Ariel error distributions, normal or not?

[Col]

For orthophotos with a ground model and an stated accuracy of +/-1500mm, are the errors in the ground model normally distributed? Has there been any research on this? If so what spreads are usual?

 

[carlorom]

This is not really my field but off hand I would say that the errors would not be normally distributed. First of all normally to what? The centre of the photo?

There are systematic errors in producing orthophotos. Usually orthophoto accuracy is a factor of the variation in topography (flat terrain versus hilly areas). See response elsewhere on this forum regarding importing maps etc.

I have put the basic question to our local university and the reponse was that more details are needed as to the actual process used to produce the orthophoto and factors as mentioned earlier. Be careful there are many packages today which will produce orthophotos but the process and accuracies obtained are suspect. So check the source of data too.

I would generally say what accuracy do you require and then work back from that to establish if the photo you got gives the correct answer. If not possible to work out in theory I suggest that a few very disparate and different types of points on the orthophoto be maesured and then verified in situ by actual survey. This will yield some idea of the reliability of the orthophoto.

 

[Col]

Thanks for the reply. I was wondering for example if errors from the actual level of points over a DTM with millions of point are normally distributed about the actual value or some close tolerance to this value, i.e. for a limit of accuracy of +/- 1500mm what percentage of errors approach this limit, assuming normal distribution quite a small percentage?
I think research on this could lead to some useful factors which could be incorporated in more accurate DTM production.