Orthorectifying and Mosaicking in PCI Geomatica
Creating a Project
Select File–> New Project and create a new project. The following screen will appear. Select your folder and file name. Ensure that the settings below are chosen.
Next set the projection and pixel spacing for both the output projection and GCP projection.
Next set the Focal Length and scale of the project. This should be available in the flight calibration report. Select Fiducial locations (middle, edges or both) and calculate locations. If you don’t have this information then select compute from length and input the length of the image edges. If you do not have this data estimate the length as best as possible.
Creating a DEM
I recommend that before creating your Ground Control Points, you recreate your DEM using the project projection in the *.pix format. To create a new DEM open up the Import & Build DEM step and select the create DEM from Raster.
Select your DEM (or DEMs) and push the down arrow button to add it to the Set of DEMs to merge box. This will make the OK button selectable.
Ok will open up the following dialogue. Defaults should be acceptable. Use Select to choose your file location and type your preferred file name into the Output DEM text box.
The first step is to open up the imagery that you need to orthorectify and mosaick. Select Data Input and choose the button to open new images as shown below. A dialogue will appear. Select your images and add them to the project.
To locate the fiducial marks on your image for orientation and model building purposes, select an image and click open. Two dialogue boxes will open including the Viewer image below and the fiducial marks dialogue to the right.
The second screen is the Fiducial Mark Collection dialogue. If you select the centre of the fiducial marks located in 4 locations on the image viewer, and then click set within the fiducial marks dialogue box, the appropriate mark location X and Y position will be filled in. Do this for all 4 fiducial marks. Ensure that you are using the correct location for fiducial marks (i.e corners or edges) from the project settings.
Make sure to save your project at regular intervals throughout this process!!!!
Selecting Ground Control Points
The next is providing GCP’s. There are several ways to do this.
1) Have ground markings that have been located on the ground and are obvious in the imagery (i.e white x’s)
2) Using a vector file such as roads to mark intersections. This is not always accurate as roads can change over time or be inaccurately located depending on the scale of your vector file.
3) Use an already geocoded image to do a comparison GCP selection. This entails selecting a feature that is duplicated in both sets of imagery (preferably close to the ground) and assigning the point a number on both imageries. This assigns the unlocated point a real location. If you load a DEM this location can use an extracted elevation. 4 points must be chosen to generate a model and provide accuracy points. The point type can be changed to GCP’s or check points. Check points are used to test accuracy. This project did not utilize check points but only considered the residual report and visual comparison.
Select a DEM and a geocoded image. Open up the image that you would like to orthorectify from the project images dialogue. Once the image is open find a location and select the preferred point. Select the matching point on each image, and on each separate viewer select use point. Not that the imagery needs to be the working image and not a reference image. Once you select use point on both viewers, you can extract the elevations and accept the point. If you checkmark “compute model” after 4 GCP’s have been selected the model will update.
After selecting 4 GCP’s you can try the automatic GCP tool to add additional points to potentially increase accuracy. I did not use this tool as my accuracy was good within 50cm.
Selecting Tie Points
Now that the imagery has been given a real world location, the images are tied to each other for a more accurate mosaicking result. This allows higher accuracy at the edges of the imagery and better correlation between images.
Points can be selected in a manner similar to GCP’s only using two aerial photography images instead of one and a geocoded image (or other method).
In this situation I was able to select the automatic tie point and establish 7 tie points between the 3 images used.
Use the project over view beside the tie point buttons to look at the foot print of the imagery over the mosaicked raster.
Compute the Model
Click the compute model button as shown below.
If successful it will simply tell you that the Bundle Adjustment was successful.
OrthoRectify your images.
Select the images to rectify and click the arrow button to add it to the images to process box.
Edit the settings shown below to preferred settings and select ok.
Mosaick the OrthoImages
The Define Mosaic tool will allow you select the images that you wish to mosaic in an overview screen.
After you select the images you want to mosaick you can run the automatic mosaick tool (tool on far right of above dialogue).
This opens up the window to the right. Set up a location for your final mosaic and your demo mosaic. Explore the highlighted settings and select generate demo to find the best match before creating the mosaic.
Print and Examine Residual Report
In the Ground Control Point option select the residual report icon.
As you can see to the right, my accuracy was .5m in ground units for GCP’s and approximately 2m for tie points. This makes sense as tie points occur the furthest from Nadir and are the last accurate locations. Higher accuracy might be obtained from selecting manual points.
Orthorectifying and Mosaicking in ERDAS Imagine
Open the LPS Extention for Erdas and create a new project. Step one is to set the projection.
Set the average flying height and select Edit Camera. Enter appropriate information (Focal length and Fiducial number).
After the project is created, right click on the images folder and select add.
Click the Int. button on one of your images to open up the following dialogue
Click the viewer button to see the image.
The image dialogue box is shown below. Check the sensor and make sure that the fiducials are appropriate (i.e middle or corners, middle should be approx (0,113 for X,Y top middle). Input as much data as you have available such as principle points, distortions etc. Navigate to the appropriate location and choose the crosshairs to choose the location of the active fiducial. Erdas will approximate the next location and move your location there. This becomes fairly accurate. Do this for all images. Make sure residuals are reasonable for your project.
Open the point measurement tool (edit).
This opens up a window wit ha side by side viewer containing 3 views of each scene based on zoom level. In order to rectify to a geocoded image select the identified horizontal (geocoded) and vertical (dem) reference buttons and click the “use viewer as reference”. This will open up the geocoded image beside one of your aerial images.
Reference Points and Tie Points
To create a reference point select the Add option on the top right of the window. This adds a point to the reference layer. Find a location in both images that match and select the cross hair to select the point in the image. You will need to select the cross hair once for each image. This will create a table below the images as shown to the right. Change the Type to Full and the Usage to Control. Repeat this process at least 4 times for each image and then select the triangle to compute the bundle adjustment.
Once complete, under right viewer select a different image and continue adding control points.
To make a tie point use the tie point setting and create a tie point between two non ortho-rectified image. I found that it works best to just reference two images to the same point. Once this is done compute the bundle adjustment and look at the report selecting process–>triangulate report.
To add more tie points automatically, in the main window go to process – automatic tie point generation and use the default settings. This will create a number of tie points. This reduced my RMSE from 22 to 11 pixels. It is assumed that additional control points could further reduce this error.
The tie point creator in ERDAS is extremely robust and allows for creation of tie points between images of one flightline as well as then tie points between multiple flightlines. You can increase your accuracy by analyzing the summary reports of these actions and creating your model more accurately. These settings can be changed within the auto tie point generation properties menu. This is beyond the scope of this comparison but there are many options within triangulation properties to aid in more accurate results.
Orthorectifying is very simple. Either select the ortho option in the viewer beside the image or go to process –> ortho rectification –> ortho resampling. Choose the appropriate save file, DEM and resampling method (cubic suggested). Select ok and it will produce an orthorectified image using your model.
To mosaick the imagery simply select process–> mosaick and select the appropriate DEM. Follow the instruction in this ERDAS mosaicking tutorial. I would recommend adjusting the seamline selection method
Selecting the Nadir option helps remove edge of image artifacts.
Overall I found my Erdas results less satisfactory and I found the process more confusing than in PCI. This is definitely the choice if you need extremely detailed control and don’t mind putting in extra time in order to achieve accuracy closer to that from PCI. As I have expressed in past posts, I much prefer the process of mosaicking in PCI and typically get better results. This example is no exception. Once again, I am sure through options it is possible to achieve results as good as PCI, but PCI seems to achieve them more readily with less user input.
Results and Comparison
Overall I preferred the results achieved from PCI Geomatica. There is no question however that Erdas has a better viewer interface for point selection and smarter fiducial choosing as it automatically moves you to the next fiducial zone. It is possible that PCI has a similar function that needs to be turned on but if so I am unaware of the tools location.
Additionally Erdas has many options tie point creation and generation and the ability to have excellent control over your accuracy and flightlines. Despite this, I found that I achieved better results in a more timely manner within PCI Geomatica with fewer edge affects on my final mosaick.