3.3 Imagery standards
99
Similarly the image analysis is not yet apart of the ISO 19100 family of standards. It includes edge detection, feature extraction, and image understanding.
Georeference
Georeference is the reference of a dataset to a location on the earth. Georeferencing information is the rationale of geographie information. From the procedural point
of view some datasets are georeferenced once they are created while other datasets
need to be georeferenced at some later time. These are called ungeoreferenced. Surveyed elevation points, small scale maps, or ortho photos are examples for georeferenced data. The raw data of remotely sensed imagery is an example for ungeoreferenced data.
Typically ungeoreference imagery data for geographie application has some additional information such as control points or orientation data that enable the process
of georeferencing. These data are called georeferenceable. Ungeoreferenceable datasets, such as a portrait of aperson, have no such additional information and are only
of theoretical interest.
There exist two basic methods to establish the georeference of an imagery dataset:
the image reference and the sensor reference.
The image reference uses image information and control points only. A geometrical model of the sensor that took the image is not required. This method has been
widely applied for remote sensing imagery. The image reference can be established
for instance with a 2-dimensional polynomial function that relates the image to the
surface of the earth based on any number of common points called control points.
The simplest way to georeference an image is the definition of the ground coordinates of the image corners. The image reference is considered an approximate
method because terrain undulations between the control points can cause undetectable height parallaxes in the georeferenced image.
The sensor reference uses a mathematical model of the sensor such as the central
perspective of a photogrammetric camera. This method requires a lot more initial parameters such as the focal length of the sensor and the position and attitude of the
sensor while the image was taken. However, two or more images ofthe same region
allow for a 3-dimensional data capture, for instance the creation of digital elevation
model. A digital elevation model is required to create georectified orthophotos (see
below) that are free from height parallaxes.
The sensor reference method is the domain of the classical photogrammetry. The
advent of high resolution satellite sensors with metre-pixels on the earth's surface
have stimulated a wider use of the sensor reference method in remote sensing. The
sensor model is defined in the ISO 19130 (Sensor and data models for imagery and
gridded data), chapter 3 section 3.2.
The most common geometry of georeferenced gridded data is the ortho projection. It denotes a vertical projection of the dataset to the reference plane like the xyplane of astate Coordinate Reference System. Coverage-type datasets are based on
points with coordinates like an elevation model or a thematic layer. If these datasets
have 3 or more dimensions they are rectified by simply selecting two of the available
99
Similarly the image analysis is not yet apart of the ISO 19100 family of standards. It includes edge detection, feature extraction, and image understanding.
Georeference
Georeference is the reference of a dataset to a location on the earth. Georeferencing information is the rationale of geographie information. From the procedural point
of view some datasets are georeferenced once they are created while other datasets
need to be georeferenced at some later time. These are called ungeoreferenced. Surveyed elevation points, small scale maps, or ortho photos are examples for georeferenced data. The raw data of remotely sensed imagery is an example for ungeoreferenced data.
Typically ungeoreference imagery data for geographie application has some additional information such as control points or orientation data that enable the process
of georeferencing. These data are called georeferenceable. Ungeoreferenceable datasets, such as a portrait of aperson, have no such additional information and are only
of theoretical interest.
There exist two basic methods to establish the georeference of an imagery dataset:
the image reference and the sensor reference.
The image reference uses image information and control points only. A geometrical model of the sensor that took the image is not required. This method has been
widely applied for remote sensing imagery. The image reference can be established
for instance with a 2-dimensional polynomial function that relates the image to the
surface of the earth based on any number of common points called control points.
The simplest way to georeference an image is the definition of the ground coordinates of the image corners. The image reference is considered an approximate
method because terrain undulations between the control points can cause undetectable height parallaxes in the georeferenced image.
The sensor reference uses a mathematical model of the sensor such as the central
perspective of a photogrammetric camera. This method requires a lot more initial parameters such as the focal length of the sensor and the position and attitude of the
sensor while the image was taken. However, two or more images ofthe same region
allow for a 3-dimensional data capture, for instance the creation of digital elevation
model. A digital elevation model is required to create georectified orthophotos (see
below) that are free from height parallaxes.
The sensor reference method is the domain of the classical photogrammetry. The
advent of high resolution satellite sensors with metre-pixels on the earth's surface
have stimulated a wider use of the sensor reference method in remote sensing. The
sensor model is defined in the ISO 19130 (Sensor and data models for imagery and
gridded data), chapter 3 section 3.2.
The most common geometry of georeferenced gridded data is the ortho projection. It denotes a vertical projection of the dataset to the reference plane like the xyplane of astate Coordinate Reference System. Coverage-type datasets are based on
points with coordinates like an elevation model or a thematic layer. If these datasets
have 3 or more dimensions they are rectified by simply selecting two of the available