Photogrammetry with PhoToPlan: Precision, accuracy and camera resolution.

So I am asked the question:

“Can you measure to 0.25” (6.2mm) precision the façade panels of a tall building using PhoToPlan, as the laser scan data won’t extend reliably to the upper floors of the building, as  it is out of range at 98 m?”

The answer is ‘yes’, but only if you are prepared to use a fairly costly camera and a pretty costly 600mm telephoto lens.

I don’t consider this to be a problem, because a quality camera is something of an essential part of the surveyor’s tool kit, but the big lens is another matter, and options like rental (with or with out the photographer attached to it!)  make sense if this is a ‘one off’ situation. In my view, the cost of professional photographic kit should be considered in comparison to the cost of a total station or a scanner because a camera is a powerful survey tool.

Ideally, to reduce the effect of perspective, the camera will need to stand off at a distance of least twice the height of the building, or greater if a raised position (such as from an adjacent building, cherry picker, fire hoist, kite, balloon, UAV or mast etc.) cannot be achieved. A range of 200m will probably be too far to get the required resolution, even with a 600mm lens. By bringing the camera in closer there will need to be more photos for coverage and with a steeper tilt but it should be possible to achieve the required resolution.

Image quality depends on camera quality Use the best camera and the best lens possible. The best performance will be from a camera with a good internal focal length and good quality glass. The sensor size is important: ‘full frame’ is best. At a minimum a professional full frame (36 x 24mm) sensor DSLR should be considered.  A calibrated camera will return better results than working without calibration.

The 3 elements of potential camera resolution:The resolution of the camera is a function both the design and component quality (particularly of the optical chain)  and the photographer’s choice of exposure  settings. There are 3 key factors to consider:

1. Optical resolving power of the lens and the optical path design. The transmission of the light to the sensor is the critical thing. This is dtermined by a mix of the aperture used and the amount of internal refraction in the elements of the lens construction. The internal focal length (or ‘principal distance’) is an important consideration:  the bigger the better. The DSLR designs have a good principal distance for most purposes,  the ‘compact’ cameras less so.

2. Sensor performance. The effective performance of the sensor and its associated processor is a function of the density of pixels, the sensitivity and size of the array. Selecting a high sensitivity (high ISO settings) introduces noise into the  image and degrades the quality, using a minaturised sensor 9 as in a comapct camera) reduces image quality.

3. Exposure: Most lenses have an optimum aperture, usually about two stops down from the widest aperture. The shutter speed will be determined by balancing the lighting conditions  with the shutter speed and the ISO setting. The camera will need to be tripod mounted and the use of image stabilisation considered. Care should be taken not to overexpose the image as detail in the highlights will soon be lost.

The largest focal length lens will give the best resolution at the greatest distance. This is a bit obvious but  the magnification of a telephoto lens is the key to getting the required precision at long ( greater than 30m) range. The photo-scale is dependent on 2 variables, the distance from the subject and the focal length of the lens, if you are forced to place the camera at a distance from the subject using ‘long’ lens will maintain resolution at the cost of smaller image area and camera stability issues ( not to mention the cost of good quality long lenses)

Photo-scale and resoloution. Photo-scale is the term used to decribe the scale of an image it is calculated by:

S=f/h

where s is the photoscale, f the focal length and h the height ( in our case the distance, this like most photogrammetry theory,  is from aerial practice). To see what happens with my camera I took a shot at known distances of a test target:

These 3 shots were taken with a 70 mm lens on a Nikon D70, they show the fall off in resolution at 50m and the complete loss of the targets at 100m. The target is 60x30mm and the photo-scale at 50m is 1:1,400 and at 100m it is 1: 7,000 clearly a photo-scale of more than 1:10,000 with this camera is useless for measuring anything smaller than about 15mm. A camera with a larger sensor and a higher pixel count (this one is a 6mp 3/4 frame one) would improve the resolution at the higher photoscales.

When I ran the 50m range image through PhoToPlan

I get a report on the pixel size achieved by the rectification:

The standard deviation between the image and the control points is given as 2.6mm which could be improved with more control points but I think the flatness of the steel door or even the thickness of my blutack target fixings might be the issue here.

What we can get out of this image is down to the size of the pixels. When scaled up to the control geomtery, this works out at 5mm to each pixel. Here is the 60x30mm target as rendered in the image pixel by pixel. It comes out neatly at 12 pixels by 6:

so the reported ‘current image resoloution’, (which is the uncorrected state) is a bit confusing, it is a simple image scale:

PhoToPlan also reports a ‘Rectification Factor’:

If we take the Rectification Factor and the Curent Image Resolution (0.00021 x 1.4)  we get 0.0029m or 3mm per pixel, allowing for the tilt and the deviation of the control from rectification plane this confirms an aproximate 5mm pixel scale.

So, looking at how the pixel size at scale effects resoloution, and the effect of the focal length on photoscale, I am confident the precision required for the tower survey is achievable and that the resolution needed is possible.

By scaling up the results with the 6mp camera with the 70mm focal length to a 10 or 12mp full frame camera with a 600mm focal length lens, the stand off distance of 140m will give a photo-scale of 0.6/140= 1:4,280 with double the pixel density.

Control is crucial. All of the above is utterly dependent on achieving good quality control in the image area, and this is crucial as the 600mm lens will involve some careful co-ordination with the photographer. Usually, photogrammetric control for facade surveys is conducted by 2 point intersection from a precise network of ground points,  so that reliance on reflectorless obs might be risky given the obliqueness of the target on the upper floor of the building.

Round trip check. Using lines digitised in PhoToPlan off the 50m range image, I was able to compare measurement on the panel by steel tape. The results I obtained were 0.219, 0.215 and 0.215.A mean of 0.216m by steel tape against 0.219m, as digitised, shows the precision achieved as : +/- 3mm.This shows that ‘sub- pixel digitising’ in PhoToPlan, provided the control and phtotography are adequate, is pretty precise!

More on PhoToPlan: http://www.theolt.com/web/photoplan/

Photoscale nicely explained: http://www.geog.ucsb.edu/~jeff/115a/lectures/scale_and_area_measurement.html

 


About billboyheritagesurvey

Heritage worker
Gallery | This entry was posted in Survey Practice and tagged , , , . Bookmark the permalink.

2 Responses to Photogrammetry with PhoToPlan: Precision, accuracy and camera resolution.

  1. Ryan says:

    Informative, but telephoto lenses are not ideal to use in photogrammetry. You cannot get accurate angle measurements when the rays of light are entering the lens near parallel. Good measurements require accurate angle measurements between points. A better technique on tall buildings would be a wide angle lens, closer to the building, and get photographs 45° from vertical to the point at the top of the building from 90° apart around the building so that the angle of intersection between opposing camera positions to the tall point will intersect at 90°. In other words, the most ideal distance from the building is exactly the height of the building. This has the minimum error ellipse in all three coordinates, provided the photographs are taken from all directions.

    • That’s a fair point Ryan but experience tells me you have to work with what is possible. Camera positions are rarely where you want them for classic ‘normal’ case cover. I disagree with your assertion on wide angle lenses: they are prone to aberration.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s