GEOG 465 Midterm

Aerial photography 1st platforms: 1858, 1860- Air Baloons, 1882 - Kites, 1903 – Pigeons

Aerial photography platforms today: 1908- Airplanes, 1961 – Satellites

Consists of an electrical field and a magnetic field
The two fields are orthogonal to each other
Direction of each field is perpendicular to the direction of wave propagation
EMR is transmitted in the form of photons (travel at the speed of light)

1. wave theory: waves travel in a sinusoidal fashion at the velocity of light
2. partical theory: waves are composed of mnay units called photons

At shorter wavelengths we can measure / sense smaller areas on the surface of the earth (higher measurable energy).
At longer wavelengths we must measure the energy from larger areas on the surface of the earth (lower measurable energy).

Ultraviolet Shortest wavelengths that are used in remote sensing applications
Visible radiation: violet - red

Blue, green, and red are the primary colours or wavelengths of the visible spectrum. They are defined as such because no single primary colour can be created from the other two, but all other colours can be formed by combining blue, green, and red in various proportion

Infrared radiation 2 parts:
1. Reflected IR (0.7- 3microm.)
2. Emitted or thermal IR (3- 100 microm.)

mostly the sun

sensor measures reflected solar radiation
reflected radiation from the sun and emitted radiation from objects
Passive because it uses natural sources (sun and moon)

Sensors emits radiation and measures the reflected radiation
emitted radiation from active sensors

Particles and gasses in the atmosphere affect EMR through scattering and absorption

The radiant flux reflected or emitted from the Earth’s surface once again enters the atmosphere, where it interacts with the various gases, water vapor, and particulates.

Thus, the atmospheric scattering, absorption, reflection, and refraction (or transmission) influence the radiant flux once again before the energy is recorded by the remote sensing system

Scattering occurs when the particles or gasses cause the EMR to be redirected
Amount of scattering depends on:
wavelength of the radiation
amount of gas/particles
distance the EMR travels through the atmosphere

Scattering decreases the quality of an image
Rayleigh: Specks of dust, nitrogen gas(N2 ) and oxygen gas (O2 ) molecules
Dominates under most atmospheric conditions
Strongly wavelength dependent
Blue sky and red-orange sunset

Mie: Dust, pollen, smoke, water droplets
Dominates in overcast conditions
produces general haze

Non-selective: Larger water droplets, large particles of airborne dust
Wavelength independent
causes fog and clouds to appear white to our eyes

Law of Conservation of Energy: Incoming Radiation = Absorption + Reflection + Transmission

The law of conservation of energy states that the total energy is constant in any process. Energy may change in form or be transferred from one system to another, but the total remains the same.

Radiation that is not absorbed or scattered in the atmosphere can reach and interact with the Earth's surface. There are three (3) forms of interaction that can take place when energy strikes, or is incident (I) upon the surface:
Absorption (A), Transmission (T), Reflection ®

In remote sensing, we are most interested in measuring the radiation reflected from targets. We refer to two types of reflection, which represent the two extreme ends of the way in which energy is reflected from a target: specular reflection and diffuse reflection.

Regions in the EMR spectrum where energy can be fully transmitted; most effective wavelengths for remote sensing

When the energy finally reaches the remote sensor, the radiance will interact with either the camera filter, the optical glass lens, and the film emulsion or optical-mechanical detector which record the number of photons in very specific wavelength regions reaching the sensor.
Ideally, the radiant recorded by a remote sensor is the amount of radiance leaving the terrain at a specific solid angle. Unfortunately, other radiant energy from various other paths may also enter the sensor’s instantaneous field of view (IFOV) or solid angle. This will introduce noise.
Various paths and factors for the noise are summarized from path 1 to path 5.
Path 1: solar irradiance (E 0) and atmospheric transmittance (T)
Path 2: diffuse sky irradiance (Ed)
Path 3: after some scattering, absorption, and/or reemission
Path 4: radiance from nearby terrain
Path 5: reflect or scatter from nearby terrain

By measuring the energy that is reflected or emitted by targets over a range of wavelengths we get a spectral response or spectral signature

By comparing the spectral response (i.e. spectral signature) of various targets in specific wavelengths we can distinguish them

Spectral signature
Reflectance: Reflected radiance in one direction is divided by incoming radiant flux density ( unitless)
Knowing in which part of the EMR to look and understanding the factors which influence a target’s spectral response is key to interpretation

Def: All vehicles, whether in space, airborne, or ground-based, that are capable of carrying remote sensors.

Sensors: Instrument that does the remote sensing - E.g., a camera is a remote sensor that can be carried on a platforms

Terrestrial platforms: Mainly used for calibration, testing, and research.

Airborne Platforms:
Balloons can carry detectors at high altitudes (up to 40 km) at comparatively low costs; we however have generally little control on their trajectory.
Aircrafts can carry heavy instruments, but the territory covered per flight is limited and their use is expensive.

Spaceborne Platforms
Satellites: Spaceborne platform, generally in orbit around the earth. They are generally very expensive, but they offer global coverage

Apogee: point farthest from the earth
Perigee: point closest to the earth
Inclination: angle that the satellite track forms with respect to the equator at the ascending node (e.g. 90o means polar orbit)
Period: time to complete one orbit (e.g. period of geostationary orbit at 36,000 km equals the earth’s rotation)

Sun-synchronous
The satellite passes over the same area at roughly the same time each pass
700-900 km altitude, rotates at circa 81-82 degree angle to equator.
Because of the low orbit, only a portion of the Earth’s surface is covered in any single day (half of it at night), although high latitude regions are better covered than lower latitude ones.
Advantage:
Similar solar illumination;
High spatial resolution (low orbits);
Global coverage.
Disadvantage
Narrow coverage at the same instant;
Limited temporal resolution (exceptions: RapidEye)

Geostationary/Geosynchronous
Stays vertically above the same sport, rotates with earth
Ex: weather images
36,000 km above equator
Advantages
Makes repeated observations over a given area (constant view area);
High temporal resolution, real-time monitoring possible
Disadvantages
Coarse spatial resolution;
Unequal spatial resolution (degradation toward the poles

Types: Field of View (FOV), Instantaneous Field of View (IFOV).

Dwell time is the time required for the detector IFOV to sweep across a ground cell. The longer dwell time allows more energy to impinge on the detector, which creates a stronger signal.

Spatial resolution
Spectral resolution
Radiometric resolution
Temporal resolution

FOV – IFOV ➔ Windshield view – eye focus

Spatial resolution depends on camera’s
Instantaneous field-of-view (IFOV)
IFOV = angular cone of visibility of the sensor (A)
IFOV determines the area on the Earth's surface which is "seen" from a given altitude at one particular moment in time (B)

Choosing a spatial resolution involves trade-offs…
Suitability to the task being performed
The cost of acquiring, processing, and storing the imagery
Compatibility with other data sets with which the imagery may be used
Which features need to be detected in the imagery

Spectral resolution describes the ability of a sensor to define fine wavelength intervals
The finer the spectral resolution, the narrower the wavelength range for a particular channel or band

Sensor’s sensitivity to the magnitude of the electromagnetic energy
Sensor’s ability to discriminate very slight differences in (reflected or emitted) energy
The finer the radiometric resolution of a sensor, the more sensitive it is to detecting small differences in energy
uses bits

Temporal resolution is the revisit period, the period after which the satellite gets back to the exact same area (NOT the same as the Orbit period - the time the satellite takes to complete one orbit around the earth)
For example, Landsat needs 16 days, MODIS needs one day, NEXRAD (weather sensor) needs 6 minutes
Temporal coverage is the time period of sensor from starting to ending.

Orbital track
Near-polar, sun synchronous, 9 degree inclination
Altitude from 705-900 km (depending on mission)
Orbital period about 100 min
Successive orbits 2875 km apart at equator
Swath widths 185 km wide

MSS
MSS = Multi-Spectral Scanner
Scanning mirror with discrete detectors
Angular field of view ranges from 12-15o (depending on mission)

TM
TM = Thematic Mapper (added to Landsat 4)
Scanning mirror with linear arrays (also called “acrosstrack” or “whiskbroom”, Figure D)
Longer dwell time

A. Direct transmission to a Ground Receiving Station (GRS) (if it is in the line of sight of the satellite)

B . If this is not the case, the data can be recorded on board the satellite for transmission to a GRS at a later time.

C. Data can also be relayed to the GRS through the Tracking and Data Relay Satellite System(TDRSS), which consists of a series of communications satellites in geosynchronous orbit. The data are transmitted from one satellite to another until they reach the appropriate GRS.

1.Gatineau Satellite Station
2. Prince Albert Satellite Station
3.Inuvik Satellite Station Facility
The combined coverage circles for these Canadian ground stations enable the potential for reception of real-time or recorded data from satellites passing over almost any part of Canada's land mass, and much of the continental United States as well. Other ground stations have been set up around the world to capture data from a variety of satellites.
Data received at the GRS ➔ in a raw digital format

Targets in remote sensing images may be any feature or object which can be observed in an image

Targets may be defined in terms of the way they reflect or emit radiation. This radiation is measured and recorded by a sensor, and ultimately is depicted as an image product such as an air photo or a satellite image. The target must be distinguishable; it must contrast with other features around it in the image.

Recognizing targets is the key to interpretation and information extraction.
Observing the differences between targets and their backgrounds involves comparing different targets based on any, or all, of the visual elements of:
Tone, Shape, Size, Pattern, Texture, Shadow, Association

No sense of depth
Unfamiliar perspective: viewing objects from directly above also provides a very different perspective than what we are familiar with.
Visible and non-visible light can be mapped in imagery

used to digitally identify and classify pixels in the data.

The computer or algorithm automatically groups pixels with similar spectral characteristics (means, standard deviations, covariance matrices, correlation matrices, etc.) into unique clusters according to some statistically determined criteria. The analyst then re-labels and combines the spectral clusters into information classes.
Analyst defines number (n) of clusters (classes) within the data
Algorithm arbitrarily ‘seeds’ or locates that number of cluster centres in n - dimensional space.
Each pixel is assigned to a cluster whose arbitrary mean vector is closest.
Procedure is iterated until there is no significant change in the location of mean vectors b/w successive iterations.

Identify known areas (known through a combination of fieldwork, map analysis, and personal experience) as training sites; the spectral characteristics of these sites are used to train the classification algorithm for eventual land-cover mapping of the remainder of the image. Every pixel both within and outside the training sites is then evaluated and assigned to the class of which it has the highest likelihood of being a member
The analyst identifies representative samples

How it works
Minimum distance: each pixel is assigned to the class whose mean is closest to class centre (in n-dimensions)
Parallelepiped: Each pixel is assigned to the class whose range it falls in (overlap = double assignment, individual new, depends on software)
Maximum Likelihood: each pixel is assigned to the class for which it has highest probability. Max likelihood can be run with a 'null class' (where pixels have no assignment to any class)

Minimum Distance to means
The mean spectral value in each band for each class is determined
An unknown class (1) is defined by computing the distance between the pixels DN value and the mean of each class.
The unknown pixel is assigned to the closest class. i.e. 1 goes to corn
is mathematically simple and computationally efficient
Limits: It is insensitive to different degrees of variance in the spectral response data i.e., Class 2 is assigned to Sand (S) instead of Urban (U).

Parallelepiped
An unknown pixel is classified based on a decision region that is defined by the range (min-max) of DNs
A pixel inside region = known, outside = unknown
Very fast and computationally efficient

Maximum Likelihood
Assumes normally distributed training data
Defines ellipsoidal equiprobability contours
If you visualize contours in 3D, the highest probability is in the center of the distribution
Each pixel is assigned to the most likely class (highest probability value) or labeled ‘unknown’ if below an analyst threshold

Sample design
Sampling design attempts to minimize bias

Collection of validation data
Possible sources: aerial photo interpretation, ground truth with GPS, GIS layers

Compare
How to summarize and quantify?
Error matrices compare, on a category-by-category basis, the relationship between known reference data (ground truth) and the corresponding results of an automated classification

uantifying accuracy
Total accuracy = number of correct plots / total number of plots
Problem with total accuracy:
Summary value is an average- Does not reveal if error was evenly distributed between classes or if some classes were really bad and some really good
Therefore, include other forms:
User’s accuracy
Producer’s accuracy

Accuracy figures are often presented from the users perspective
If I select any forest pixel on the classified map, what is the probability that I'll be standing in a forest stand when I visit that pixel location in the field?
and from the producers perspective

If I know that a particular area is forest what is the probability that the digital map will correctly identify that pixel as forest?

From the perspective of the user of the classified map, how accurate is the map?
For a given class, how many of the pixels on the map are actually what they say they are?
Calculated as: Number correctly identified in a given map class / Number claimed to be in that map class

From the perspective of the maker of the classified map, how accurate is the map?
For a given class in reference plots, how many of the pixels on the map are labeled correctly?
Calculated as: Number correctly identified in ref. plots of a given class / Number actually in that reference class

K = observed accuracy - chance agreement / (1 - chance agreement)
Interpretation of kappa
Closer to 1 → higher accuracy

Overall accuracy = (# pixels correctly classified) / (total # of pixels) = (90) / 100 = 90%
Omission Error: Excluding a pixel that should have been included in the class (i.e., omission error = 1 - producers accuracy)
Commission Error: Including a pixel in a class when it should have been excluded (i.e., commission error = 1 - user's accuracy.
Kappa coefficient: Measures the improvement of the classified map over a random class assignment

Depends on audience
Depends on the objective of your study
Most references suggest full reporting of error matrix, user’s and producer’s accuracies, total accuracy, and Kappa

Errors in reference data
Positional error
Better rectification of image may help
Interpreter error

Errors in classified map
remotely-sensed data cannot capture classes
Classes are land use, not land cover
Classes not spectrally separable
Atmospheric effects mask subtle differences
Spatial scale of remote sensing instrument does not match classification scheme

Land use/land cover: incorporate other data (in spectral signature…)
Spectral inseparability: add spectral data
Atmospheric effects: Atmospheric correction may help
Spatial pattern recognition – brings in context via image texture, pixel proximity, feature size, shape, directionality, repetition
Temporal pattern recognition – dynamics of spectral response patterns in multi-date imagery
Spectral Unmixing – deconvolving ‘mixels’ or mixed pixels

Rarely a 1-to-1 match between spectral classes and information classes
Unique spectral classes may occur that do not belong to information classes of interest
A broad information class may include a number of spectral sub-classes (e.g. dry soil, wet soil)
Must decide on the utility of the spectral classes and decide how to aggregate or distribute them

normally required prior to the main data analysis and extraction of information, and are generally grouped as radiometric and geometric corrections…

include correcting the data for sensor irregularities and unwanted sensor or atmospheric noise, and converting the data so they accurately represent the reflected or emitted radiation measured by the sensor.
Ie removal of sensor or atmospheric 'noise', to more accurately represent ground conditions - improve image‘fidelity’:

include correcting for geometric distortions due to sensor-Earth geometry variations, and conversion of the data to real world coordinates (e.g. latitude and longitude) on the Earth's surface.
Ie conversion of data to ground coordinates by removal of distortions from sensor geometry

All remote sensing imagery are inherently subject to geometric distortions.
Geometric corrections are intended to compensate for these distortions so that the geometric representation of the imagery will be as close as possible to the real world

Systematic distortions –
Sensor, platform movement ➔ can be predicted
are well understood or predictable
Accounted for by modelling sensor and platform motion and its geometric relationship with the Earth
e.g While in orbit, the sensor may experience shifts and moves…

Random / unsystematic distortions
Can not be predicted ➔ Geometric registration
Image-to-map Registration
The true ground coordinates are typically measured from a map either in paper or digital format
This is image-to-map registration

In general, a pixel in the output image will not directly overlay a pixel in the original image

Geometric registration may also be performed by registering one (or more) images to another image

Resampling
In order to actually geometrically correct the original distorted image, a procedure called resampling is used to determine the digital values to place in the new pixel locations of the corrected output image.
The resampling process calculates the new pixel values from the original digital pixel values in the uncorrected image.

Nearest Neighbour (NN)
NN resampling uses the digital value from the pixel in the original image which is nearest to the new pixel location in the corrected image
This is the simplest method and does not alter the original DNs, but may result in some pixel values being duplicated while others are lost

Bi-linear Interpolation (BI)
BI calculates a weighted average of four pixels in the original image nearest to the new pixel location
The averaging process alters the original pixel values and creates entirely new digital values in the output image
This may be undesirable if further processing and analysis, such as classification based on spectral response is to be done

Cubic Convolution
CC calculates a distance weighted average from a block of sixteen pixels from the original image which surround the new output pixel location.
Results in completely new pixel values.
However, this method produces images which have a much sharper appearance and avoid the blocky appearance of the NN method.

Radiometric correction is made on the brightness (gray level) values of the image.
* The need to perform radiometric corrections, more specifically the atmospheric correction depends directly on the particular application
* Every time two images need to be combined (e.g., in a mosaic) or compared (image-image, imageground data), the corrections become obviously important

The radiance L refers to the radiation according to a certain angle of observation, indicating the radiant outgoing flux per surface unit and per solid angle unit, measured on a perpendicular plane to the considered direction.

Irradiance: Incoming radiant flux density in Watts m-2 or J/(m2 . s)

Radiance: (L ) is the amount of EMR leaving a point on a surface, is the most precise remote sensing radiometric measurement. It is measured in Watts per meter squared per steradian (W . m-2 . sr -1 ), or it is measured in Watts per meter squared per wavelength per steradian (W m-2 . sr -1 . m–1 ) (= Spectral Radiance: Radiance per unit wavelength)

Reflectance: Reflected radiance in one direction divided by incoming radiant flux density (or irradiance) – unitless

Exitance: outgoing radiant flux density Watts m-2 or J/(m2·s)

Albedo: Outgoing radiant flux density (exitance) divided by incoming radiant flux density (irradiance) from a flat surface in all directions - unitless

Step 1: Sensor Calibration Converts DN values back to radiance. Includes: Sensor malfunction corrections (noise removal)
Step 2: Conversion of At-Sensor spectral radiance to At-Sensor Spectral Reflectance (relative to Incident Radiation – unitless)
Step 3: Atmospheric Corrections

noise is any unwanted disturbance in image data that is due to limitations in the sensing, signal digitization, or data recording process •
Systematic
Striping
Valid data are present in the defective lines but they must be normalized with respect to their neighbouring observations. Normalized values are applied only to problem lines, the others are not altered.
Line Drop
Occurs when there are system errors which result in missing or defective data along a scan line
Are normally 'corrected' by replacing the line with the pixel values in the line above or below, or with the average of
Random
Random noise i.e., “salt and pepper” or “snowy” appearance.
is characterized by nonsystematic variations in gray levels from pixel to pixel called bit errors

Eliminates the effect of atmospheric scattering and absorption on radiation values.
Source
Scattering may attenuate (reduce) some of the energy illuminating the surface
Corrections
Image based/histogram minimum method
Physical models
Image based/histogram minimum method
in an area of shadow or for a very dark object (such as a large clear lake - A) and determine the minimum value (B)
The correction is applied by subtracting the minimum observed value, determined for each specific band, from all pixel values in each respective band
Physical models simulate the physical process of scattering at the level of individual particles and molecules

1. Physical Basis for Remote Sensing of Vegetation Factors that determine the spectral signature of vegetation
2. Spectral Vegetation Indices
3. Remote sensing of vegetation for ?

photosynthesis
- Absorption (A) Transmission (T) Reflection ®
- interaction radiation with target ➔ spectral signature the specific combination of emitted, reflected or absorbed electromagnetic radiation (EMR) at varying wavelengths which can uniquely identify an object.
Leaf structure
- Pigment absorption dominates visible light!
Plant water content
- Middle-infrared energy interaction with water in the spongy mesophyll

(NIR - RED) / (NIR + RED)
most common
NDVI is the most widely used index making it easier if you want to compare your data. As well as this NDVI is great at gauging the health of vegetation, so I would use this index if that was my goal. Also, NDVI is great at reducing noise so if my image had a lot of noise, I would go with NDVI.

NIR / RED
If measuring in an area with lots of atmospheric conditions such as clouds, then SR may be a better choice as NDVI’s accuracy can be overly reduced by these conditions. Also, SR may be a better choice in overly dense vegetation areas as NDVI may become too saturated and SR can better avoid saturation.

Quantitative analysis of “greenness”:
“Green chromatic coordinate (gcc)” = Green DN / (Red DN + Green DN + Blue DN)
“Green Excess (ExG)” = 2 * Green DN – (Red DN + Blue DN)

The integrated NDVI is the sum of all NDVI values (per location) over the growing season
The integrated NDVI has been interpreted as equal to Net Primary Production (NPP) e.g. amount of photosynthesis to have taken place over the growing Season
Timing of plant growth (when they grow, when they senesce)