Using aerial imagery is the most widely used approach in making maps with OpenStreetMap. Mappers typically use Bing or Mapbox satellite imagery as a background layer, or imagery from another provider. We have already seen this in action. In this module we will learn a little bit more about aerial imagery, and we will learn how to solve the one important problem of using aerial photographs - imagery offset.
Aerial imagery is the term that we use to describe photographs that are taken from the sky. These can be taken from airplanes, helicopters, or even kites and balloons, but the most common source of imagery comes from satellites orbiting the Earth.
In the module on GPS we learned about the dozens of satellites orbiting Earth which allow our GPS receivers to identify latitude and longitude. In addition to these GPS satellites, there are also satellites which take photos of the earth. These photos are then manipulated so that they can be used in GIS (mapping) software. Bing Aerial Imagery is made up of satellite photos.
Let’s look at a couple of the components of aerial imagery - resolution and georeferencing.
All digital photographs are made up of pixels. If you zoom in very close on a photograph, you will notice the the image starts to get blurry, and eventually you’ll see that an image is made up of thousands of little squares that are each a different colour. This is true whether the photograph is taken with a handheld camera, a mobile phone or a satellite orbiting Earth.
Resolution refers to the number of pixels wide by the number of pixels high that an image is. More pixels means higher resolution, which means that you are able to see greater detail in the photograph. Resolution in handheld cameras is often measured in megapixels. The more megapixels your camera is able to record, the higher the resolution of your photos.
Aerial imagery works the same, except that we talk about resolution differently. Measurement is important with aerial photographs - hence, a pixel represents a certain measurable area on the surface of the Earth. We usually describe imagery as something like “two metre resolution imagery”, which means that one pixel in the image is equivalent to two metres on the ground. One metre resolution imagery would have a higher resolution than this, and 50cm resolution would be higher still. This is generally the range of imagery that is provided by Bing, though it varies between locations, and in many places it is worse than two metres, at which point it becomes difficult to identify objects in the image.
The higher the resolution of an aerial image, the easier it is to use in making maps.
Every aerial image is georeferenced, meaning that it is manipulated so that it can be shown in its correct location on the Earth. Georeferencing is a relatively complex process, because images are flat and the Earth is round, and the images need to be positioned and stretched so that the pixels are accurately positioned on the planet.
The imagery available to us is already georeferenced, so it is not something that we need to concern ourselves with too deeply. We can happily use imagery to help add to OSM, so long as we understand a little bit about the imagery we are using, and so long as we are aware of one common pitfall - imagery offset.
Imagery providers usually do a pretty good job of georeferencing their imagery, but occasionally the images can be a little bit out of position. This is particularly true in hilly or mountainous areas, where it can be difficult to stretch a flat image over an area of the Earth with many contours. When you load imagery in JOSM, it can sometimes be ten metres or more from its true position. This is called imagery offset.
Aerial imagery layers are composed of many photographs of the Earth’s surface that have been georeferenced and then stitched together. Imagery providers cannot verify the accuracy of every photo, so some images may be shifted from their actual positions. This might be a shift of a couple metres, or in rare instances up to hundreds of metres. In mountainous areas, imagery may be distorted non-linearly, which means that nearby parts of the same image may be shifted in many different directions.
Notice in the following image that two separate aerial photographs have been georeferenced and merged together. Because georeferencing is not a perfect process, the images do not line up perfectly with each other. Hence one, or both, must be inaccurate.
We’ve learned about two major ways of making maps - one is by utilising aerial imagery to identify features on the ground, and another is by using GPS to record tracks and waypoints and then add them to OSM. The advantage of aerial imagery is obvious. It enables you, the mapper, to see the whole picture, to observe various details from the image, consider your knowledge of the area, and easily trace roads, buildings and areas. One key advantage of GPS however, is that it doesn’t suffer from offset like imagery. A GPS will always provide you with a correct latitude and longitude. The only exception is when the satellite signals are interrupted by tall buildings or mountains, but in this case it is easy to recognise the error. Observe the GPS trace in this image, compared with the Bing aerial imagery layer beneath it:
Because of what we now know, it is clear that the GPS trace is likely to be accurate, and the image beneath it is out of place.
So now we must ask, “if the imagery may be out of place, how can we still use it and make accurate maps?”
The answer to the preceding question is that we can move the imagery so that it aligns with things that we know are in the correct location, such as GPS tracks. It is easy to correct imagery offset in JOSM.
The best references for adjusting imagery are GPS tracks that follow roads. And the more GPS tracks that you have to reference, the more accurate you will be able to align your imagery. Since OSM users often upload their GPS tracks to the OSM database, we can download them and use them to align our imagery.
What if there are no GPS tracks on OSM, and you don’t have a GPS? Without GPS tracks, it is difficult to align imagery. If it is a relatively empty area (not much mapping done), you might choose to simply use the imagery as it is and correct the data later. It’s better to map an area 20 or 30 metres offset than to not map it at all.
If you can positively identify the latitude and longitude of one object on the ground, you can ensure the imagery is correctly placed by following these steps:
If, on the other hand, the area has already been extensively mapped, then hopefully the previous mappers have drawn objects in their correct locations. In this case, you can align the imagery to the OSM map, but beware! Other mappers may not be aware of imagery offset, and they may have made mistakes when they mapped.
Now you know how to watch out for and correct imagery offset, but there is one major problem with this approach that we have overlooked thus far. If every OSM user adjusts the imagery differently, everybody will be mapping with slightly different backgrounds.
Imagine that you are mapping a small town, and you realise that Bing imagery is offset by 15 metres to the north. So you adjust the imagery and then use it to map the whole town accurately. But then somebody else wants to add something to the map, so they download the data and load Bing imagery, but they don’t know about the imagery offset you discovered! They will think that something is wrong and all of the objects in town are misplaced by 15 metres, and they will start to move them, which is not correct! This can be disastrous for the town’s map data.
For this reason it is important that all users are aware of imagery offset, and should always check for it before mapping an area. To help with this problem, some smart people created a plugin that allows users to save offset information in a database and share it with others. Let’s see how this works:
In the same way that you are able to save offsets as bookmarks, this plugin allows you to save offsets to a central database, and to access the offsets that other users have created. Hence, if one mapper creates an imagery offset in an area, other users can use the exact same offset to map with.
When using aerial imagery layers, you should ALWAYS check for existing offsets, and when you create your own offset, you should ALWAYS save it to this database.
Now that we have marked this user’s offset as “deprecated”, we should add an improved offset to the database.
Oh No! Somebody mapped this area with misaligned imagery, so the area is not correctly mapped. This will take some time to fix.
For more information on the offset database, you can visit the website at http://offsets.textual.ru/. This lists all the offsets that have been uploaded to the database, and it also has a map feature that visualises where the offsets are located, as you can see here:
One last thing to remember is that the imagery may not be offset the same distance everywhere! This is especially true in regions where there are lots of hills and mountains. So if the imagery seems to be offset differently in different areas, you’ll need to move it again.