpness of the image must be achieved by selecting the aperture. The aim is always to work with as small an aperture as possible. Optimal f-stops are normally in the range 5.6 - 11, if the f-number becomes larger, the exposure times become too large or the images too dark.
But beware: For freehand shots the exposure time should not be longer than one 1/60 second (danger of camera shake). When using a telephoto lens, the longest exposure time must be 1/125 second or less.
Illumination by means of lamps always makes sense if cast shadows are to be expected or strong reflections from the object are to be expected. Reflections must be taken into account, especially in the case of factory buildings, pipelines, etc.
Dropshadows occur indoors or in highly structured objects. If lamps are used in such cases, shadows can be removed by moving the lamps.
The best natural light in the open is when the sky is slightly overcast. In this case, there are no hard shadows and the entire object is usually uniformly illuminated.
Imaging set-ups for 3D photogrammetry
- When shooting for automatic orientation, the convergence angle between 2 images should not exceed 30°, if the images are manually oriented, the convergence angle should be limited to 90°, so that points can still be easily recognized.
- Points, which should be measured three-dimensionally (X-, Y-, Z- coordinates), must be visible in at least 2, better 3 or 4, images from different points of view.
- The photos should be taken as format-filling as possible, so that accuracy is not wasted over the image scale. Grass, trees or blue sky are of no interest in most cases.
- Despite the considerations of accuracy, it often makes sense to approach the object closer than necessary. The recognizability of the object details is thereby considerably improved, the evaluation can be carried out much more detailed.
Capturing highly structured objects
Highly structured objects and large base distances often lead to visual shadows in which it is not possible to measure:
The solution here is either to go up in the air or better to cover the object completely with a more distant series of pictures and then additionally photograph the presumably hidden parts with 2-3 pictures each:
Shooting around an object corner
It is best to photograph around object corners with 4-5 images, thus keeping the maximum convergence angle of 30° for automatic orientation:
Photography of a small building or object
It is best to walk around the object and take a picture every 1-2 steps, details are best photographed with a 2nd pass:
Photography of interiors
When taking pictures indoors, the arrangement of the pictures and also the number of pictures strongly depend on the interior design of the room. This is due to the concealment caused by pieces of furniture in the room. It is best to photograph the rooms with the back to the wall, the wider the angle of the lens the better:
and then afterwards furniture and visual shadows are covered as best as possible with additional pictures.
Passage to the next room
The best way to photograph through doors is with pictures taken in pairs; the walls opposite each other then provide the connection points:
Recording of a long corridor
It is best to make pairs of stereo images that are aligned slightly to the opposite wall. If possible, the floor should be included as well, because often there are more points for the connection:
Aerial surveys: With drones, aircraft, helicopters ...
The object should be flown in strips, the images and strips should overlap by 75%. This way every point on the ground will be visible in about 6 images: The orientation of the image formation will be very stable and a very good high-density point cloud can be calculated, where any point on the object should be visible in 4 or more images.
Here is a table that shows how far a camera moves while taking a picture at different flight speeds and exposure times:
|3,6 km/h = 1 m/s||1,67cm||1,00cm||0,80cm||0,40cm||0,20cm||0,10cm|
|7,2 km/h = 2 m/s||3,33cm||2,00cm||1,60cm||0,80cm||0,40cm||0,20cm|
|10,8 km/h = 3 m/s||5,00cm||3,00cm||2,40cm||1,20cm||0,60cm||0,30cm|
|14,4 km/h = 4 m/s||6,67cm||4,00cm||3,20cm||1,60cm||0,80cm||0,40cm|
|18,0 km/h = 5 m/s||8,33cm||5,00cm||4,00cm||2,00cm||1,00cm||0,50cm|
|21,6 km/h = 6 m/s||10,00cm||6,00cm||4,80cm||2,40cm||1,20cm||0,60cm|
|25,2 km/h = 7 m/s||11,67cm||7,00cm||5,60cm||2,80cm||1,40cm||0,70cm|
|28,8 km/h = 8 m/s||13,33cm||8,00cm||6,40cm||3,20cm||1,60cm||0,80cm|
|32,4 km/h = 9 m/s||15,00cm||9,00cm||7,20cm||3,60cm||1,80cm||0,90cm|
|36,0 km/h = 10 m/s||16,67cm||10,00cm||8,00cm||4,00cm||2,00cm||1,00cm|
|54,0 km/h = 15 m/s||25,00cm||15,00cm||12,00cm||6,00cm||3,00cm||1,50cm|
|72,0 km/h = 20 m/s||33,33cm||20,00cm||16,00cm||8,00cm||4,00cm||2,00cm|
Photogrammetry differs fundamentally from other measuring methods when the achievable accuracies are considered: In principle, one can achieve any desired accuracy with corresponding effort.
Or viewed the other way around:
If you know which accuracy you need, you have to keep a maximum shooting distance with a given camera/lens combination to achieve this accuracy, or with a given shooting distance, what focal length the lens and/or what resolution the camera must have to achieve this accuracy.
Rules of the thumb
- The accuracy you achieve with ELCOVISON 10 is at natural points about the pixel size of the object.
- If points are signaled with ELCOVISON 10 targets, an accuracy in the range of 0.1 - 0.01*[pixel size at object] is normally achieved at these points, i.e. approx. 10 to 100 times more accurate than at natural points.
- The higher the quality of the camera-lens combination, the better the images, the better the accuracy.
- Fixed focal length lenses are faster than zoom lenses. With the same lighting conditions, the f-number can therefore be increased, the images become sharper and the accuracies better.
- Do not zoom or refocus while taking pictures. If this is done, the corresponding images must be simultaneously calibrated by ELCOVISION 10 and the achievable accuracy may decrease.
- If non-calibrated cameras are used, the accuracy will deteriorate by a factor of 1-4 depending on the camera and recording configuration.
Precision in aerial photography e.g. Drone images
In the case of aerial photographs, a distinction is usually made between position accuracy and height accuracy. The reason for this is that you normally have mainly vertical shots and no oblique shots.
The position accuracy is identical to the pixel size of the object, and is mainly determined by the focal length and resolution of the camera.
For practical purposes, altitude accuracy depends almost exclusively on flight altitude. The camera resolution plays hardly a role here, provided it is a normal or high-quality camera, as a rule of thumb one has an altitude accuracy of approx. 0.003*flight altitude.
Here are some tables with the typical precision for cameras with different sensors for different focal lengths and flight altitudes, Position precision is in italic script
APS-C Sensor: approx. 28*19mm; 12 MPixel: 4000*3000 Pixel
|Altitude||Height prec.||8 mm Lens||15 mm|
|10 m||0.3 cm||0.61 cm||0.33 cm||0.25 cm||0.16 cm|
|15 m||0.5 cm||0.92 cm||0.49 cm||0.37 cm||0.25 cm|
|20 m||0.7 cm||1.23 cm||0.65 cm||0.49 cm||0.33 cm|
|30 m||1.0 cm||1.84 cm||0.98 cm||0.74 cm||0.49 cm|
|40 m||1.3 cm||2.45 cm||1.31 cm||0.98 cm||0.65 cm|
|50 m||1.7 cm||3.06 cm||1.63 cm||1.23 cm||0.82 cm|
|60 m||2.0 cm||3.68 cm||1.96 cm||1.47 cm||0.98 cm|
|80 m||2.6 cm||4.90 cm||2.61 cm||1.96 cm||1.31 cm|
|100 m||3.3 cm||6.13 cm||3.27 cm||2.45 cm||1.63 cm|
|120 m||4.0 cm||7.35 cm||3.92 cm||2.94 cm||1.96 cm|
APS-C Sensor: approx. 28*19mm; 24 MPixel: 6000*4000 Pixel
|Altitude||Height prec.||8 mm|
|10 m||0.3 cm||0.41 cm||0.22 cm||0.16 cm||0.11 cm|
|15 m||0.5 cm||0.61 cm||0.33 cm||0.25 cm||0.16 cm|
|20 m||0.7 cm||0.82 cm||0.44 cm||0.33 cm||0.22 cm|
|30 m||1.0 cm||1.23 cm||0.65 cm||0.49 cm||0.33 cm|
|40 m||1.3 cm||1.63 cm||0.87 cm||0.65 cm||0.44 cm|
|50 m||1.7 cm||2.04 cm||1.09 cm||0.82 cm||0.54 cm|
|60 m||2.0 cm||2.45 cm||1.31 cm||0.98 cm||0.65 cm|
|80 m||2.6 cm||3.27 cm||1.74 cm||1.31 cm||0.87 cm|
|100 m||3.3 cm||4.08 cm||2.18 cm||1.63 cm||1.09 cm|
|120 m||4.0 cm||4.90 cm||2.61 cm||1.96 cm||1.31 cm|
Full format Sensor: approx. 36*24mm; 24 MPixel: 6000*4000 Pixel
|Altitude||Height prec.||8 mm|
|10 m||0.3 cm||0.51 cm||0.27 cm||0.21 cm||0.14 cm|
|15 m||0.5 cm||0.77 cm||0.41 cm||0.31 cm||0.21 cm|
|20 m||0.7 cm||1.03 cm||0.55 cm||0.41 cm||0.27 cm|
|30 m||1.0 cm||1.54 cm||0.82 cm||0.62 cm||0.41 cm|
|40 m||1.3 cm||2.05 cm||1.09 cm||0.82 cm||0.55 cm|
|50 m||1.7 cm||2.56 cm||1.37 cm||1.03 cm||0.68 cm|
|60 m||2.0 cm||3.08 cm||1.64 cm||1.23 cm||0.82 cm|
|80 m||2.6 cm||4.10 cm||2.19 cm||1.64 cm||1.09 cm|
|100 m||3.3 cm||5.13 cm||2.73 cm||2.05 cm||1.37 cm|
|120 m||4.0 cm||6.15 cm||3.28 cm||2.46 cm||1.64 cm|
Full format Sensor: approx. 36*24mm; 36MPixel: 7360*4900 Pixel
|Altitude||Height prec.||8 mm|
|10 m||0.3 cm||0.43 cm||0.23 cm||0.17 cm||0.11 cm|
|15 m||0.5 cm||0.64 cm||0.34 cm||0.26 cm||0.17 cm|
|20 m||0.7 cm||0.86 cm||0.46 cm||0.34 cm||0.23 cm|
|30 m||1.0 cm||1.28 cm||0.68 cm||0.51 cm||0.34 cm|
|40 m||1.3 cm||1.71 cm||0.91 cm||0.68 cm||0.46 cm|
|50 m||1.7 cm||2.14 cm||1.14 cm||0.86 cm||0.57 cm|
|60 m||2.0 cm||2.57 cm||1.37 cm||1.03 cm||0.68 cm|
|80 m||2.6 cm||3.42 cm||1.83 cm||1.37 cm||0.91 cm|
|100 m||3.3 cm||4.28 cm||2.28 cm||1.71 cm||1.14 cm|
|120 m||4.0 cm||5.14 cm||2.74 cm||2.05 cm||1.37 cm|