Sie sind auf Seite 1von 23

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [Nathan Page]

NIAGARA COLLEGE

Geometric Correction, Orthorectification and Mosaicking


GISC9216-Deliverable 3
Nathan Page 3/12/2014

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

March 12, 2014 GISC9216-D3 Mrs. Janet Finlay GIS-GM Program Coordinator Niagara College 135 Taylor Road Niagara-on-the-Lake, ON L0S 1J0 Dear Mrs. Finlay, RE: GISC9216-Deliverable 3-Geometric Correction, Orthorectification and Mosaicking Please accept this letter as my formal submission of Deliverable 3: Geometric Correction, Orthorectification and Mosaicking for GISC9216-Digital Image Processing. My report can be found on the X drive under X:\Students\NPAGE4\GISC9216\Assignment3\PageNGISC9216D3. This assignment served to greatly further and encourage skills in ERDAS, specifically processing digital images; having zero background in remote sensing and very little in ERDAS, I found all the skills learned here very interesting. Utilizing the two processes to correct photos brought about the realization of the advantages and disadvantages when using Geometric Correction versus Orthorectification; Geometric Correction is less accurate and produces and inferior mosaic but requires slightly less work. Overall I feel Ive learned a great deal from this assignment and truly did enjoy the work that it consisted of. Should you have any questions regarding the enclosed documents, or if there are technical issues regarding the files please contact me at your convenience at nathanpage90@gmail.com. Thank you for your time and attention. I look forward to your comments and suggestions. Warm Regards, Nathan Page, BA, GIS-GM GeoTech Solutions Inc. NP/

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Contents
1. 2. 3. Introduction ...................................................................................................................................... 2 Methodology..................................................................................................................................... 2 Discussion.......................................................................................................................................... 3 3.1 Question A: Prediction Process ....................................................................................................... 3 3.2 Question B: Resample Photo Pixel Size ........................................................................................... 7 3.3 Question C: Geometric Polynomial Correction-Mosaic Output Quality ......................................... 9 3.4 Question D: Prediction Process..................................................................................................... 13 3.5 Question E: Resample Photo Pixel Size ......................................................................................... 16 3.6 Question F: Orthorectification-Mosaic Output Quality ................................................................ 18 4. 5. Conclusion ....................................................................................................................................... 20 Bibliography .................................................................................................................................... 21

Figure 1: Methodology Diagram ................................................................................................................... 2 Figure 2: GCP #4 ............................................................................................................................................ 3 Figure 3: GCP #5 ............................................................................................................................................ 4 Figure 4: GCP #6 ............................................................................................................................................ 5 Figure 5: GCP #7 ............................................................................................................................................ 6 Figure 6: Photo 1 Resampled Cell Size .......................................................................................................... 7 Figure 7: Photo 2 Resampled Cell Size .......................................................................................................... 7 Figure 8: Photo 3 Resampled Cell Size .......................................................................................................... 8 Figure 9: Mosaic of Polynomially-Corrected Photos..................................................................................... 9 Figure 10: Example of Mosaic Overlap ....................................................................................................... 10 Figure 11: Example of Mosaic Overlap ....................................................................................................... 11 Figure 12: Mosaic Vector Overlay Example ................................................................................................ 12 Figure 13: Mosaic Vector Overlay Example ................................................................................................ 12 Figure 14: GCP #5 ........................................................................................................................................ 13 Figure 15: GCP #6 ........................................................................................................................................ 14 Figure 16: GCP #8 ........................................................................................................................................ 15 Figure 17: Photo 1 Resample Cell Size (Ortho) ........................................................................................... 16 Figure 18: Photo 2 Resample Cell Size (Ortho) ........................................................................................... 16 Figure 19: Photo 3 Resampled Cell Size (Ortho) ......................................................................................... 17 Figure 20: Orthorectified Mosaic ................................................................................................................ 18 Figure 21: Example of Mosaic Overlap ....................................................................................................... 19 Figure 22: Example of Mosaic Overlap ....................................................................................................... 20

1|Page

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

1. Introduction
This report will detail and explain the results of the orthorectification, geometric correction and mosaicking of three aerial photographs. Orthorectification utilizes one image, known as the Reference Image to correct other images of the same general location by locating areas of commonality between the two known as Ground Control Points. Using these ground control points together with the Reference Image, the photos are resampled and re-projected, adhering to the parameters of the reference image and any fiducial marks which are added on in the orthorectification process.

2. Methodology
Carrying out the geometric correction, orthorectification and finally the mosaicking of both techniques required multiple tools and steps to be utilized from the Erdas Imagine 13 software program. A diagram can be seen below detailing the flow of steps and tools utilized in both techniques.

Figure 1: Methodology Diagram

2|Page

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

3. Discussion
3.1 Question A: Prediction Process Following the input of the first 3 Ground Control Points (GCP) for the polynomial geometric correction, theres an automatic setting which helps speed up and increase accuracy of the process. Erdas has the ability to predict any further GCPs that you wish to enter by using the initial points to triangulate and localize where the ground control point may be located on your Reference Image according to where you place it on your Input Image. As is visible in the Figure 2, 3 and 4 below, during the creation of this deliverable the prediction process was giving a relatively good localization on the latter GCPs. Figure 1 displays the first prediction (GCP #4), where it can be seen how accurate the prediction was; the predicted GCP was only meters from the actual location, most likely under roughly 10 meters. When examining the screen shots for GCPs 5 through 7 (Figures 2, 3 & 4), it can be seen that the localization prediction somewhat improves, although they still appear to be off by a few meters. The controlling factor behind the quality of the localization predictions is how accurately the original first few GCPs can be placed between the input image and the reference image; accurately placed initial GCPs result in more accurately predicted GCPs.

Figure 2: GCP #4

3|Page

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 3: GCP #5

4|Page

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 4: GCP #6

5|Page

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 5: GCP #7

6|Page

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

3.2 Question B: Resample Photo Pixel Size The pixel size of the 3 resampled and corrected photos slightly varied. As can be seen below in Figures 5 6 and 7, the default pixel size was around 0.50 meters. Photo 1 was 0.5025457754 meters, photo 2 was 0.502128055 meters and photo 3 was slightly smaller at 0.4987115249. Although the photos were all acquired by the same sensor under the same conditions at the same time, they still resulted in slightly varied cell sizes. This was because of the geometric corrections that took place while placing the GCPs. The GCPs modified and stretched the images according to the Reference image, resulting in pixel sizes which are specific to each individual photos GCP placement.

Figure 6: Photo 1 Resampled Cell Size

Figure 7: Photo 2 Resampled Cell Size

7|Page

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 8: Photo 3 Resampled Cell Size

8|Page

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

3.3 Question C: Geometric Polynomial Correction-Mosaic Output Quality

Figure 9: Mosaic of Polynomially-Corrected Photos

As seen above (Figure 8: Mosaic of Polynomially-Corrected Photos), the resulting mosaic from the polynomial corrected photos was, at first glance, very successful and accurate. At a full extent view 9|Page

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

there doesnt appear to be any geometric problems whatsoever; only when zoome d in to the point where features are recognizable can any geometric problems be seen. Figure 9 is a zoomed in snap shot of a section of road. It is clear at this level of scale that the overlap is spatially offset along the x-axis, but only by a matter of meters.

Figure 10: Example of Mosaic Overlap

Further examination of the mosaic at a zoomed in scale provides more examples of geometric problems. Figure 10, pictured below, shows a section that is spatially offset, still by only a matter of meters, along the x-axis as well as along the y-axis.

10 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 11: Example of Mosaic Overlap

When examining the overlay of the vector files, a similar amount of offset can be seen. Figure 11 (pictured on the following page) shows how the building polygons are offset on the x and y axis; again, only by a matter of meters. Figure 12, also pictured on the following page, shows how the roads layer is off as well by only a matter of meters. These slight offsets in the overlaps have been caused by the combination of the three photos and their GCPs. Each photo correction required choosing where to reference from (GCPs) manually and arbitrarily; we didnt have set features or location to use in the referencing and so when the resampling was carried out each photo would have been geometrically corrected slightly different, which in turn resulted in the slight offset or shift between the photos as well as between the photos and the vector files.

11 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 12: Mosaic Vector Overlay Example

Figure 13: Mosaic Vector Overlay Example

12 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

3.4 Question D: Prediction Process Following the input of the first 3 Ground Control Points (GCP) for the Orthorectification, the same setting for predicting further GCPs is utilized again. As is visible in Figures 13, 14 and 15 which can be seen below, during the creation of this deliverable the prediction process was giving good localization on the latter GCPs. Figure 13 displays the first prediction (GCP #5), where it can be seen how accurate the prediction was; the predicted GCP was only a couple of meters from the actual location. When examining the screen shots for GCPs 6 and 8 (Figures 14 & 15), it can be seen that the localization prediction improves even more, practically predicting to within centimeters, instead of within meters as was seen with the polynomial correction process. The controlling factor behind the quality of the localization predictions for orthorectification comes from the fiducials that were entered in. Unlike the polynomial correction, for orthorectification we enter in coordinates for 8 fiducials marks along the edges of the photos which act almost like a reference grid for the orthorectification process. This allows the predicted GCPs to have a reference grid which works together with the control points to accurately predict more GCPs.

Figure 14: GCP #5

13 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 15: GCP #6

14 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 16: GCP #8

15 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

3.5 Question E: Resample Photo Pixel Size The pixel size of the 3 orthorectified photos slightly varied. As can be seen below in Figures 5 6 and 7, the default pixel size was around 0.50 meters. Photo 1 was 0.5017155399 meters, photo 2 was 0.5013081474 meters and photo 3 was slightly smaller at 0.4998729349. Although the photos were all acquired by the same sensor under the same conditions at the same time, they still resulted in slightly varied cell sizes. This was because of the geometric corrections that took place while placing the GCPs. The GCPs modified and stretched the images according to the Reference image, resulting in pixel sizes which are specific to each individual photos GCP placement.

Figure 17: Photo 1 Resample Cell Size (Ortho)

Figure 18: Photo 2 Resample Cell Size (Ortho)

16 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 19: Photo 3 Resampled Cell Size (Ortho)

17 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

3.6 Question F: Orthorectification-Mosaic Output Quality

Figure 20: Orthorectified Mosaic

18 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

As seen above (Figure 19: Orthorectified Mosaic), the resulting mosaic from the Orthorectified photos were, at first glance, near perfect except for one section near the bottom. Near the southern end of the area of interest the roads can be seen not lining up and shifting over roughly 50 meters, even at the full extent scale. Figure 20 is a zoomed in snap shot of a section of road. It is clear at this level of scale that the overlap is spatially offset along the x-axis by roughly 50 meters.

Figure 21: Example of Mosaic Overlap

This shift in the overlap was the largest shift among the two types of photo processing used in this report. This mosaic would have been perfect if it were not for the problems with photo 3. These geometric problems all fall within the area of photo 3 overlap. This suggests that the orthorectification of photo 3 was not done properly; perhaps the fiducials were not located accurately or the predicted GCPs were not accurate and not corrected for that photo. As a result when the photo was mosaicked with the other two it was considerably offset. Figure 21, pictured below, shows how the vector layers on top of the photo 3 areas are shifted as well. The red squares depict how far off the road network was and the small arrow shows where that building polygon should really be located if it was 100% accurate.

19 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

Figure 22: Example of Mosaic Overlap

4. Conclusion
This report detailed and explained the results of orthorectification, geometric correction and mosaicking of three aerial photographs. Orthorectification utilizes one image, known as the Reference Image to correct other images of the same general location by locating areas of commonality between the two known as Ground Control Points. Using these ground control points together with the Reference Image, the photos are resampled and re-projected, adhering to the parameters of the reference image and any fiducial marks which are added on in the orthorectification process. As depicted in the discussion, the geometric correction process did not predict points as accurately as the orthorectification, and as well did not utilize fiducial marks, resulting in a mosaic of lesser quality; there were overlap offsets and shifts throughout the entire mosaic. Furthermore the vector layers (buildings, roads) were also shifted and offset throughout the entire mosaic. In conclusion, the orthorecitification method of processing aerial photos is superior to the geometric correction process.

20 | P a g e

[135 Taylor Road Niagara-on-the-Lake Ontario L0S 1J0] [GeoTech Solutions Inc.]

5. Bibliography
"ERDAS Field Guide." Intergraph. (2013) Online. <http://geospatial.intergraph.com/Libraries/Misc_Docs/ERDAS_FieldGuide_PDF_Intergraph_br and.sflb.ash&xgt.

*Cover photo courtesy of iWakes Geomad

21 | P a g e

Das könnte Ihnen auch gefallen