PCI’s Historical Airphoto Processing (HAP) System – Tech TV Demo

In today’s Tech TV episode we’ll be
providing you with an overview of PCI’s Historical Airphoto Processing
workflow, better known as HAP with HAP, you can double your current
production throughput significantly reduced manual labor and achieve high positional accuracy. The power of HAP is in its automated and
iterative approach to collecting hundreds of Ground Control and Tie Points to build math models that are used to
create highly accurate ortho mosaics. the HAP workflow is a new and
innovative system that PCI has developed, making it possible to correct
a large volumes of historical imagery. The system is comprised of mostly
automated steps with minimal manual setup and quality assurance. Hhere’s an overview: Step one, Data Prep consists of preparing
data to be used in the system. Step two, data ingest, will ingest the imagery and
produce initial nominal geo-referencing as a starting point for the alignment
steps. Sstep three, coarse alignment refines the model using
the reference imagery to collect initial ground control and tie points. In Step four, fine alignment, we will
complete the model using the coarse alignment as the starting position with
improved ground control points and tie points. Once the model has been refined PCI’s
ortho mosaic technology can be used to automatically creates seamless
mosaics. The data preparations step gets the data
ready for the HAP system. Here we have a folder on our machine with all all the data. The input directory contains all of the
raw scanned TIF airphotos for this project. In this case we have over two hundred
and fifty different images that we will be processed. We also have a reference folder that
contains all of our control data. So in this folder we have our reference
imagery and DEM that will be used to collect Ground Control Points automatically to produce Orthos. Finally we have some important metadata that is associated with our airphotos. The metadata includes a minimum amount
of information required such as approximate center coordinates for the air photos, focal length and flying altitude. So as you can see, at this point our air
photos have no geo-referencing. In their current form, only visual analysis is possible. We need to associate accurate geo
referencing to use the air photos for proper geospatial analysis. In Step two, Data Ingest, we begin by
launching an automated process that reads and all of the input imagery, reference data, and metadata into a suitable format for the HAP system. So we’ve got our project open, now let’s
collect some fiducial marks on the images. The HAP system has some great automation
tools, we will collect fiducials on one air
photo, and use it as a template for the
remaining images. This will save us lots of time. Now that we have our template fiducials collected, let’s use the automation in HAP to
collect the remaining fiducials on our airphotos. As you can see, the automatic fiducial collection has completed, and all of the hundreds of images in our
project have automatic fiducials that have been collected. Let’s take a look at the initial nominal
georeferencing produced through this step. now let’s zoom into an intersection. As you can see, there’s a significant offset. This is only our initial model which
will be further refining iteratively, through the use of automated GCP and Tie Point collection. In Step Three, Coarse Alignment, we will launch an
automated process that will automatically attempt to collect hundreds
of GCPs and tie points per image to further a refine our model. Now that’s done let’s look at the
results of the coarse alignment. The HAP system has collected tens of
thousands of GCPs and Tie Points with an acceptable RMS for this stage of
the process. Now that we have a coarse model, let’s go
back and see how we’ve improved upon the initial nominal georeferencing. This is a reference image, here is the nominal georeferencing
image, I’m now going to toggle on our results from the coarse alignment step. As you can see we are much closer, approximately within twenty pixels. In Step Four – Fine Alignment, we will
once again launch an automated process that will attempt to collect hundreds of
GCPs and tie points per image to further refine our model. This iterative approach is a proven method to achieve accurate results. That, combined with the automated GCP and tie point collection is what makes the HAP system innovative. Once again perhaps system has collected
tens of thousands of GCPs and Tie Points, with an improved RMS. Now we have a suitable model to produce
accurate orthos. Let’s take a look at the results from
the fine alignment step. Once again, this is our reference image. Here’s the nominal georeferencing from
Step Two, Data Ingest. As you can see the accuracy is off by
roughly three hundred pixels. And now, the final Ortho produced with the model from the fine alignment is very accurate. The HAP system has done a great job orthorectifying these airphotos. Our final ortho mosaic which was
produced using automated cut lines and colour balancing is of high quality. This two hundred and fifty seen project
was completed in under a day. For more details on metrics, go to our website to download a white
paper on the HAP system. Thanks for watching!

Leave a Reply

Your email address will not be published. Required fields are marked *