1. Retrieval/classification/measurement

2.Overlay (arithmetic, various conversions)

3.Neighbourhood

4.Connectivity

Retrieval: Selective search

-addresses selected because they fall within a circle

Reclassification

-overlays, combine

-identifying a set of features as belonging to a group

-defines patterns

(Vector)

-dissolving to aggregate polygons

-When there are identical polygons next to each other, can merge them into the same polygon

(Reclassify by Area Size)

-ie. work with areas over Xsize...

-eliminates all the smaller areas

(Reclassify by Contiguity)

-work with individual areas, rather than the class as a whole

-ie.. # of areas that are the same... can reclassify them as #1,2,3, etc

(Reclassify Values)

-zones with elevations between 20 and 40 fet

-can change feet to metres

Measurements

-buffers

-distances, lengths, perimeters, areas

(Buffering Operations)

-raster difficult, vectors easy

-creates a distance from a feature

-works with points and lines and polygons

-legal descriptions.. Nature rarely pays attention to specific distances

**streams flood by elevations, not distance from stream

**Eagles hunt where the food is, not within a 1/4 mile circle of their nest

**Volcanoes erupt based on internal mechanics influenced by topography, not neat rings of hazard

Vector Distance Operation: Buffers and Setbacks

-buffers go outward from lines or areas (rounded ends/corners), while setbacks run inside of areas (and cannot exist for lines)

Buffer creation

-each line/segment of polygon will have a rectangle and a half circle on the ends

To generate a buffer, construct these objects around each segment, overlay all the objects, aggregate to remove duplicate areas

arithmetic/various conversions

First law of Geography: Everything is related to everything else, but near things are more related to each other.

Arguable, the most impotant feature of any GIS is its ability to combine spatial datasets...

How?... Arithmetic! (addition, subtraction, division, multiplication)

-Use overlay functions to find wehre specified conditions occur (x and y, x or y, >, <, etc)

Raster and Vector methods differ...

-vector good for sparse datasets (many polygons makes difficult overlays)

-raster grid calculations easier (as long as the overlay is meaningful).... (prob the most misused thing in GIS)

Most common overlay is a Suitability Analysis

ie. gov administrative units over land use over soils over suitability for industrial developemtn

-Spatial location is key!

End up with a bunch of overlain polygons or rasters.. differences do mean something!!

1. Enumeration rules: each attribute preserved in output

-all unique combinations recognized

2. Dominance Rule: one value wins

-operation consists of choosing one value. Examples: maximum; highest bidder.

3. Contributory Rule: Each attribute contributes to results

-operations like addition allow each source to contribute to result

-when you don't need to preserve each attribute

-this one gets people in trouble... can't just add everything together and assume it is the same output.

-Don't add things to nominal/ordinal

-for the sake of GIS, interval/ratio are the same thing.. depend on zero.. can still add/multiply/etc.

4. Interaction RuleL pair of values contribute to result

-decisions in each step may differ

-complicated

-the way we add things together may differ.

3 Main types of Vector Overlay

Point-in-Polygon (find the points in polygon)

-ie. how many weather stations are in forest/non-forest categories

-result: table of intersections

Line-in-Polygon

-ie. how a road runs through same region

-result: new line segments coded with forest classes

Polygon-in-Polygon

-use logical operators (and/or/not)

-don't really use these anymore

AND= spatially coincident

OR= I don't care which (can be both also)

Simple Addition

-level of measure important: some kinds of data can not be added

ie. scored  grid of resource map A + B + C = composite grid

Boolean Combine

-Geometric phase (common grid framework ensures that each pixel overlays in the same position

-Attribute phase (threshold chosen between suitable and not.

-operation (combination rule) produces result

-uses AND/OR with rasters (ie. where they overlap, or all of them)

Composite Combine

-combine elements of the geometric phase

-operation creates new categories for all distinct combinations discovered

-From the composite categories, any result can be obtained

-can select for category types

Vector Overlay: Composite Structure

Geometric Phase: First phase discovers all intersections between input linework, creates topological structure

-second phase identifies all polygons with unique ID and link to parent attributes

Attribute Phase: Using the attribute table, a query can extract combinations of attributes (like in Composite combine)

**STudy diagrams for last two!!!***

3 Outcomes:

-Target locations

-specifications of neighbourhood

-Derivation of data from a function performed on elements in neighbourhood

Basic Functions

-Average, diversity, majority, min/max, and total

-basically summary statistics

Parameters to define

-Target location(s)

-specification of a neighbourhood

-Function to perform on neighbourhood elements

Search Operation

-most common neighbourhood operation

ie. count the # of customers within 2 miles of grocery store

Poit or Line in Polygon Operation

-Vector model (specialized search function)

-Raster model (polygons one data layer, points or lines in separate data layer)

Thiessen Polygons Operation

-Defines the individual areas of influence around a point

-Used to predict values at surrounding points from a single point observation

-Can produce olygons with shapes unrelated to phenomenon being mapped.

-Lines are all halfway between points

-How far away do I need to be before I build another airport? Hospital? etc..

Polygon Overlay Operations

Simplest Form: Boolean (Sieve mapping)

4 Steps:

-determining the inputs

-getting the data

-getting the spatial data into the same coordinate system

-overlaying the maps

Overlay Operations (in ArcGIS)

-Union

-Intersect

-Identity

*Unlikely to ask for Boolean names for intersections... more likely to ask how it works...

According to Chou (1997) a Spatial Model:

1)Analyzes phenomena by identifying explanatory variables that are significant to the distribution of the phenomenon and providing information about the relative weight of each variable

2)Is usefor for predicting the probable impact of a potential change in "control" factors (independent variables)

Models can be:

-descriptive or Prescriptive

-deterministic or stochastic

-static or dynamic

-deductive or Inductive

STEPS in modeling process

1)Define the goals of the model

2)break down the model into elements

3)implementation and calibration of the model

-model validation (sometimes difficult or not feasible)

GENERAL

1)Descriptive: characterization of the distribution of spatial phenomena

2)Explanatory: deal with the variables impacting the distribution of a phenomena

3)Predictive: once explanatory variables are identified, predictive models can be constructed

4)Normative: models that provide optimal solutions to problems with quantifiable objective functions and constraints.

SPECIFIC

1)binary models (descriptive): use logical expressions to identify or select map features that do or do not meet certain criteria... how?

2)Index models (desc.): use index values calculated for variables to produce a ranked spatial surface ... How?

-weighted linear combination model

3)Regression models (explanatory/predictive): a dependent variable is related or explained by independent variables in an equation.. How?

-linear and logistic regression

4)Process (explanatory or predictive): integrate existing knowledge aboutenvironmental processes into a set of relationships and equations for quantifying those processes.. How?

How can GIS enable spatial modeling?

-GIS is a tool that can integrate a myriad of data sources

-GIS can incorporate raster and/or vector data into modeling schemes

-Modeling may take place within a GIS, or require linking to other computer programs

---loose coupling

---Tight coupling

---Embedded system (writing own software)

Important Issues in Conducting Spatial Analysis:

-Delineation of geographic units of analysis

---How do you choose geographic units of analysis so that spatial analyses are valid? (no answer... everything you do depends on the Q you ask)

-Identification of structural and spatial factors that impact spatial analysis

---structural - impact site

---spatial - impact situation (absolute and relative location, neighbourhood effects)

Surface is the start point

Surface Functions....

-Density, contour, interpolation (take points and turn into areas)

-aspect, slope, hillshade, etc.

-watershed analysis and modeling (flow direction, flow accumulation, flow length, watershed delineation, stream ordering)

-visibility modeling/mapping (determine the area that can be 'seen' from the target location)

Surface Analysis Function in modeling

Interpolation Types: Generating Spatial Data

1)Deriving continuous data from irregular spaced points

2)DEMs

3)Global vs Local

4) Conceptual break down of the processes

5) Some work by considering entire point ditributions (global)

6) Some work by taking a sample of the points

Regular spaced data is already continuous*

CLASSIFICATION OF SPATIAL INTERPOLATION

Global vs Local

A global interpolation method uses every known point available to estimate an unknown value. A local interpolation method uses a sample of known points to estimate an unknown value.

Exact vs Inexact

exact interpolation predicts a value at the point location that is the same as its known value. In other words, exact interpolation generates a surface that passes through the control points (no error @ point).

Inexact interpolation or approximate interpolation predicts a value at the point location that differs from its known value. Almost every points has error.

SCALE DEPENDENCY

If you have a high density of sample points:

-capture local variation

-approximate for large-scale (small-area) studies

If you have low density of sample points:

-lose sensitivity of local variation and capture only the regional variation

-appropriate for small-scale (large-area) studies

DETERMINISTIC VS STOCHASTIC

A deterministic interpolation method provides no assessment of errors with predicted values.

A stochastic interpolation method offers assessment of prediction erros with estimated variances

** Any surface you generate is always and ESTIMATE!**

GLOBAL

Deterministic: Trend surface analysis (inexact)*

Stochastic: Repression (inexact)

LOCAL

Deterministic:

-Theissen (exact)

-Density estimation (inexact)

-inverse distance weighted (exact)

-splines (exact)

Stochastic: Kriging (exact)

*= given some required assumptions, trend surface analysis can be treated as a special case of regression analysis and thus a stochastic method.

**Exact means it MUST hit the points (ie. has to use the points it is given)

DETERMINISTIC PROCESSES

Defines a condition where the outcome is known and not subject to random variation.

Equations without error terms etc (vastly oversimplified).

Newtonian mechanics.

There is method - not madness

But can error be evaluatied??

2 ways..

1)reduce search area

2)Randomly hold out points, then throw them back in once surface is created

-how well did observed meet predicted?

STOCHASTIC PROCESSES

A stochastic process is a random function.

The domain over which teh function is defined is a time interval (a time series) or a regiona of space (a random field).

Has uncertainty in the prediction (an error term)

Has some degree of indeterminancy in the outcome

-From the initial condition there are many possible ways that you could cover space.. some are more probablyl than others.

Error Term

Ie... 3 points connected with elevation and distance... some ways of connecting them are more likely than others. Use stats.

IMPLICATIONS

Global methods are restrictd to trend surfaces

Global methods are all inexact

We can only gather error information from Krigin.

Creats surfaces that are largely correct, but not complete

-any GIS surfae is not complete

-it represents a sampling and guess

-quality of guess determines goodness

First order (linear) trend surface. (ie. a hill)

or

Second order (quadratic) trend surface (ie. a valley)

Global Fit..

An isoline map of a third-order trend surface created from 105 points with annual precipitation values

- end up with negative precipitation values in the corner (not possible)

Thiessen (Voronoi) polygons:

-assume values of unsampled locations are equal to the value of the nearest sampled point (climatology)

Vector-based method

-regularly spaced points produces a regular mesh

-irregularly spaced points produces a network of irregular polygons

(final compare and contrast)

Thiessen Polygon Construction..

-Get all distances b/w points

-Then take each distance, and divide it by half

-can do for elevation, etc.

The "other" vector-based method.

Another vector-based method often used to create digital terrain models (DTMs).

Adjacent data points connected by lines (vertices) to create a network of irregular triangles

-calculate real 3D distance b/w data points along vertices using trigonometry

-calculate interpolated value along facets b/w three vertices

-creates topology in 3D

-everything has to be a triangle

Used for video games...

-easy to render

-can be done fast

-looks right

-Widely used in GIS for medeling!!!

DEMs and their usage in GIS

-Geovisualization (making things pretty)

-Contours

-Modeling

Terrain Derivatives are the rates of change of elevation (ie. slope)

-Calculus

-Fitting surfaces

-Significant for moedeling systems

Slope and Aspect

The vector that defines magnitude of slope has a direction (Aspect) in calculus.

Aspect shows direction the slope faces.

Now regard to magnitude of slope.

CALCULATION

Only way to calculate these is to do it over a kernel of some size, usually 3x3.

A kernel is a square, or sub array

-there is NO slope to a single cell!

-Slope can be over 3x3, 5x5, 7x7, etc..

The value calculated is then assigned to the central cell of the kernel.

So the accuracy of slope and aspect (and ALL derivative mapping) is NOT the same as the original data!!!

It represents a slightly larger area!

(~60% larger)

-Qhwn you get to the outer layer, cannot surround. Either use zeroes, copies, or extrapolations

Determining slope always smoothes (filters)

-Since these values are for a 3x3 kernel, the image is effectively smoothed and simplified.

Removes high frequency data

HOW IS IT DONE?

Finite Difference Approach

-2 ways, in 3x3..

rook's case: based on two slopes (N/S; E/W)..

-only uses 4 data points

-can't move diagonally

-don't use middle elevation for calculation

-uses only some of the data, but processing time is quicker..

queen's case: based on 6 slopes (3NS/3EW)

-uses 8 data points

-again, not the middle one

-we can calculate slope for AB at 20m elevation.. takes a day

-will get same result as rook's...

Caclulation is done using the summation of the partial derivatives in X and Y (ie. rise over run)

-change x/change y

OTHER Slope Calculation Possibilities

Statistically fitting a surface to elevation points

Then take the surface formula and take the first derivative.

-A curved surface is fit to the elevation points...

-a complex operation

Generally slower, no real benefit

so, THREE ways: queen/rook/surface

-same results

-Rook is faster

____

Slopes govern certain processes which occur on them!!

-ie. soil development..

Curvature isthe rate of change of slope (second derivative of elevation.. slope of slope)

Important for classifying terrain

Tells us where the terrain is concave, convex, or flat

MEASURING CURVATURE

Can be measured as an absolute magnitude (non-directional)

-simply how curved the surface is.. no concave/convex

Laplacian curvature: retains the concavity/convexity (sign)

-error in magnitude

Directional curvature:

-down-slope and across slope

-profile plan

SIGNING CONVENTION

Calculus:

++ = concave

(cross coss smile)

-- = convex

***Geomorphological convection is the reverse...

USE?

Can indicate strength of erosive forces, areas of accumulation, and transitional zones.

DEMs are often used in the presentation of geographic data.

Provide context and visual interest.

GIS packages are capable of producing attracting renderings.

Will not work without you.

Whar are the rules?

Do not accept the default for anyting.

Artistic ability is important

-or at least stick to something that looks good.

All cartographic rules apply (Scale, design, orientation, colour selection.

COLOUR SELECTION

Colours in a map shouldn't be discordant.

Try to avoid using all the colours you have

Select colours from a palette

-pre-select colours that look good.

-controlled by use

Different set of colours for a general purpose overview map (oceans are blue, terrain is earthy)

For special purpose maps, stick to pastels.. avoid distracting colours.

___

back to... DEM PRESENTATION!

Used to show elevation chages or topographic detail.

ArcGIS defaults are B/W

OK for jounral reproduction or texts due to cost, but it could be better..

Try some of these tricks....

-defaults

-greyscale

-colour

-hillshade

-hillshae with classed greyscale

-hillshade with colour (continuous vs classed)

Three different types of errors pop up all the time.

1) Data entry errors (clearly wrong)

2) Tiling errors (putting together large data sets)

-from tilted photoraphy - misalignment of the base map.

3) Quilting (surfacing errors)

BP pipeline routing

Modeling wildfire risk (increased pop growth into the forest/urban interface raises the threat of disaster... Identify areas most likely to be impacted by fire. Effective pre-treatment, suppression, and recovery plans can be developed)

Characterizing major terrain features

-identifying extented high ridges

-identifying valley bottom land

-identifying dam reservoir extent

Forest Access model: forested areas are first assessed for Availability considering ownership and sensitive area designation and then characterized by Relative access considering intervening terrain factors of steepness and stream buffers, plus human factors of housingdensity and visual exposure to roads and houses.

Study area: UofL campus.. calculating flood areas

What is it?

Constellation of orbiting navigational satellites

-transmit satellite position and time data

Handheld receivers calculate

-lat/long

-altitude

-velocity

Developed by US Department of Defense

-battlefield positioning

-ICBM targeting

HISTORY

1969= Defense Navigation Satellite System (DNSS) formed

1973 = NAVSTAR Global Positioning System developed

1978= first 4 satellites launched (Delta rocket launch)

1993= 24th sat launched; initial operational capaitlity

1995= full operational capability

2000= military accuracy available to all users

Three Segments

SPACE SEGMENT

CONTROL SEGMENT (master and monitor stations, and ground antennas)

USER SEGMENT

24 satellite vehicles

Six orbital planes (4 satellites on each plane)

-inclined 55 degrees with respect to equator

-orbits separated by 60 degrees

20200km elev above earth

Orbital period of 11hr 55 min

Five to eight satellites visible from any point on Earth (block I sat vehicle)

GPS SATELLITE BUS AND PAYLOAD

Four atomic clocks

Two solar panels

-battery charging

-power generation

S band antenna - sat control

12 element L band antenna - user communication

Block IIF satellite vehicle (fourth generation)

Sat Bus.

-2370 pounds

-16.25 fee (tall)

38.025 feet (wide including wing span)

-design life = 10 years

 GPS antennas and reeiver/processors

Position

Velocity

Precise timing

Used by:

-aircraft

-ground vehicles

-ships

-individuals

Master control station (Colorado)

Five monitor stations

Three ground antennas

Backup control system

Located in US and ~ low latitudes

Satellite ranging
•Satellite locations
•Satellite to user distance
•Need four satellites to determine position

Distance measurement
•Radio signal traveling at speed of light
•Measure time from satellite to user

Pseudo-Random Code
•Complex signal.
•Unique to each
satellite.
•All satellites use
same frequency.
•Economical.

Distance to a satellite is determined by measuring
how long a radio signal takes to reach us from that
satellite.
To make the measurement we assume that both
the satellite and our receiver are generating the
same pseudo-random noise codes at exactly the
same time.
By comparing how late the satellite's pseudo-
random noise code appears compared to our
receiver's code, we determine how long it took to
reach us.
Multiply that travel time by the speed of light and
you've got distance.

THE GPS SIGNAL

-read over... slide 21 lecture 22

•Accurate timing is the key to measuring distance
to satellites.
•Satellites are accurate because they have four
atomic clocks ($100,000 each) on board.
•Receiver clocks don't have to be too accurate
because an extra satellite range measurement can
remove errors.

•To use the satellites as references for range
measurements we need to know exactly where
they are.
•GPS satellites are so high up their orbits are very
predictable.
•All GPS receivers have an almanac programmed
into their computers that tells them where in the
sky each satellite is, moment by moment.
•Minor variations in their orbits are measured by
the Department of Defense.
•The error information is sent to the satellites, to
be transmitted along with the timing signals.

If one satellite, know that the receiver is somewhere on the sphere

With two, know it is where the spheres intersect

With three, there is only one point = 2D position

With four, get 3d position

Standard Positioning System

-100m horizontal accuracy

-156m vertical accuracy

-Designed for civilian use

-No uer fee or restrictions

Precise Positioning System..

-22m horizontal accuracy

-27.7m vertical accuracy

-Designed for military use

Selective availability

-Intentional degradation of signal

-Cotrols availability of system's full capabilities

-Set to zero May 2000

-Reasons:

---enhanced 911 services

---car navigation

---adoption of GPS time standard

---Recreation

The earth's ionosphere and atmosphere cause delays in the GPS signal that translate into position errors

Some erors can be factored out using math and modeling

The configuration of the satellites in the sky can magnify other errors.

Differential GPS can reduce errors

Position and coordinates

The distance and direction between any two waypoints, or a position and a waypoint

Travel progress reports

Accurate time measurement

The Defense Department dithered the satellite time message, reducing position accuracy to some GPS users.

S/A was designed to prevent America's enemies from using GPS against us and our allies.

In May 2000 the Pentagon reduced S/A to zero meters error.

S/A could be reactivated at any time by the Pentagon.

Sources of GPS Error...

-satellite clocks

-orbital errors

-ionosphere

-troposphere

-receiver noise

-multipath

-selective availability (no error.. for now)

User error (up to km or more)

** REceiver Errors are Cumulative!!!

System/other flaws= <9m

User error = +/-1km!

Sources of interference:

-atmosphere

-solid structures

-metal

-electromagnetic fields

Position Fix

A position is based on real-time satellite tracking

It is defined by a set of coordinates

It has no name

A position represents only an approximation of the receiver's true location

A position is not static. It changes constantly as the GPS receiver moves (or wasnders due to random errors)

A receiver must be in 2D or 3D mode (at least 3 or 4 satellites acquired) in order to provide a position fix.

3D mode dramatically improves accuracy

A waypoint is based on coordinates entered into a GPS receiver's memory.

It can be either a saved position fix, or user entered coordinates.

It can be created for any remote point on earth. It must have a receiver designated code or number, or a user supplied name.

Once entered and saved, a waypoint remains unchanged in the receiver's memory until edited or deleted.

PLANNING A NAVIGATION ROUTE

-receiver sees your route as a series of points

There is a GPS waypoint circle of error.

Satellite geometry can affect the quality of GPS signals and accuracy of receiver trilateration

Dilution of Precision (DOP) reflects each satellite's position related to the other satellites being accessed by a receiver.

There are five distinct kinds of DOP

PositionDOP or PDOP is the DOP value used most commonly in GPS to determine the quality of a receiver's position.

It's usually up to the GPS receiver to pick satellites which provide the best position triangulation.

Some GPS receivers allow DOP to be manipulated by the user.

Ideal geometry = satellites 33degrees in every direction, and one straight above

Good Geometry

-4 equally spaced..

-scanning zones intersect at perpendicular angles

Poor Geometry

- all right above you

-all to one side

-lots of intersecting

dark overlap area = zone of potential (when @ right angles, zone is minimized)

Wide Area Augmentation Service

-Created by the Federal Aviation Administration (FAA) to improve GPS performance (launched satellites to improve satellites)

-WAAS enables aircraft to rely on GPS for flight including approaches.

-Network of ground stations that measure GPS variations

-Transmitted every 5 seconds or less to the WAAS satellites (Geostationary) - sat moving same speed as earth's rotation

-Rebroadcast corrections to Earth

WAAS coverage is essentially all of North America

How Good is WAAS?

Differential GIS = slight time delay

-With Selective Availability set to zero, and under ideal conditions, a GPS receiver without WAAS can achieve fifteen meter accuracy most of the time*

-Under ideal conditions a WAAS equipped GPS receiver can achieve three meter accuracy 95% of the time*

*Precision depends on good satellite geometry, open sky view, and no user induced errors.

Aircraft

-most modern technology is air based (get the equipment back)

-Low, medium and high altitude (closer you are, better you see)

-Higher level of spatial detail

-much slower than GPS

-trying to generate an image

Satellite

-Polar orbiting, sun-synchronous

-slight inclination (8-12 degrees)

-nothing at the poles!

-700-900km altitude, ~100min/orbit

Geosynchronous (36km alt, 24hr orbit... same as earth)

-stationary rel. to earth

-weather satellites

What are the primary benefits of satellite remote sensing?

-View from a long way away (much harder to "shoot down")

-Most of these systems operate at ranges of about 500km

-Repeat cycle

-Large area covered quickly

-Synoptic view

-Automatic image gathering (no film)

Polar orbits allow for lower altitudes and more image resolution (geostationary far away)..

-These satellites take multiple passes of the earth before returning to the same locations (download data @ night)

SUMMARY

-Geosynchronous Orbit (GEO) 36kkm above earth. Includes commercial and military communications satellites, satellites providing early warning of ballistic missile launch.

-Medium Earth Orbit (MEO): includes navigation satellites (GPS, Galileo)

-Low Earth Orbit (LEO): from 80-2000km above earth.. includes military intelligence satellites, weather satellites, Earth observation satellites.

STYLES OF IMAGING SYSTEMS

**study this slide (lecture24, slide 24)

-first is most complicated..

-the others are all just more advanced versions of the same thing.

Digital Imaging is vital to our understanding of the global environment system.

Imaging provides the key information for decision makers in government ad industry.

Basics of image understanding come from out roots in photographic interpretation adn measurement.

Digital imaging systems do more than produce "pictures" they are science-grade instruments that are measuring physical quantities in space.

(picture not the same as image... one art, one science)

Resolution implies that we can measure something in a finite manner.

Resolutions in remote sensing involve 4 fundamental measureable quantities:

-spatial resolution (division of ground being imaged)

-spectral resolution (division of spectrum) (multi/hyper/ultra)

-temporal resolution (Revisit time)

-radiometric Resolution (ability to sense energy)

These are the basic quantities.

All have a physical basis.

FROM COLOUR TO SPECTRAL RESPONSE

Helps to understand colours and spectral responses

Discrete bands are recorded

Usually done with broad spectral bands (multi)

Physics version of colour (more complicated biologically)

Main operating principle for all multispectral canners.

*On a Reflectance/Wavelength graph, the peaks are where energy is coming to us (if it isn't it is being absorbed).

-plants don't like infrared, so you'll see a spike in this region.

**look at dif pictures and their colors (Nov 8.. also have notes)

Images come from the sensor as a digital recording of the photons received at the sensor.

Digital images don't have colours.

-Colours are made... they don't exist on an iage per se.

-colour is used as an interpretive tool, not just pretty pictures.

This is what we call spatial resolution.

Relates to the number of pixels/image area.

Combined with the concept of an instruments IFOV

Not necessarily the same distance as IFOV

-sometimes the IFOV is arger than the recorded ground resolution

Not the same measure as for photographic film

Refers to the spatial dimension of each pixel

NOT THE SIZE OF THE SMALLEST OBJECT THAT CAN BE RESOLVED

-that on square is one colour and can't be resolved

-need multiple pixels to see something

RESOLUTION VS DETECTION

-can't do anything about this problem

-can see tennis court lines at 61cm resolution?

Instantaneous Field of View

Cone angle within which incident energy is focused on the detector.

Determined by the optical system and the size of the detector.

All energy focused at the sensor is recorded within the IFOV

Mixed and pure pixels

-mixed if object is smaller than IFOV

-pure if object is larger than IFOV

TM = thematic mapper

Syn synchronous orbit

705 vs 900 km (improve resolution and for potential recovery)

Cross equator at 9:45 am; 99min orbit; 14.5 orb/day

14.92 FOV (> than MSS)

30m ground resolution

-thermal is 120m

Bi-directional scanner.

-7 bands (6 visible/IR, 1 thermal)

A/D converter (produces 8-bit data; 256 shades of grey)

Currently on Landsat 7 (EATM. enhanced thematic mapper)

-has blue-green band (very noisy)

-troposphere gets most of this radiation before it gets back to the satellite

GEOMETRIC CHARACTERISTICS

Tangential-Scale Distortion

Severe distortion of across-track imagery perpendicular to flight direction

Irregular speed of IFOV across the landscape

-mirror rotates at a constant velocity

-creates a different velocity of the IFOV

-covers more distance per unit time further from the nadir

-pure geometric correction (we know the parameters)

Band 1 (blue-green)

-noisy due to scatter

=penetrates water - the band of choice for aquatic ecosystems

-used to monitor sediment in water, mpaping coral reefs, and water depth

-rarely used for "pretty picture" type of images

BAND 2 (green)

-similar qualities to band 1, but not as noisy

-green wavelength for the vegetation monitoring

BAND 3 (red)

-chlorophyll absorption band

-useful for distinguishing b/w vegetation and soil

-monitoring vegetation health

BAND 4 (near IR)

0measures the peak reflectance in vegetation

-corresponds to veg vigor

-water absorbs all IR - easy to map wate rboundaries

-less affected by atmospheric attenuation - longer wavelengths do not interact with atmosphere.

BAND 5 (mid IR)

-monitors the dip in the spectral response curve associated with vegetation moisture (ABSORPTION feature - darker = more wa ter)

-used to monitor vegetation water stress and soil moisture

-useful to differentiate b/w clouds and snow

BAND 6b (thermal IR)

-measures furface emittance, not reflectance

-not used for much

-geological applications

-differentiate clouds from birght soils since clouds tend to be very cold

-the resolution is twice as coarse as the other bands (60m instead of 30m) - Longer wavelengths are less energetic.

BAND 7 (mid IR)

-also used for vegetation moisture, although band 5 preferred.

-commonly used in geology - mineral exploration

-good at detecting burn scars

-can detect high surface temperatures (Chernobyl)

SPOT

Systeme pour L'Observation de la Terre

Commercial orientation (not experimental)

Began in 1986.. 20m multispectral (IR, red, green)

-10m panchromatic (HRV)

*Viewing area i 1/3 the size and 3 times resolution (60km viewing bands)

**10m is 4x better than 20m

200 employees.. worldwide ground station network

8 distributors worldwide.

No stations in US...

-at SPOT 5 now (@ least 3-4 still functioning)

IUNCTUS GEOMATICS

Canadian/US Rights to distribute SPOT data since 2002

Ground station on UofL campus

Infrastructure to process SPOT data in North America

Focus on commercial markets via Telus Alliance

Academic markets: sublicensed to ATIC (alberta terrestrial imaging centre)

ATIC's academic prices for Canadian imagery substantially below commercial prices

Spatial Resolution: imagery detail

Spectral Resolution: type of data

Temporal Resolution: frequency of observation

Financial Perspective: price/sq km

SPOT 5 = launched in May 2002

-2.5m resolution = panchromatic B&W

-5m resolution = panchromatic B&W

10m resolution = multi-spectral (colour)

60kmx60km scenes or 3600 sqkm/scene

+/- 30 degree look angle

Three dual camera satellites

Data storage ranges from 40Mb to 5050 Mb/scene

60km swath for nadir

80km for oblique

832km altitude (sun synchronous)

-return 26 days for nadir

**Off nadir viewing creates image parallax..

-allows for the creation of digital elevation models

Focal length is huge (have to fold up light)

-can look to left an dright..

-first time using stereoimaging from space.

Resolution Cell Variation

-IFOV gets larger in an off nadir direction

-Different dwell times per unit area at the nadir and end of the rotation angle

One Dimensional Relief Displacement

-Relief displacement occurs only perpendicular to the flight line

-Photographs have radial displacement from a point... scanners produce a line.

(Relief Displacement)

-know dif b/w satellite and single image

-sat scane along a line... get directly above for line, and displacement above/below

*ALL imaging systems!

Digital imaging represents a systematic sampling of the area imaged with a fixed number of pixels.

Two types of sensors:

-CCD (Charge-Coupled Device)

-CMOS (Complementary Metal Oxide Semiconductor)

Borth produce an electric charge when photons of light strike the surface.

GENERATING AN IMAGE WITH A CCD

•A CCD is an array of silicon atoms whose bonds
can be broken by absorbing light of various
wavelengths.
•Electrons are released but maintained in a region
of the chip called a potential well.
•Charge collected in the well is read out.
•An analog to digital converter converts the charge
to a digital number (DN).

***It is analogue.. we don't really have digital!

-Do an analogue to digital conversion

CCD Operation... rectangles!! (ever seen a rectangular pixel?)

IMAGE FORMATS

Multispectral imagery is composed of a number of discrete bands.

-3 dfferent ways of organizing these data...

Band Sequential (BSQ)... digital cameras

-3x3 of a colour, 3 times

Band Interleaved by Line (BIL)... SPOT

-1x3, as above

Band Interleaved by Pixel (BIP)... Landsat

-1x3, each the same, each pixel a different colour

to see diagram... lecture 27, slide 19

Imaging that takes place outside of hte photographic spectrum (0.3-1.1 micrometers)

Usually in the Near Infrared (1.1-3 micrometers) or the thermal infrared (8-14)

Passive sensors

Great deal of interest in microwave wavelengths (0.8-100cm) RADAR

dark world concept: shoot energy out. Only reason we know where it is, is that there is nothing else there

Acronym for RAdio Detection And Ranging (although we rarely ue radio frequencies (1-10m))

Active imaging system.. transmits short pulses or bursts of microwave radiation and receives the reflection from the target (detection)

Can tell where the target is by timing the pulse (ranging)

Reflections from the target are called echo or backscatter.

Active system. Day-Night imaging

Sends out single wavelengths of EMR

Records the strength of the return

-good reflectors are bright

-poor reflectors are dark

3 Types of Reflector

1)Specular (smooth ice) (look dark)

2)Diffuse Reflectors (Vegetation)

3)Corner reflectors (building

**All are wavelength/feature dependent

*the further away the pulse ribbon gets, the wider it gets

• The pulseof electromagnetic radiation sent out by the
transmitter through the antenna is of a specific wavelength
and duration (i.e., it has a pulse lengthmeasured in
microseconds, sec).
• The wavelengths are much longer than visible, near-
infrared, mid-infrared, or thermal infrared energy used in
other remote sensing systems. Therefore, microwave energy
is usually measured in centimetersrather than micrometers.
• The unusual names associated with the radar wavelengths
(e.g., K, Ka, Ku, X, C, S, L,and P)  are an artifact of the
original secret work on radar remote sensing when it was
customary to use the alphabetic descriptor instead of the
actual wavelength or frequency.

USES AND BENEFITS

•All weather imaging
•penetrates clouds
•no atmospheric scattering
•Day night imaging (twice the imaging frequency of
optical satellites).
•Ice detection.
•Flood mapping.
•Vegetation differences.

•Canada’s Radar Satellite.
•Run by RSI in Richmond.
•C band (5.6cm wavelength).
•7 different beam modes (8-100m resolution).
•6 different area coverages(50km -500km).
•3 different incident angles.

RADARSAT 1

• In operation since 1996 –confirmed
continued  operations until 2009
• Demonstrated potential for routine polar
observation (world-wide ice services, Arctic
and Antarctic missions)
• Archive collected over both polar regions
(>250,000 images over the Arctic)

RADARSAT2

-enhanced modes and capabilities

-multi-polarization, polarimetry

-ultra-fine resolution

-left and right looking

Canadian Government data allocation available for International Science Projects

WHAT IS THE RATIONALE BEHIND NEEDS ASSESSMENT?

What is needs assessment?
•Process of identifying what is expected from the project
•Can be thought of as a blueprint (i.e. planning)

Why needs assessment?
•Cost of not planning?
•Expectation management
•Helps you identify potential problems

Why do we have to know about organization
needs?
•Is need-to-know question in alignment with goals of
organization?

WHAT ARE THE BENEFITS OF GIS PROJECTS?

•Save money/cost avoidance
•Save time, thus increase efficiency
•Increase accuracy
•Increase productivity, same time more product
•Increase communication and collaboration
•Generate revenue
•Support decision making
•Automate workflow
•Build an information base
•Manage resources
•Improve access to government…

The level of benefits

-Large-scope GIS projects more likely to reduce cost since it will make the use of shared database (e.g. centralized database) and benefit from automation.

The roles of GIS in an organization

-Large-scope GIS is more likely to change the way the organization does business

Strategies for needs assessment

-The larger the scope of GIS projects, the more it is necessary to find out the organization and business work flow.

UNDERSTANDING BUSINESS WORK

Models of complex business process

Thought of as the way an organization does business

-for example, land-use developent approvals, building loan approvals, permitting, conservation land planning, service delivery, and distributed facilities management.

Information products come out of that work flow.

GIS functionalities can be utilized within the context of business work flow, and also have a potential to change/redesign business work flow.

Application

-two or more information products linked by a business workflow model

Examples of applications:

-forest practices permitting

-warehouse management and distribution optimizing

-land records managemtn

Information products go into Business workflow model, use GIS to extract data..

To find out business workflow, ask "what do you do?"

To find out information product, ask "what do you need for a business function?"

WHAT IS A HIERARCHICAL VIEW OF NEEDS ASESSMENT?

Goal>Function>Facility>Entity>Attribute

In the real world, start form outside.

Goal: One of the major strategic directions for the organization

-maintain the urban infrastructure

-reduce crime

-create jobs for citizens

Function: a major activity within the organization which supports a goal of the organization.

-repair city streets

-enforce building codes

-assess the value of property

Facility: the physical, legal, or other asset upon which a function of the organization is performed (ie. database: a collection of related entities)

-street, building, parcel

Entity: Components of the facilities that are managed by the various responsibilities of the organization (ie. feature class in ESRI speak)

-pavement, curbs, cutters

Attributes: descriptive data that define the characteristics of the entity

-location, condition, size, date, type

How do we identify information products?

What are outcomes needed from a GIS?

What are answers to need-to-know questions?

Core of information product is an information
structure; how are chunks of information
arranged?
•List of something xxxxx (text list requirements); perhaps
a ranked list
•Narrative of something xxxx (text requirements)
•Mapping variables for xxxxx (map requirements)
•Map showing changes in xxxxx (map requirements)
•Diagram that shows the potential database design
(schematic requirements)

For each information product, input data has processing needs before it can become an information product. Which GIS functionalities are needed for each step?

Data acquisition

Mapping

Pre-processing

Data Structure

Database

Spatial analysis

Modeling

Display

Application

•Finding a subset of  locations from a set
of potential or candidate locations that
best serve some existing demand so as
minimize some cost.
•Locatesites to best serve allocated
demand.
•Application areas are warehouse
location, fast food locations, fire stations,
schools.

INPUTS

•Customer or demand locations.
•Potential site locations and/or existing facilities.
•Street network or Euclidean distance.
•The problem to solve.

OUTPUTS

•The best sites.
•The optimal allocation of demand locations to
those sites.
•Lots of statistical and summary information about
that particular allocation.

Spatial Data and Spatial Analysis..

-Imagine you are a national retailer

-You need warehousees to supply your outlets

-You do not wish the warehouses to be more than 1000km from any outlet.

*look at ideal sites

*look at feasible sites

* find optimal

Vancouver Maps

•“Enterprise GIS” -many layers (property, crime,
census, schools & parks, capital improvements,
development)
•OrthophotographyPhotography Included
•Zoom capabilities
•Capacity to Look at Databases
•http://vancouver.ca/vanmap/g/gis.htm

Look through these slides!!

THE INTERNET REVOLUTION

E-Government

-Activities that focus primarily on providing information and trans-active type services to customers of government.

E-Governance

-Activities that focus on the public in its role of citizen and include such attributes as on-line dialoging nad oplling among others all designed to make government more accessible and transparent.

DESIGN & EVALUATION: E-PARTICIPATION

“E-Participation” Language.
•Right to Know, Informing, Restricted Participation,
Public Participation in Defining Agenda, Public
Participation in Analysis, Public Participation in
Decisions.
•Online Service Delivery, Communication Barrier, Online
Discussion, Online Opinion Surveys, Online Decision
Support Systems.

Public Participatory GIS

Participatory GIS is an emergent practice in its own right' developing out of participatory approaches to planning and spatial information and communication management.

WHAT IS PPGIS?

•Participatory GIS implies making GIS available to
disadvantaged groups in society in order to
enhance their capacity in generating, managing,
analysingand communicating spatial information.
•PGIS practice is geared towards community
empowerment through measured, demand-driven,
user-friendly and integrated applications of geo-
spatial technologies.
•GIS-based maps and spatial analysis become
major conduits in the process.
•PGIS is part of the spatial decision-making
processes.

It is flexible.

Different socio-cultural and bio-physical environments,

Depends on multidisciplinary facilitiation and skills

Inclusionary vision of Urban GIS.

•Builds essentially on visual language.
•Expert skills with socially differentiated local
knowledge.
•Promotes interactive participation of stakeholders
in generating and managing spatial information.
•Uses GIS to facilitate decision making processes
that support effective communication and
community advocacy.

(UofL doesn't even have software for this)

•3D urban modelling is the geometric and
graphical object reconstruction the
physical structure of a city.
•Includes all buildings and the
surrounding environment (terrain,
streets, vegetation etc.).
•Makes use of RS/GIS Technology.

•Use of LiDAR and Object Modelling.
•Why 3D urban modelling?
•Photo-textured and three-dimensional
models enable easy understanding.
•It is relatively easy to layer abstract
phenomena over a detailed model.

Visualization in an Urban Environment

•User would be able to recognize specific
elements, spatial position, scale and to
relate plan details and other information
within the area under investigation.
•The computational power of this
technology to transform and instantly
compare alternative representations
provides decision-makers with
unprecedented flexibility .

3D URBAN MODELLING

•Last 10 years has seen a transformation
in 3D visualisatioin tools.
•LiDAR data are captured with image
data (sometimes hyperspectral).
•City planning processes not involve
dialogue sessions.

•Synthesize proposed changes to a city
with a systematic investigation of the
visual implications of a design.
•Building design, traffic flow, air flow,
visual impact, sunlight.
•Other uses...growing daily.
•Seen as a critical development in e-City
concept.

3 Categories:

-Remotely Captured (col/laser/sat/aerial)

-Close range (geodetic/photogrammetry/laser scanner/texture images)

-Derived Source (DSM, 2D Maps/GIS)

LiDAR (laster scanning systems)

•Laser light emitted –recorded.
•Same principle as RADAR.
•Thousands of measurements of height vs. traditional
methods.
•Crude LiDAR has been around for decades.
•Laser DSMs (digital surface models) provide very reliable
information and accurate surface information.
•Costly to produce –still requires manual manipulation

2D GIS

•Currently used in operational urban management
systems.
•Long history of use and recent integration of data
(different sectors of management).
•Powerful analytical capabilities.
•Restricted use.
•Beginning of web accessible GIS.
•Integration with Geospatial technologies.

Technology Convergence

•Realistic 3D visualisation.
•Animation technology.
•Texture mapping (realistic visual appearance,
false impression of higher level of geometric
detail.)
•Transform aerial data and ground collected data
into photo-realistic representations of current
architecture.
•Continue to develop means of incorporating
information into new developments

Interactive/Manual

•Terrestrial or aerial photographic data on virtual
geometry can be very expensive and time
consuming.
•Major advantage = control over the outcome
despite the time and cost limitations.
•Usually consists of a GUI allowing human
operator to map photographic data on
corresponding virtual geometry.

Automated

•Aerial imagery, and panoramic image acquisition.
•Registering aerial images to the virtual geometry
from different angles including oblique
perspectives.
•Ground based acquisition systems that can scan
urban building façade as seen from the street level
(usually with a camera mounted on a vehicle)

• Consist only of façade graphics with no thematic
features or connection to external information for
buildings (example; year building was created, owner
information etc.)
• This is also directly linked with level of detail (LoD)
modelling (again lacking in most current systems).
• Example: first LoD-> building, LoD2 -> bounding
surfaces are differentiated semantically: wall roof
ground surfaces, LoD3 -> openings: doors, windows,
LoD4->rooms: interior doors, interior walls / ceilings
etc.

• Urban areas are becoming more complex and are more
difficult to model sufficiently by the techniques and
methods described.
• Gap between research results and production tools, for
example using a single semiautomatic or automatic
approach to reconstruct both ac athedral and a simple
gabled roof is not effective

• Most systems fail to take into account that urban areas
are evolving and change all the time.
• Finally, there are many3D city models available but
there is the distinctive lack o fa unified standard evident
(many different systems, formats, approaches, different
representations of geometry, difficult to integrate 3D city models etc.)

2 main types are LiDAR (light detection and ranging) and IFSAR (Interferometric Synthetic Aperture Radar)

-much higher, onboard processing, no limitations

LiDar has multiple laser returns... can as for

first' and 'last' returns (look through trees)

Extraction of features can be automated by measuring shape

LiDAR Analhyst, developed by Dr. Vincent Tao at York Univeristy

Bought by US Company

Can extract bare Earth, buildings, generate contours, vegetation extraction and terrain characteristics

Simplifies LiDAR data extraction (key limiting factor)

Interoperable with other GIS

AUTOMATED FEATURE EXTRACTION (RS)

Manual interpretation is costly and time consuming.

Would like to go from orthophoto or satellite image to GIS layers in an automated system.

RS DATA EXTRACTION

-Acquire image

-Mosaic images- orthophoto production

---geo-corrected data- projected and rectified

-Image classification

-Determine spatial objects

-Deal with trees etc.

-Semi-automated process

IMAGE PROCESSING

Image processing involves the processing and manipulations of aerial photography and satellite imagery to suit the needs of GIS and information extraction

To accurately derive geospatial information from image data, it is critical that imagery be properly registered, enhanced, and prepared.

Processing Tasks

-Format conversion

-Geo-registration

-Orthorectification

-Image Reprojection

-Image Enhancement

-Image Mosaics and Clipping

-Stereo Model Creation

•Higher-resolution imagery is used to extract
geospatial information. 
•Simple image classification.
•Definition of types of surface targets.
•Makes use of spectral signature information.
•Often use image texture and spatial context
classifiers assist in the  conversion of these
features into a GIS environment.

•Man-made features.
•Natural features.
•Land-cover.
•Change-Detection.

1)GIS-ready Image Data

2)Feature Extraction

3) GIS layers

•Visualization is a powerful rendering capability to
interpret geospatial relationships betweenobjects
and features extracted from imagery.
•Multiple layers of data can be viewed at once.
•Desktop visualization can make time in the field
more efficient.

•Perspective views
•Fly-through
•Scenario Modeling
•Veiwshed Analysis
•Disaster Planning

•http://www.youtube.com/watch?v=GXS0bsR0e7w

Very testable

In 15 years, want everyone in Portland living 20 mins from everything they need.

UPS spends 1billion/year on technology

SYSTEM CONSIDERATIONS

Central processing unit (CPU).
•Serial or parallel processing.

Operating systems.
•Unix, windows NT, 98, XP.

Ram.

Storage.

Rapid access on hard disk.
•8 bit data, 16 bit data, 32 bit real.
•Optical disks, CD, DVD, exabyte tape.

Spatial and colour resolution.
•1024 x 1024, 1248 x 1024; 6000 x 6000.
•Grey scale / colour depth more demanding for image
display than graphics.
•24-bit colour displays (millions of colours).

Image processing applications software.

Peripheral devices (image output).

IMAGE STATISTICS

Univariate Statistics

-Measure of central tendency

-No information b/w pixel values in different bands

---mean

---variance

---standard deviation

Multivariate Statistics

-Insight into data quality and redundancy

---Are the data suitable for the planned task?

---Do the pixel values in different bands vary independently?

-Provide information for PCA, feature selection and image classification.

---covariance

---correlation

Mean

Variance (squared difference from mean, divided by population size)

Standard Deviation

Covariance

-A measure of mutual interaction.. the joint variation of two variables about their common mean.

Correlation

-the ratio of the covariance of two variables to the product of their standard deviations

-estimates the degree of interaction b/w two variables.

Mode

•most frequently occurring value in a distribution of
scores
•modal class (classified data, nominal/ordinal)
•Remotely sensed data often have more than one mode.
•There are several relatively bright areas on an image,
that are not vary variable.
•Used for post-classification clean-up.

MEDIAN

•the value (X) with an equal number of ranked scores
above and below (ie 50th percentile, or P50)
•Count the number of observations, divide by two.
•Rank observations...and select the middle number (50th
percentile).
•Used for processing imagery and removal of localized
errors.

Range (not central tendency)

•Gives a crude idea of the amount of dispersion (total).
•Can also be used in a class by class fashion.
•Plays an important role for image classification.

The ability to detect differences in tone between
similar features.
•Low contrast images have a small range of tones (little
difference).
•High contrast image have a large range of tones (large
differences).

Low contrast is related to targets having similar
amounts of radiant flux within a portion of the EM
spectrum.
•Vegetated/water scene will have low contrast in the blue,
high contrast in the IR and low contrast in the thermal.

Also related to the sensitivity of the detectors.

Satellites are designed to record a wide range of
scene brightness values without becoming
saturated.
•Must handle imaging conditions from ice caps to jungles
in a single pass.

Atmospheric effects also contribute to poor image
contrast.

•Can go from image to histogram for analysis of
data.
•Can look at the frequency and the variability of
information.
•First step in statistical analysis.

Positive skew is a high frequency of lower number

Neg skew is a high frequency of high number

Typical image histogram is frequency of brightness value

Thresholding

-Simple image segmentation

-If DN > value then = 0 (ie. 85)

If DN< value = 255

-Binary Mask

Level Slicing

-DN values are divided into a number of intervals (Bins)

-0-255 split into categories

-Single Band classification

Linear Contrast Enhancement

-Used when all the brightness values fall generally within a single narrow range

-low contrast images

-"stretching"

-determine the min and max ofthe distribution

-compute a simple linear transform

-Some waste as there may be erroneously high pixel values

-Saturation = changes the min and max level

-Conditional statements do the saturation

---series of it statements (if less than min, then min. If greater than max, then max).

Eyes have cones and rods.

cones= colour

rods= seeing things in dark (much more of these @ higher density)

We have lots of green cones (more adapted to seeing this color)

Our eyes don't see a really nice red... our brain has to make the adjustment.

In a digital system, you get a complete band, while cones all overlap broadly.

The simultaneous red+blue response causes us to perceive a continuous range of hues on a circle. No hue is greater than or less than any other hue.

***STudy the graphs of receptor spectral sensitivity*** (lecture 34, page 5)

The eye has  3 types of photoreceptors: sensitive to red, green, or blue light.

The brain transforms RGB into separate brightness and color channels (e.g. Luminance, hue, saturation)

In a digital image, point=pixel.

Point processing transforms a pixel's value as function of its value alone.

It does not depend on the values of the pixels neighbours.

Brightness and contrast adjustment

gamm correction

Histogram equalization

histogram matching

colour correction

Histograms are calculated based on the
frequency of a greyscale.

Most images have 0-255 (8-bit) greylevels.

Controlled by the AD converter from the
imaging sensor.

•Take the analog voltage recorded in the
potential well and converts it to a digital signal.

Frequencies are calculated by counting the
number of pixels in each greylevel.
•Also known as bightnessvalues (BV’s).

8-bit = radiometric scale

AD= analogue to digital

•Used in remote sensing to ensure that the range of
satellite images are similar between two images.

•Images could be from the same sensor on different
dates or from different sensors (same or different
dates).

•Elementary calibration of images.

•Produces a NORMALIZED image.

•Trying to produce a physical quantity rather than
a pretty picture.

MATCHING USING LUT

•LUT = Look Up Table.
• One image histogram matches another directly.
• Often faster or more versatile to use a lookup table (LUT).
• Don’t have to remap each pixel in the image separately,
one can create a table that indicates to which target value
each input value should be mapped in the output image. 
• Create LUT’s to match image types once a consistent
match is found.
• Often based on comparisons of Invariant Targets (through
time and over space).
• ---Deserts/snow etc.
• ---Related to physical quantities.

What you see isn't what you get.

Need to correct for the use of non-linear systems used in interpretation of images

-you visual system- CRT's

If a digital system was used to record image data - the relationship between brightness values is LINEAR

Generally results in darker areas being harder to see/interpret.

Loss of interpretable data at lower/higher ends of spectrum.

WHY GAMMA?

Relates to the slope of the characteristic curve from photographic film.

Make maximum use of the dynamic range.

Changes image contrast characteristics (sensitivity to light)

Gamma = 1 = regularish
Gamma = 1.6 brighter

Gamma = 0.8 = darker

A chromaticity diagram has a fixed brightness for all colours.

Colours associated with a single wavelength are on the curved rim but nonwavelength colours like magenta are on the flat part of the rim.

-insides are the less saturated colors, including white at the interior.

ADDITIVE

RGB are the primaries.

-additive mixing is done by mixing primary colour lights with dif intensities

-G+R = Y

-G+B = cyan

-R+B=magenta

-middle venn diagram = white

complementary colours= when two complementary colours mix, they produce white light

-B+Y/G+M/R+C

SUBTRACTIVE COLOUR MIXING

The subtractive primaries = Cyan, Yellow, Magenta

-white light passed through a cyan filter plus a magenta filter appears blue

-Magenta passes blue and red - cyan passes blue and green. -colour in common is blue

Unlike point-to-point of per pixel processing, neighbourhood processing considers spatial context.

Areas vs poitns (pixels)... think spatial

Based on odd-numbered moving widow processing procedures (convolution)

AKA filter kernel

More or less the way we like to think of images - we are better at spatial relationships than spectral ones.

General image processing procedure

When this procedure is used the image is being convolved.

An odd-numbered (3,5,7,etc) window is  positioned over the data and a series of calculations or weights are applied to the area.

Value for the area is returned to the location of the centre of the moving window (kernel)

A new image is created as the window moves.

DETAILS

Two different types of windows

Neighbourhood operators - collection of values that are multiploed and then summed.

-weighted values

-computationally efficient

Neighbourhood functions - mathematical or statistical computation conducted for the winder

-may not be efficient... not everything can be made into an operator

PROBLEMS

What to do with the edges..must use odd sized windows.

For larger window sizes this can be significant.

3 approaches...

1)leave edges out... make images smaller

2)pad image (input or output) wit 0s to keep the output image the same size or keep the entire area (pad the input)

3)Copy the first processed pixels (rows and columns) into the "vacant" locations

Filter: the process of removing or emphasizing a feature.

Often used as a pre-processing procedure.

Emphasizing spatial features in the image based on Local spatial properties.

FILTERING BASICS

-filtering comes from Signal Processing (EE)

-Named for what is allowed to get through (pass)

-for us this means spatial frequency

-spatial frequency is the amount of change in DN per pixel.

-Low spatial frequencies are spatial frequencies that do not change frequently

-High spatial frequencies change rapidly per unit distance

LOW PASS FILTERING

-take every square, and multiply it by the fraction of how many pixels you are using in the grid (ie. 1/9 for 3x3)

-then go through and add groups of 9 together.

-get a more spread out distribution (ie. blurrier lines)

High Pass

-simply take the low-pass filtered image and subtract it from the original

-edge enhancement

(Nirvana awaits with image classification)

The primary goal of statistical classification is data reduction.

Assumes that we are producing information

Major area of RS research

Originaly conceived as a replacement for manual mapping.

Some techniques are almost completely automated, some require user intervention.

CONCEPTUAL UNDERPINNINGS

All classification is based on the likeness of groups

-many different ways to measure likeness

-squint

We select, or have the computer select, pixels that represent groups.

The computer takes these initial data and converts the rest of the image based on degrees of similarity between pixels in the sample and the image.

SIMILARITY

Based on the collection of a sample and the extension of that sample to the whole.

Uses the concept of spectral signature

If the "signature" of a feature is invariant, then we can find all similar features.

How to define similarity..

-closeness (pixel colour, not space)

-normal distribution

-connectedness (degree of likeness - decision tree)

SPECTRAL SIGNATURES

Limited # of Bands

Multispectral images record radiance/reflectance in a series of discrete spectral bands, rather than over a continuous range (hyperspectral)

Spectral response is represented by the discrete digital number (DN).

Wavelength is indicated by the band number.

Crude spectral signature curves can be simply constructd by plotting the image pixel value as the function of band number.

Higher contrast between signatures = easier to distinguish between them.

The greater the potential is for fast and accurate image interpretation and mapping.

Data reduction

-major goal of all classification is to reduce or geenralize the number ofcolours to a manageable number (ie. from 17 million to 8).

Used to extract general patterns (interpretation - extract some pixels... find out their average spectral properties and go find the rest of the pixels that look like the sample)

Only spectral - no spatial relations are considered.

Point process.

SIMPLE CLASSIFICATIONS

Level Slicing

-take an image with 256 shades of grey and reduce that to the most common.

-DN values are divided into a number of intervals (bins) (ie. 0-70,71-140,141-200,201-255)

-single band classification

Multi-band Level Slicing

-Parallel piped (parallelepiped)

-same logic as the level slice except that it involves two or more bands (dimensions)

-colour!