Historical Airphoto Processing
Using modern, automated image correction methods to tap into valuable historical imagery.
Historical aerial photography archives contain valuable information that remains untapped. Digitally scanned and properly geo-referenced historical aerial imagery can bring this information to life, making it possible to analyze/visualize the historical information in modern GIS systems. These historical images can reveal hidden patterns, provide a deeper understanding of changes over time thus leading to better decision making.
PCI Geomatics offers a customized solution to automate the correction of historical imagery. With it, users can properly prepare historical data and set up workflows that will create perfectly aligned and orthorectified mosaics for use in numerous GIS applications.
Start realizing the value of your vast archives of historical airphoto imagery today and turn hundreds / thousands of archive images into GIS ready digital mosaics.
18. HAP Value Proposition
Increased Production
• The HAP system can more than double
production capacity
Increased Efficiency
• The multi-pass automatic GCP and Tie Point
collection reduces manual effort
High Accuracy
• Accurate ortho-mosaics that often exceed the
quality produced by manual techniques
•
In many cases can handle poor interior model information better
than manual approach
19. HAP Features & Benefits
Features
Benefits
Automated Fiducial Collection
Reduced labor, high accuracy
Automated GCP/TP
Significantly Reduced labor,
high accuracy
Ortho-Mosaic – automated
cutline and color balancing
High quality mosaics in less time
Works with multi-source and
varying quality historical imagery
Rescue your valuable data
21. Case Study
Project Details
-
130 raw aerial images acquired over a large urban area in 1954
-
Reference image of urban area acquired in 2011
-
Population grew by more than 10 million people between 1954 and 2011
Processing Details
Traditional Approach
GCP Collection
Tie Point (TP) Collection
Quality Assurance
Manual Labour (Hours)
Project Turn Around (Days)
HAP System
Manual
Automatic
Automatic
Automatic
Manual (TPs only)
Manual
24
8
3 Days
1.5 Days
22. Case Study
The same customer performed 7 additional tests
over two areas in Asia Minor.
23. Case Study
Project
# of Photos
Processing
time (Hours)
Accuracy
(CEP 90)
001-1946
122
16.5
±8m
001-1954
117
14.7
±6m
001-1972
90
10.5
±6m
001-1993
87
9.2
±5m
002-1975
69
10.8
±7m
002-1991
73
10.0
± 6.5 m
002-1999
80
9.0
±5m
24. Case Study
“All ortho-photo projects performed by the
operator through conventional methods were
found to be three times slower than the same
processes performed on the HAP System.”
- HAP customer
Notes:The HAP System has successfully processed projects containing imagery with an initial positional error of greater than 2000 pixels.Orientation accuracy is another important concept, which is related to the initial positional accuracy of images in a flight line.
Notes:Land cover change can be a big issue with historical imagery. The ability to automatically collect GCPs from a reference image can be hindered by a large degree of land cover change between the reference image and the raw data. However, there are many different techniques that can be applied to overcome this problem, such as copying GCPs successfully collected on overlapping areas of neighboring images or by using the HAP System to create a new reference image of a date closer to that of the raw imagery. For example, if you are orthorectifying many different datasets of a given area over different years, correct imagery closer to the year of the reference image and then use that output mosaic as a reference to correct datasets from prior years.
Notes:Land cover change can be a big issue with historical imagery. The ability to automatically collect GCPs from a reference image can be hindered by a large degree of land cover change between the reference image and the raw data. However, there are many different techniques that can be applied to overcome this problem, such as copying GCPs successfully collected on overlapping areas of neighboring images or by using the HAP System to create a new reference image of a date closer to that of the raw imagery. For example, if you are orthorectifying many different datasets of a given area over different years, correct imagery closer to the year of the reference image and then use that output mosaic as a reference to correct datasets from prior years.
Notes:Scratches and damages on the imagery can hinder automation when collecting GCPs, Tie Points and fiducial marks. Scratches often appear as white or relatively bright grey marks that can confuse the correlators, making it more difficult to obtain accurate GCPs automatically.
Shawn
Shawn
Notes:The time it takes to complete a project is highly variable. Often it depends on the quality of the input imagery, which impacts the number of alignment iterations required and or manual effort. In general, most projects complete in about half the time it takes an operator to complete the same project using the traditional manual approach when attempting to achieve the same accuracy.
Shawn
Shawn
Shawn
Shawn
Notes:The optimal configuration of the HAP system is generally 2 QA/QC Machines for 1 Processing system. However, the configuration is determined on a customer-by-customer basis.The Processing System performs all of the automated process, which can be launched from a remote location (i.e. remote desktop). The QA/QC machines are installed with a special QA build of Geomatica and the QA/QC operations are performed on the machines, while accessing the data remotely over the network (data always stays on the processing System).
Notes:Data Preparation is an automated step, but can often be optimized with scripts that setup the required inputs. PCI can customize scripts for customers, based on how their data is stored
Notes:Data Preparation is quite simple, but depending on what information is available for the imagery, it can be time consuming. Data Preparation requires:Copying all of the raw input imagery into the same folder Input imagery must have the same fiducial types (i.e. corner or edge or corner and edge) Input imagery must have a focal length that is almost the sameInput imagery should be continuous (no major gaps or missing flight linesEnsuring you have a suitable reference image and DEM copied to the Reference folderCreating a Metadata file that containsApproximate center Longitude and Latitude coordinates Focal LengthInput Image nameApproximate flying Altitude (Feet above MSL)Physical Image DimensionsPCI can create scripts to help optimize the creation of the metadata file
Notes:Data Ingest is the process of reading in the input information and outputting the imagery in working formats (PIX). Linked PIX files are generated to minimize the I/O time when ingesting data. Data ingest also consists of fiducial mark collection, which can often be automated.
Notes:The Data Ingest stage also creates nominal (initial) georeferencing for the images, based on the input information.
Notes:The purpose of the coarse alignment step is to improve the initial positional accuracy of the imagery so that it is suitable for the fine alignment step.Some more difficult projects may benefit from running the coarse alignment step multiple times, which iteratively improves the quality until it is suitable for the fine alignment.
Notes:The coarse alignment step attempts to collect hundreds of GCPs and Tie Points for each image. This number of GCPs and Tie Points can effectively compensate for poor internal orientation due to limited information (i.e. missing radial distortion coefficients).The output of the coarse alignment step is an improved georeferenced image, which provides a better starting point for the fine alignment step.
Notes:Quality Assurance is an important step. The level of manual edits required depends on the success of the automated coarse alignment run, which is often related to the quality of the input imagery.QA/QC of the coarse alignment step often requires removing bad GCPs and tie points and collecting GCPs on images that do not have any or that are not tied to a subset of imagery with GCPs.
Notes:The fine alignment step uses the math model produced by the coarse alignment step as a starting point so that it has a better chance of successfully collecting GCPs and tie points on the images with the accuracy required to produce ortho images.It should never be required to run multiple fine alignment steps (all necessary improvements should be applied during the coarse alignment steps)
Notes:QA of the fine alignment step should be limited to deleting bad GCPs and tie points to obtain the final desired accuracy. However, it is sometimes necessary to manually collect GCPs in images with poor distribution. For example, if an image has most of its GCPs in the top half of the image, it may be necessary to remove some GCPs from the top (called pruning) and manually add some GCPs to the bottom half.
Notes:Orthorectification is a high speed automated process that can be performed using Geomatica or GXL
Notes:The Mosaic Preparation step automatically generates cutlines and color balances the imagery. The output is a preview mosaic with cutlines and topology polygons for the images.
Notes:Users can use PCI’s Mosaic Tool to manually edit cutlines and color balancing in order to ensure that the output is a seamless mosaic. PCI’s Mosaic Tool allows multiple people to connect to a single mosaic at the same time from different computers. A user can ‘check out’ a region of the mosaic to work on, which locks other users out of that area until it is checked back in.
Notes:The final mosaic is a full resolution mosaic image produced using the cutlines and color balancing levels accepted in the QA/QC of the preview mosaic.