Discussion: Determining scale

What is the easiest way to determine scale from an aerial photo? I have a high resolution photo but no indication of scale or height of pic. I was thinking of just pacing out 100m on the ground, seeing how far that is on the map - and going from there? Easier way?

If using a Google Earth Image - is there a way to determine scale using the elevation of the image?

Mappers?

Find a school with a track and measure the length of the straight, (or look at the lower left of the map for the scale lines in metric and imperial format :)
http://maps.google.com/

Well, if it's an aerial photo with no annotations (such as the scale bar in Google Earth) then your options are limited. Although instead of pacing it on the ground, I would recommend finding the same area in Google Earth (or Maps) and measuring a known interval with the ruler tool.

I'll assume the photo is printed in a piece of paper, since it's harder to define a unique scale for a digital image... then it depends on how you print it. Height above the ground shouldn't really matter, since that would be dependent on FOV + orthorectification, and scale is what you want anyway.

As an example, if the photo has a road on it: pick two distinctive points on the road (sharp bends, or junctions) as far apart as possible, and measure the straight-line distance between them in cm on your map. Let's say it's 14.7 cm (=0.147 m). Then, go to GE and locate the area, use the ruler tool and measure the straight-line distance between the same two obvious points. Let's say it's 2.3 km. Then your aerial photo is at a scale of 1:(2300/0.147), or 1:15,646.

Assuming the photo is orthorectified and at a size typical to O-maps (i.e., no spherical distortion), this result should be accurate to within a couple of percent.

Post back if you're working with the image digitally, in OCAD or the like. If so, meters/px is a more relevant measure.

I have a high resolution jpeg I am working with and will be using it as my template in OCAD to start a new map. Your suggestions above are kinda what I thought would be the best way. Thanks.

Meters/px? Explain please.

Use gmaps pedometer. Its the coolest website ever.

He means meters per pixel.

OK, so the photo is in fact digital. If it's a hires JPEG I'd guess it doesn't contain resolution information (and that might not help even if it did.)

The basic reason for using m/px is that it's not meaningful to ask "how many (real world) cm wide is this JPEG?" - that will depend on several factors including zoom factor, what it's printed or displayed on, etc. Since the one constant thing about the image is its pixel size, that's what to use in the ratio instead of cm. Use a ruler tool in Photoshop or GIMP or other editing software to get the number of pixels in a straight line between the distinctive features - let's say it's 2,874. Then your m/px will be (2300 m/2874 px), or 0.80.

For the type of image you're talking about, I'd expect it to be somewhere in the range of 0.5-2 m/px.

Once you know that number, you can open the template at the right real-world scale directly in OCAD (I'm still using v8, so hopefully this hasn't changed.) Do Template > Open and get to the dialog box where you enter the scales. This isn't the only combination of numbers that will work, but it's what I use:

Resolution: 254
Draft Scale: 10,000 times the m/px (in this case, 8000)
Rotation: whatever you want
Map Scale: whatever the map scale is (don't change it)

The photo should open up as a template that will be sized properly on the paper, and work with the "real world" coordinate mode.

Hope this helps.

Thanks for the feedback.

Consider a combination of a pedometer just to count steps, and creating a GPS track. Over the course of several terrain types, a good average for a given terrain can be established. I found that my stride is reduced by about 10% for easy hilly terrain vs. level ground.

I try to keep it simple, since traveling cross country, reading a map, counting steps and doing math in my head tends to slow down one of these processes. Seldom do I count steps anymore unless absolutely necessary.

Counting telephone poles could be useful. You just need to know the disstance from one pole to another.

If you've got a good clear view of the sky and a GPS, you can set it to UTM coordinates, find the easting and northing of several points you can identify on the map, subtract to get x and y distances, then use Pythagoras to find the hypotenuse between the two points. It's a lot easier than it sounds.

See also:

http://www.attackpoint.org/discussionthread.jsp/me...

Th methods given above may once have been "good enough" but they are approximate. Suppose you measure between two road junctions on the photo the accuracy may be +/- a road width. Then what are you comparing with? Pacing, well how accurate is that? Google Earth, well they stitch their photos together in a quick and dirty way, and there seems to be various types of distortion.

If an orienteering map was a self-contained entity it wouldn't matter, we treat the (single) photo as the complete and utter truth and no-one will know if the scale is actually 1:15,500 and north 3 degrees out.

Over the years I've found myself working on maps which have bits tacked on from other sources and then the fun starts. Of course the original source(s) are lost in the mists of time and you don't know what to trust. When there is a dodgy bit, what is right and what is wrong? How do I fit a modern aerial photo to it and what do my GPS readings mean?

The answer is (must be) using a trustworthy rectangular framework. In other words one of the national grids. We are lucky in NZ that we can get digital orthophotos (ie distortion-free) for about half of the country. Resolution 2.5m/pixel is a bit grainy for buildings but fine for roads. These have known coordinates at the corners. You set them up as templates in OCAD (using the "real-world coordinate" system). Then you fit everything else to them, or the road network traced from them. Small downloads from Google Earth can be fitted like tiles, minimising the problems with GE. GPS files sit in exactly the right place when you import them.

There are still errors in this process (eg the width of a road, the natural errors in the GPS process). But the errors do not build up. They are localised. You can improve one bit of map knowing that it's not making the next bit worse. Separate maps can have the in-between piece filled in as its exactly the right size. It's a no-brainer.

IMHO no map should be started without being related to the national grid. And existing maps should be progressively related to the national grid.