4.3 Georeferencing and geocoding

Question 1

Georeferencing

Answer 1

• The main objective is to relate the image coordinates to a specific map coordinate system.
• Image coordinates are linked to map coordinates using a transformation model.

(x,y) = T (i,j)
(i,j) = T-1 (x,y)

Question 2

Georeferencing two major steps

Answer 2

–Selecting the appropriate type of transformation
•Depends on the sensor system and the platform.
•Linear transformations
•Polynomial transformations ( 1 st , 2 nd , … n th order).

–Determination of transformation parameters
•Ground control points are used

Question 3

Image transformation

Answer 3

Linear transformations:
–Translation
–Rigid (Isometry): Translation and rotation
–Similarity: Translation, rotation, and scaling
–Affine: Translation, rotation, scaling, and shear
–Projective: Translation, rotation, scaling, shear, projective distortion

Question 4

Linear transformations - Translation

Answer 4

Two degrees of freedom
x2 = x1 +a
y2 = y1 +b

Question 5

Linear transformations - Rigid (Isometry transformation)

Answer 5

– Is a combination of translation and rotation

– Three degrees of freedom ( 2 translation + 1 rotation)

Question 6

Linear transformations - Similarity

Answer 6

–Is a combination of translation and rotation and scaling
–The ratio of two lengths are preserved.
–Four degrees of freedom ( 2 translation + 1 rotation + 1 scale)

Question 7

Linear transformations - Affine (allowing for or preserving parallel relationships.)

Answer 7

–Is a combination of translation and rotation and scaling and shear
–Parallel lines remain parallel.
–Six degrees of freedom ( 2 translation + 1 rotation + 1 scale + 1 aspect ratio + 1 shear)

Question 8

Linear transformations - Projective (has 8 degrees of freedom or 8 parameters)

Answer 8

–Most general form of linear transformation
–Parallel lines may not stay parallel, but lines preserved in transformation.
–This represents perspective distortion caused by the camera converting a3 D world to a 2 D image.
–Eight degrees of freedom
–For aerial photographs when the terrain is flat projective transformation iscommonly used.

Question 9

Polynomial transformations

Answer 9

•1 st order transformation
–Six degrees of freedom

•
Higher order polynomials can also be used.

Question 10

Transformation parameters

Answer 10

• Transformation parameters can be determined by means of Ground controlpoints (GCPs)
• GCPs are points that can be clearly identified in both image and on a map (e.g.road crossings and noticeable morphological structures
• Alternatively, GCPs can be identified on the image and then the coordinates measured in the filed by GPS
• The coordinates of GCP in the map and image are then used to estimate the parameters of the transformation model.
• The transformation model s degree of freedom defines the minimum number of needed GCPs, but it is highly recommended to use more number of points to identify errors in the measurements.
• In low-resolution images it is difficult to identify reliable GCPs.

Question 11

Geocoding

Answer 11

• Georeferencing assigns map coordinates to each image pixel.
• Geocoding is georeferencing with subsequent resampling of the image.
• The row-column structure of the image is also transformed to the map coordinate system. As a result a new raster image is defined along the xy axes of map projection.

Question 12

Geocoding is required when

Answer 12

–Data fusion or analysis of data from different seasons.
–Multi temporal analysis such as change detection
–Mapping and cartography applications.
–Overlaying images with existing datasets or maps.

Question 13

Major steps of geocoding

Answer 13

–Transform the image to map coordinate
–Define an image grid in map coordinate
–Each new image element is projected onto original image
–Pixel value for the new pixel is determined and stored

Question 14

Image Resampling

Answer 14

• The orientation and size of input and output images are different
• There is no one to one relationship between their pixels.
• We have to apply an interpolation method to obtain the pixel value inoutput image.
• Usually simple interpolation methods are used.

Question 15

Interpolation methods - Nearest Neighbor (1)

Answer 15

–The value of nearest pixel in input image is assigned to output
–Edges can be offset in a step like pattern
–spectral information is retained
–Spatial information may be altered because some original pixels may be omitted or appear twice

Question 16

Interpolation methods -Bilinear Interpolation ( 4

Answer 16

Average of four nearest pixels are assigned to the output

Question 17

Interpolation methods -Bicubic Convolution ( 16

Answer 17

Average of 16 nearest pixels are assigned to the output

• Smoother image
• Radiometric information altered because of averaging.