orthorectified, 2-foot resolution multispectral imagery with less than 2 meter overall error. for background imagery in cartographic products and for heads-up digitizing of features No problem! Someone HAS to have written a detailed how-to document for this!
Right?… Found 30+ journal articles, but few clues and no step-by-step methodologies
1. Mosaic (unchip) by scene 2. Source your DEM 3. Orthorectify the panchromatic and multispectral bands 4. Pan-sharpen your multispectral imagery 5. Georeference using ground control points
In order to squeeze the imagery on to DVDs, Digital Globe splits up the scenes into tiles. Not an issue until you realize that the RPC files needed to orthorectify the imagery are delivered by SCENE not by tile. So you have to re- assemble the scenes before orthorectifying.
Cannot stress this enough - when using Imagine to orthorectify you NEED a DEM that does the following: covers your full area of interest and has horizontal and vertical reference EXACTLY the same as your imagery
Done using the LPS Extension of ERDAS Using sensor and image information provided by scene, and using the DEM, you end up with 2 orthorectified images per scene (one panchromatic, one Multispectral)
At this point, you should have a pan-sharpened, ortho-corrected image. If desired, you could have used any ground control points (GCPs) during the orthorectification process in Imagine. In our case, since the imagery follows road corridors and the spread of the available GCPs is linear, processing the imagery with GCPs at that point resulted in unsatisfactory products. Instead, we are using our network of ground control points to run georeferencing in ArcMap after the pan-sharpened product is finished. This adds an extra step to the process, but results in a higher quality image.
error of less than 1.5 meters using control points for verification Recent points taken by geology field crew at Old Faithful are within 0.5 meters (using Trimble GeoXH) Contacted by NAIP processing crew, and they are using our GCP database to rectify the 2009 NAIP imagery.
Collecting ground control points ad hoc for smaller projects has a long tradition at Yellowstone NP. We had many projects from GIS and Geology with GCPs scattered throughout project folders. With ~1600 km of high-resolution QuickBird imagery we realized that we had a choice: register the QuickBird to the existing NAIP imagery (with a ~3m accuracy in areas of low relief, ~7 m in high) or make the QuickBird imagery our best data set to which we could then register other imagery. We chose the latter.
Used short-term volunteers to collect points over summer 2008 & 2009 1,200+ points Not all are useful…
Check for free imagery on the NGA WARP site and use the CIDR tool! (these are changing all the time) Order full scenes rather than by convoluted Area of Interest (such as buffers along roads) Purchase 2 external hard drives – one to backup your data daily and one to work from. We used one 1 TB hard drive for processing and backed up to our network drive. When dealing with large amounts of imagery (multiple scenes), keep a log file of your imagery – spreadsheet works fine. Most processing was done on a computer with a quad-core processor and 8 GB of RAM. We did spread the work around though, and some was done on dual-core processor with 2 GB of RAM. It just ran MUCH more slowly – some processes will run for 8+ hours. Train your ground control point collection crew to a specific method before setting them loose with GPS units!
Available upon request: How-to documents for the image processing and Ground Control Point collection 31.pdf reference papers used to create this methodology email@example.com