Georegistration is the process of adjusting a raster image or vector drawing to match the coordinate system and conformation of a known-good reference window that contains one or more spatially accurate layers. Georegistration is also known as georeferencing. Images and drawings in Manifold are components of a project. Georegistration works the same way for either images or drawings, so in this documentation we use the word component to mean either raster image or vector drawing.
Georegister an image or drawing:
Beginning users will do the above process in a simple way. For example, beginners tend to mark all the control points they will use in the source window, and then they will switch to the target window and mark all corresponding control points in the same order. That is simple and reliable workflow.
Control points can be marked in the target window in whatever order desired by first clicking the desired control point in the Register pane list and then clicking the location for that control point in the target window.
Experienced users tend to go back and forth between marking control points in the source window and marking control points in the target window, panning and zooming each window to find distinctive features that appear in both windows. When marking control points in the target window, experienced users know how to pick the control point the control point in the Register pane list that they intend to mark, so they can mark control points in whatever order is most convenient for them. Experienced uses also know how to remove control points they might have marked inaccurately in either the source or target windows, and how to add more control points in regions where the preview shows lower georegistration accuracy. Experienced users will also often experiment with different georegistration methods.
None of the above is rocket science, but it does require reading the Register pane topic to learn how to use various capabilities that make life easier.
Georegistration is simple: we place control points on distinctive features in the component to be georegistered, and then we place control points on those same distinctive features in a known-good component. When we press Register, the system adjusts the component to be georegistered, casting it into the known-good coordinate system, so the control points in the georegistered component line up with the corresponding control points in the known-good component.
The known-good window contains layers that have correct coordinate systems, are located in the correct geographic location, and accurately show features of interest. Such layers could be web served layers such as Bing, Google, or OSM layers, or any other image or drawing layers that are known-good.
When adjusting a source component to match a known-good target window, the georegistration process will usually resize, rotate, and otherwise distort the source component, as if it were a malleable rubber sheet, in addition to casting the spatial data it contains (geometry for vector drawings and the locations/sizes of pixels for rasters) into the coordinate system of the known good target window. That makes it possible to georeference somewhat inaccurate components, such as scanned images of hand-drawn historical maps, into a correct geographic context for use within GIS.
Keep in mind that windows, like a map window, can use a coordinate system that is different from the coordinate systems used by layers within the window. When layers in a map window use coordinate systems different from the map window, they will be reprojected on the fly into the coordinate system used by the map window. For example, a map window may use Latitude / Longitude projection while the layers within the map window, such as Bing streets or Google satellite, might use Pseudo-Mercator projection.
If a map window uses a coordinate system that is different from the coordinate systems used by layers within the map window, the coordinate system used by the map window is what will be used for the georegistered result. That is a great convenience, because it allows us to use "known good" layers, like Bing streets, which are in Pseudo-Mercator, to guide the placement of control points to georegister a vector layer, but within a map window that uses Latitude / Longitude, so that the result will be a vector layer that is georegistered into Latitude / Longitude projection. Latitude / Longitude is often a convenient projection to use for point data, because it is easy to read out the coordinates of points and other locations in decimal degrees of latitude and longitude.
Georegistration works by aligning visible features in the source component with corresponding, visible features in the target window. Features to be used are marked by placing control points at their position. Control points are marked with a cross cursor giving the position and a label giving the name of the control point.
To place control points in a window, with the focus on the window we choose Edit Control Points for the cursor mode button. We can then click into the window and each click will place a control point.
For example, we open the component to be georegistered in a source window, and then we use the Edit Control Points cursor tool to mark distinctive features visible in the source window with control points. The system will automatically provide each new control point with a unique name.
The illustration above shows six control points that have been placed in the source window, at locations such as joints between concrete and asphalt surfaces or the corners of buildings, that we know are also distinctly visible in the target window.
We then open the known-good, target window, and we use the Edit Control Points cursor tool to mark features visible in the target window that correspond to control points at the same locations in the source window.
The georegistration process adjusts the source component, the drone photo in the illustration above, so that control points in the source component end up in positions as close as possible to the corresponding positions marked for those control points in the target window, the Google satellite layer. In the illustration above, features that are visible in both windows, such as joints between concrete and asphalt surfaces or the corners of buildings, have been marked with control points in the source window and at corresponding features visible in the target window.
The orientation (rotation) of either window does not matter, nor does the starting coordinate system, if any, assigned to the source component: Manifold automatically will transform the source component as necessary to make as close a match as possible to the target, doing all necessary coordinate system transformations. In the illustration above, for example, the source component must not only be scaled and warped to match the destination, target, coordinate system, it must be rotated as well. Manifold will do all that automatically, as guided by the control points in each window.
We manage control points using the Register pane. Choosing Edit Control Points in the cursor mode button for a window will automatically pop open the Register pane if it is not open or hidden in the tab strip. When we place a control point in a source or target window, the list of control points in the Register pane automatically will be updated.
The illustration at left above shows the Register pane with a focus on a source window to which six control points have been added. The topmost box reads (current window) because the control points listed are for the current window, and are not brought in from an associated source.
The illustration at right above shows the Register pane with a focus on a target window, for which a source window (a drawing window) called EV_001 has been picked. The control points listed are those created in the source, EV_001 image window. All of the six points have had a matching control point added in the target window, indicated by a cross symbol for those seven control points in the list.
When we press Preview or Register in the Register pane, Manifold uses mathematical algorithms to warp, reposition, rotate, and re-project the source component so the control points we have marked in the source window line up as best as possible with the corresponding control points in the known-good, target window.
We can press the Preview button to get a preview of what will happen using the control points we have marked in the source and target windows. The preview will appear using blue colors to show how the source component will be georegistered into the coordinate system of the target window. Based on what we see in the preview, we might decide to add more control points, to use different control points that better cover the source and target windows, or to choose a different algorithm from the available list of georegistration methods. .
When we like what we see in the preview, we press the Register button to georegister the source component, saving the georegistered result to a new image or drawing. Manifold always writes the georegistered result to a new component, so the original source component is never altered by accident.
The illustrations above show georegistration of a raster image. The procedure for vector drawings is exactly the same.
See further discussion of georegistration, with step by step procedures and detailed information on Register pane controls in the Register topic.
Use ground features for control points - A frequent desire when georegistering images or drawings is that they line up with background layers taken from Google, Bing, OpenStreetMap, or some other web server layer.
Whether or not Google imagery is accurate is often beside the point for many users, since customers and other users will often think our work is wrong if it does not line up with Google. For that reason, a very popular choices for a "known-good" target component is to use either a Google or Bing satellite image server layer in a map as the target component.
That is fine, but very often Google and Bing satellite imagery is not nadir imagery, that is, it is not imagery that was photographed by the satellite from directly overhead the scene shown. Google and Bing satellite imagery is often off-nadir, that is, photographed from one side or the other, so that we can see the sides of buildings, like in the off-nadir satellite image of Manhattan seen at right.
With off-nadir imagery it is a mistake to mark control points at any locations other than ground level. To see why, consider the two control points marked on the image at right, both of which mark locations on the same corner of the building. The difference is that the P 1 control point is marked on the corner at a higher floor, so the location of that control point is skewed from the location of the P 2 control point that marks the corner of the building at ground level.
Because of the off-nadir angle, the location of the P 1 control point is many meters from the corner of the building at ground level. If we were georegistering a CAD drawing of building footprints, using the P 1 control point instead of the P 2 control point for the corner of this particular building would shift the georegistration of the CAD drawing many meters to the North from the actual location of the corner of the building footprint.
Note that if the Google image were a nadir image there would be no part of the building's sides that would be visible. Marking the corner of the building at a higher floor would be directly above the corner of the building at ground level.
Therefore, when using off-nadir imagery as the target component for georegistration, always use locations at ground level, such as street intersections and similar locations.
5 Minute Tutorial - Georegistration - In just five minutes we learn how to georegister (gereference) a vector drawing with an unknown coordinate system to a known-good map. Georegistration is a key capability that allows us to cast raster images and vector drawings into geographic context, so they can be used as GIS layers in maps. We can georegister aerial photos and drone photos, scan paper maps and georegister those for use in GIS, we can georegister CAD drawings, and we can rescue vector drawings and raster images that once had coordinate systems but were published in formats that failed to preserve coordinate systems. Super! Works in the free Viewer, too.
5 Minute Tutorial - Georegister a Drone Photo - See the fast and easy way to georeference drone photos for use in GIS and online web mapping: Learn how to georegister (georeference) a drone photo to line up with Google imagery for full GIS use and for use within Google Maps and other web mapping applications. This video uses exactly the same drone photo used in ESRI's ArcUser example of how to georeference a drone photo in ArcGIS Pro. The difference is that using Manifold is faster and easier.
Example: Georegister a Drone Photo - We take a raster image, a drone photograph in Everson, Washington, that was imported from an ordinary .jpg file, and we georegister it using a map that shows a Google Satellite view of the same region, casting the drone photo into Pseudo-Mercator projection. We use previews to see how well the control points we have added will work, before creating a georegistered image.
Example: Georegister a Vector Drawing - We take a vector drawing with an unknown coordinate system that shows the provinces of Mexico and we georeference it to a map containing a Bing Streets web-served layer, casting the Mexico drawing into Pseudo-Mercator coordinate system. We begin the process using only two coordinate points and then we do a preview to see where accuracy of the proposed georeferencing result should be improved by adding more control points. We add more control points and then georeference the Mexico drawing with good accuracy.
Example: Georegister a Whole World Image - We make a screenshot of a map we see on the web that covers the whole world, from +/- 90 degrees of latitude and +/- 180 degrees of longitude. The image shows the position and geology of continents as they were 200 million years ago. We georeference the image using four control points placed at the +/- 90 degrees and +/- 180 degrees corners, using a target map with a Bing streets background layer. We use the Show Coordinates option in the Register pane to quickly set exact target control point locations. As a bonus, we show how to knock out "background" pixels if our image is a palette image.