Visual interpretation Chapter 10. Visual interpretation in important to GIS development and...

Post on 20-Dec-2015

261 views 0 download

Tags:

Transcript of Visual interpretation Chapter 10. Visual interpretation in important to GIS development and...

Visual interpretation

Chapter 10

• Visual interpretation in important to GIS development and application.

• It is the interpretation of aerial photography by a photo interpreter.– Regarded as somewhat useful but old-

fashioned likely to be replaced by digital analyses

• Visual interpretation is now seen in the new context.– Skills that a photo interpreter uses are applied

equally to photographs as well as digital images

– So a new name has come into use that of image analyst or image interpreter

• In addition, remote sensing data has become more detailed, much like photography

• Finally, computer science is not evolved to the point where image analysis can be done completely automatically

Characteristics of aerial imagery

• Visual interpretation is used to extract information from fine resolution imagery at scales of 1:40,000 or larger– And includes almost all aerial photography

and fine scale satellite imagery (e.g. IKONOS)– Excludes coarse resolution imagery produced

by satellites such as Landsat

Characteristics of aerial imagery

• GIS requires the following conditions be met– Geometric errors removed– Detailed spatial information is contained– Can provide taxonomic details such as separation of

corn from wheat– Spatial relationships can be examined such as

vegetation distribution and water bodies– Patterns of change from one day to the next are

monitored– Equipment and experienced staff are available– Historical phenomena can be examined such as land-

use change

Sources of imagery

• Two common sources– Archival imagery previously acquired for

another purpose.– Imagery acquired specifically for the project at

hand

Archival imagery

• Several sources are available.– USGS– US National Archives and Records

Administration– State agencies

• Department of Transportation• Department of natural resources.

• Such imagery must often be bought

Archival imagery

• In most cases, computerized databases permit online searches.

• Many states update photography on a regular basis.– These are useful in examining change

Custom acquisition

• Imagery is acquired, according to the user specifications.– The date scale emulsion coverage and other

characteristics are determined by the user.

• The disadvantages include – high cost – requirement for planning well in advance. – effort devoted to preparing and monitoring the

specifications– delays and cancellations due to weather or equipment

malfunctions

Custom acquisition

• Users should prepare a contract or statement of work– Specifies in detail the products and services

required from the firm with respect to cost, deadlines, and products to be prepared

• Contract is a means for controlling the completeness and quality of the imagery

Custom acquisition

• The photo interpreter may participate in making decisions about scale, date, time of day, deadlines, choice of film, and coverage

Custom acquisition

• Photo interpreter should ask:– Will imagery be adequate for the purpose of

the project?– Is the film emulsion satisfactory for the task?– Is the season suitable?– Will planned time of day provide appropriate

shadowing?– In stereophotography, what will the overlap

be?

Elements of image interpretation

• Image analysis requires explicit recognition of eight elements of image interpretation that form the framework and understanding of an image– Shape– Size– Tone– Texture– Shadow– Site– Association– pattern

Shape

• The outline of a feature– Important to note that shape depends on

perspective– Overhead perspective, introduces scale effect

Size

• The first to the dimensions of a feature

• Relative size determined by comparing the object with familiar nearby features

• Absolute size refers to the use of the aerial image to derive measurements

Tone• Refers to the average brightness of an area or,

in the case of color imagery, to the dominant color of the region– Depends on the nature of the surface in the ankles of

observation and illumination.– Smooth surfaces behave like specular reflectors,

they tend to reflect radiation in a single direction• These features may appear bright or dark

– Rough surfaces behave this diffuse reflectors.• Scatter radiation in all directions.• A peer is medium gray tones

Texture

• Refers to the variation in tone over a surface or the apparent roughness of the surface as seen in the photo

• Created by micro shadows in small irregularities in the surface.

Shadow

• Refers to large distinctive shadows that revealed the outline of a future as projected onto a flat surface.– Depends on the nature of the object, angle of

illumination, perspective, and slope of the ground surface

Site

• Refers to a futures position with respect to topography and drainage.– Some things occupy a distinctive topographic

position because of their function• Sewage treatment facilities at the lowest feasible

topographic position.• Power plants located adjacent to water for cooling

Association

• Association refers to the distinctive spatial interrelationships between features– Schools often associated with athletic fields.– Large parking lots often associated with malls

Pattern

• Refers to distinctive arrangement of features– Orchards have trees plant can rows– Mobile home parks have rectangular buildings

arranged in rows

Image interpretation tasks

• Classification - Assign objects to classes– Detection, recognition, identification.– Interpreter confidence is ranked as possible,

probable

• Enumeration - Count items• Measurement- mensuration

– Uses image scale to derive measurements such as length with distance and volume

• Delineation – demarcation of regions

Interpretation strategies

• Field observation.– Required when the image and its relationship to

ground conditions are imperfectly understood

• Direct recognition.– Interpreter derives information directly from inspection

of the image

• Inference– Based on visible features can derive information

about invisible features– Certain relationships and identify features

Interpretation strategies

• Interpretive overlays.– Useful when relationships between visible patterns

are used to reveal patterns not directly visible.– Soils may be revealed by relationships with

vegetation patterns, slope, and drainage

• Photomorphic regions– Identify regions of uniform appearance on an image.– Does not attempt to resolve individual landscape

components.• Mostly used with small-scale imagery in which coarse

resolutions tend to average separate components

Interpretation strategies

• Image interpretation keys.– Keys are reference information designed to

enable rapid identification of features.– Usually a collection of annotated images or

stereograms and a description, which may include sketches or diagrams

Aerial mosaics and image maps

• IMAP is characterized by the planimetrically accurate representation of your surface and uses symbols to represent select features

• An aerial image is not really a map because Tilton relief displacement introduced your metric errors– Its features are represented without selection

or symbolization

Aerial mosaics and image maps

• A aerial images form the basis of useful map-like representation of the Earth’s surface such as mosaics and orthophotos

• Aerial mosaics are produced by assembling adjacent aerial photographs to form a single image

Aerial mosaics and image maps

• Uncontrolled mosaics performed by assembling adjacent photographs without strict concern for geometric integrity

Aerial mosaics and image maps

• A controlled mosaic presents the detail of an aerial photograph in planimetrically correct position

Aerial mosaics and image maps

• An aerial index is created by placing adjacent area photographs together in their approximate relative position without represent econometric relationships– These are to be used as a guide for

identifying those photographs that represent an area that is needed without searching the entire collection

Aerial mosaics and image maps

• Orthophotos are aerial photographs prepared using stereoscopic parallax and photogrammetric principles to remove the effects of relief displacement and tilt to provide a planimetrically correct aerial image

Aerial mosaics and image maps

• And orthophoto quadrangle is an orthophoto that represents the same region as the USGS topographic quadrangle– These can be thought of as image maps

• The digital orthophoto quadrangle (DOQ) is a digital version

Field observations and accuracy assessment

• Field observations are necessary because aerial imagery is not infallible or a total source of information– The information is derived through

interpretation

Field observations and accuracy assessment

• Think of a map or GIS as a statement about conditions of a certain place– The more precise the statement, the more

useful it is to the user• Delineate small parcels, specific labels, etc.

– The less precise the statement, the less useful

• The larger the parcels, more general labels, etc.

Field observations and accuracy assessment

• For example, a broad area labeled as a forest– Versus

• Smaller areas each individually labeled as oak, pine, and fir.

Field observations and accuracy assessment

• Field data are first-hand observations collected on the ground– Often use copies of images or maps to

annotate in the field

• Field work can create delays, increased expense, and other difficulties– But it is essential for establishing the validity

of the work

On-screen digitizing

• Traditionally, photointerpreters used transparent (mylar) overlays to show outlines, mark features, and make annotations – These then had to be digitized, transferred

and registered to a planimetrically correct base layer

Manual Digitizing

Most common form of coordinate data input

Requires a digitizing table – Ranging in size (25x25 cm

to 150x200cm)

• Ideally the map should be flat and not torn or folded

• Cost: hundreds (300) to thousands (5000)

On-screen digitizing

• In digital imagery it requires taking a raster image and putting a drawing a vector layer on top of it.– This all can be saved as one file– Can then be transferred directly to a GIS

• This is called heads-up digitizing

• Raster-scanned image on the computer screen

• Operator follows lines on-screen in vector

mode

Heads-Up Digitizing II

Digitizing Errors

• Undershoots

• Dangles

• Spurious Polygons

On-screen digitizing

• Not inherently difficult

• Requires consistency and attention to detail

• Can be challenging– Edges can be difficult to identify– Shadows– Low image contrast

Accuracy Assessment

• Procedures well established to evaluate the accuracy of interpretation– For maps with polygons of specific types, an overall

map accuracy and individual class accuracies can be determined

• Requires a set of reference data points that have known characteristics– Ideally should be random– That may miss some categories, so, often random

selection combined with collecting across all classes and across the entire study area

Accuracy Assessment

• Should keep field collected data for input separate from, field data used for accuracy assessment

• Assessment is a table called the error matrix– Each reference point and its matching map

label are compared

The kappa statistic

• There are so many values in the error matrix that it becomes confusing to people.– Want a single number that encapsulates the

significance of the error matrix.– This is the kappa (κ) statistic

• Difference between the observed agreement reported by the diagonal and the agreement that might be due solely to chance

The kappa statistic

• Kappa is estimated by k hat

• Observed = accuracy reported in error matrix– Sum correct / total

• Expected = correct classification due to chance

ected

ectedobserved

exp1

exp

The kappa statistic

• The statistic adjusts the percentage correct by subtracting the estimated contribution of chance agreement– A κ=0.83 means that the classification is 83%

better than would be expected from chance assignment of pixels to categories

– Kappa ranges between +1 and -1

Interpretation Equipment

• Paper prints or transparencies

• Magnification

• Flatbed scanners– Dpi – up to about 3500

• Light tables

• Densitometers

• Stereo imagery equipment