Sei sulla pagina 1di 23

March

13/13

GISC9216- Digital Image Processing


Assignment 3- Geometric Correction, Orthorectification and Mosaicking

Brooke Le Blanc

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

March 13, 2013 GISC9216-D3 Janet Finlay Niagara-on-the-lake Campus, Niagara College Room E301 135 Taylor Road, S.S #4 On, L0S 1J0

Dear: Janet Finlay: Re: Submission of GISC9216-D3 Please accept this letter as formal submission of Deliverable 3- Geometric Correction, Orthorectification and Mosaicking for the GISC9216 Digital Image Processing course. For the third submission of the Digital Image Processing course, students learned the basics of orthorectification and mosaic processes. This assignment was completed using three different photos that have been fixed together in a mosaic form after being corrected by both the camera and polynomials method. Students were then asked to compare the results of both of the corrected and then fixed mosaic images. To conclude, I have attached a copy of my GISC9216D3- Geometric Correction, Orthorectification and Mosaicking. Please do not hesitate to contact my for any additional information at 1-416-509-8070. Thank you for your time and attention. We look forward to hearing from you.

Respectfully, Brooke Le Blanc Brooke Le Blanc BA, GIS-GM Graduate Candidate B.L/

Encl: 1.) BrookeLeBlancGISC9216D3 2.) mosaicpolynomial.img 3.) mosaiccamera.img

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Executive Summary This result of this assignment was to successfully mosaic together three given .tiff images to create an .img file. The act of the mosaic was to be done twice using the same three images. The first mosaic was using a third order polynomial method, while the second was while using an orthorectified method that required the use of a camera. Both types of mosaics required the use of georeferenced ground control points. For this assignment the order of transformation decided upon was 3, which required a minimum of 10 ground control points per image. This assignment compares the results of both of these mosaic .imgs. The way this was done was by comparing the features on the pictures in the same mosaic to see how accurately they were combined together. A conclusion was then reached on which of the two methods proved to be most effective.

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Table of Contents
Introduction ..................................................................................................................................... 4 Purpose............................................................................................................................................ 4 Use of GCPs (Ground Control Points) ............................................................................................. 4 Polynomial Model (Reference ArcGIS Website) .............................................................................. 4 Orthorectification Model ................................................................................................................ 6 Errors with GCP ............................................................................................................................... 7 Resampling .................................................................................................................................... 10 Image Clipping ............................................................................................................................... 12 The Mosaic Process ....................................................................................................................... 13 Conclusion ..................................................................................................................................... 14 Appendix........................................................................................................................................ 15

Table of Figures
Figure 1: Polynomial Equation......................................................................................................... 5 Figure 2: Highlighted fiducial ........................................................................................................... 6 Figure 3: Sum Button from ERDAS .................................................................................................. 7 Figure 4: Polynomial 1 RMS Error.................................................................................................... 7 Figure 5: Polynomial 2 RMS Error.................................................................................................... 8 Figure 6: Polynomial 3 RMS Error.................................................................................................... 8 Figure 7: Orthorectification 1 RMS Error......................................................................................... 9 Figure 8: Orthorectification 2 RMS Error......................................................................................... 9 Figure 9: Orthorectification 3 RMS Error....................................................................................... 10 Figure 10: Resample Button in ERDAS........................................................................................... 11 Figure 11: Resampling Popup ........................................................................................................ 11 Figure 12: Description of Nearest Neighbor Resampling .............................................................. 12 Figure 13: ERDAS Polygon Tool ..................................................................................................... 12 Figure 14: Creation of Clipped Subsets ......................................................................................... 13 Figure 15: Mosaic Express Selection ............................................................................................. 14

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Introduction
The process of creating a mosaic joins images with the same overlapping feature through the use of different methods. This mosaic process can be done with the computer program ERDAS imaging. The methods used for this assignment were a third order polynomial and an orthorectification method. The methods both require the use of georeferenced ground control points. The methods differ because the orthorectification method also involves the use of a camera that uses fiducial points around an image to consider elevation in the specific area.

Purpose
The data given to complete this assignment is three .tiff files, a subset .img, a digital elevation model .img and shapefiles of buildings and roads to help determine which mosaic was most successful. All of the data for this assignment was provided in the Niagara College X: Drive. This assignment will give students insight into the fundamentals of orthorectification and mosaic process. It will also allow students to create their own georeferenced ground control points and show how accurate these points really are through the RMS errors given for each georeferenced ground control point.

Use of GCPs (Ground Control Points)


As previously mentioned, in both mosaic processes georeferenced ground control points will be necessary. GCPs are what is used to find a relationship between the uncorrected .tiff and the corrected subset image that already has a coordinate system. It is important that ground control points are spread evenly throughout an uncorrected image so they have an even spread of exact georeferenced locations.

Polynomial Model (Reference ArcGIS Website)


A polynomial model uses ground control points to create a polynomial transformation. This transformation will alter the existing dataset from its current location to the spatially correct location of the subsets geographic coordinates. For this assignment the decision was made to complete a 3rd order polynomial. This means that for each .tiff image being georeferenced, at least 10 ground control points are necessary. Where ever possible the GCP points should be spread out. Best results

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

usually occur when there is at least one GCP near each corner of an image and then the rest spread evenly throughout the image. After all of the required GCPs are collected the polynomial transformation can be completed. The transformation itself consists of a polynomial that is created from ground control points and a least square fitting algorithm. The polynomial itself is composed of two formulas, one for the x- coordinate output and one for the y-coordinate output. As shown on the Arc GIS help website, the polynomial can be seen below.

Figure 1: Polynomial Equation

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Orthorectification Model
Similar to the polynomial model, the orthorectification model also uses GCP points to correct an image. The difference between the two is the use of fiducials to project elevation. Fiducials are the eight points surrounding an aerial photo. One of those fiducials is highlighted below in the yellow circle.

Figure 2: Highlighted fiducial

The fiducials are the first 8 GCP points and are created in ERDAS through a camera model. Then after the camera model is set up the addition of more GCPs are required to be georeferenced throughout the image to allow for a more accurate result. Because the
6

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

camera model is being used with the DEM image, parts of the image that may occur on top of buildings with elevation changes will be georeferenced correctly.

Errors with GCP


The addition of GCPs to an image is done in a manual process. In saying that, it is without question subject to human error. After the completion of adding all of the GCPs the sum button, which is shown below can be clicked and it will populate the (RMS) field in the attribute table of the GCP points.

Figure 3: Sum Button from ERDAS

RMS stands for Root Square Mean and can be defined simply as the difference between the GCP location on the uncorrected image and the corrected image. The larger the RMS error, the less accurate the resampled image will be, and in relation to this assignment, the less accurate the final mosaic will be.

Figure 4: Polynomial 1 RMS Error

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Figure 5: Polynomial 2 RMS Error

Figure 6: Polynomial 3 RMS Error

Image Image 1 Image 2 Image 3

Polynomial RMS Error Average 0.026 0.024 0.047

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Average of RMS Error in Polynomial Mosaic = 0.032

Figure 7: Orthorectification 1 RMS Error

Figure 8: Orthorectification 2 RMS Error

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Figure 9: Orthorectification 3 RMS Error

Image Image 1 Image 2 Image 3

Orthorectification RMS Error Average 0.019 0.019 0.014 Average of RMS Error in Orthorectification Mosaic = 0.017

After calculating the amount of RMS error in each of the mosaics it is apparent that the Orthorectification mosaic has the lesser amount (0.032 > 0.017 ). A reason why this is likely to occur is because the Orthorectification models are georeferenced by elevation through fiducial points, as well as the ground control points that the polynomial model uses.

Resampling
Before the mosaic process can take place, each image must be resampled before added to the mosaic. Resampling is actual process of increasing or reducing the number of pixels that an image consists of. When the pixels of an image are altered the actual size and resolution of an image can be significantly altered. Below is the Resample button that is to be selected in ERDAS that will open the Resample popup.
10

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Figure 10: Resample Button in ERDAS

Figure 11: Resampling Popup

In the Resample popup it is visible that the output is saved as a .img file. It is also visible that Nearest Neighbor has been chosen as the type of resampling. This method was selected for all of the resampling that took place for the assignment. Nearest Neighbor resampling takes the value from the pixel in the original uncorrected image and uses the nearest to the new pixel location in the corrected image. A visual of the Nearest Neighbor sampling method is shown below.

11

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Figure 12: Description of Nearest Neighbor Resampling

Image Clipping
Another step in the process before the mosaics can be created is the clipping of the resampled .img files. The images were clipped so that the final mosaic result would look a lot cleaner without writing and fiducials on the photo. To clip an image the tool highlighted below must be selected.

Figure 13: ERDAS Polygon Tool

12

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

The user then proceeds to draw a polygon around the image and double click to create the selection. When the selection is highlighted the create subset image function can be selected with the following fields selected below.

Figure 14: Creation of Clipped Subsets

The Mosaic Process


The process of piecing images together is called mosaicking. It is beneficial to mosaic images together to get a larger picture of the same region. The process is extremely common with aerial photos. For this assignment two mosaics were created, one for the polynomial model and one for the orthorectification model. The results of the mosaic will be reliant on how accurate the ground control points were created. If there were a lot of RMS errors the separate pictures will not be aligned when mosaicked. There are different methods to mosaic images, the one selected for this assignment was mosaic express. Below is the mosaic express popup after three of the images to be mosaicked had been added.

13

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Figure 15: Mosaic Express Selection

Conclusion
The completion of this assignment has helped to prove that the Orthorectification mosaic has a smaller amount of RMS error. This mosaic also uses fiducials associated with DEM data to decrease error within the mosaic by taking elevation of the landscape into consideration. After working with the two separate mosaic images it has become extremely evident that the success of these procedures relies heavily on the users ability to place the ground control points accurately on both the incorrect and corrected image.
14

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Appendix
Third Order Polynomial

Polynomial Image 1

15

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Polynomial Image 2

Polynomial Image 3

16

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Orthorectified

Orthorectification Image 1

17

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Orthorectification Image 2

18

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

Orthorectification Image 3

19

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

20

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

21

Brooke Le Blanc [15 Strathmore RoadSt. Catharines Ontario L2T 2C4] [416-509-8070b-leblanc@live.com]

22