Remote Sensing Final

Landsat

Name the spectral (by band name) of TM on Landsat 4,5 (not including the thermal band)

Band 1: 0.45-0.52 (blue-green water soil vegetation)

Band 2: 0.52-0.60 (green, chlorophyll reflection)

Band 3: 0.63-0.69 (red, chlorophyll absorption)

Band 4: 0.76-0.90 (near infrared , vegetation biomass)

Band 5: 1.55-1.75 (mid infrared, vegetation moisture, cloud, snow, ice)

Band 7: 2.08-2.35 (mid infrared, geologic rock formation, plant heat stressaq)

Name the spatial resolution of TM on Landsat 4,5 (not including the thermal band)

30 × 30 meters

Name the spectral and spatial resolution of SPOT 1,2,3, for the Pan bands

spectral resolution: 0.51 - 0.73 microm

spatial resolution : 10 × 10 meters

Name the spectral and spatial resolution of SPOT 1,2,3 for the XS bands

spectral resolution:

band 1: 0.50-0.59 (green)

band 2 : 0.61-0.68 (red)

band 3 : 0.79 - 0.89 (near infrared)

Spatial resolution : 20 × 20 meters

What is the purpose of band ratioing?

The purpose is to reduce redundancy caused by overall brightness differences, to transform raw band images to component images and to enhance spectral differences between surface features.

What are the types of bands that should be used

Red and NIR Bands.

Bands must be from the same sensor, acquired the same time.

Explain the design principle behind band ratioing

NDVI = (NIR-RED) / (NIR+RED)

What is the purpose of IHS (Intensity-Hue Saturation)

is to separate brightness from color information in an image.

How IHS (Intensity-Hue Saturation) works

1) RGB space is used to represent XS bands

2) RGB is transformed into IHS space

3) The pan band is fused into the intensity component of create a modified IHS space

4) The modified IHS space is back transformed to RGB

RESULT: The pan band spatial detail is integrated with XS spectral (color) information

How does the RGB cube turn into the IHS hexagon

1) the hex cone model projects the RGB cube into a plane, resulting in a hexagon.

2) the plane is perpendicular tot he gray line and tangent to the RGB cube at the white corner

why is it called “Unsupervised” classifications

because the process requires minimal input from the analyst.

What are the two passes for the Chain Method

1) First Pass = goes through the image pixel by pixel, line by line and builds clusters. Calculates mean vectors as well

2) Second Pass = goes through the image again, assigns each pixel of the image to clusters based on the minimum distance rule.

what are the purposes of the two passes for the chain method

1) First Pass = is to create initial clusters on spectral similarity

2) Second Pass = to refine and stabilize clusters which improves class consistency

ISODATA method

how to determine the initial means

based on the mean and standard deviation of each band

the recruiting and re-calculation of the mean in each iteration

In each iteration, pixels are classified and cluster mean vectors are calculated.

Iterations are continuous until parameters are met

T: the maximum percentage of pixels whose classes are allowed to be unchanged between iterations

M: the maximum amount of iterations to classify pixels and calculate cluster mean vectors.

what is the motivation behind the development of spectral mixture analysis?

was that the variation in a image is a mixture of limited features. Instead, SMA compares mixed spectral signatures to pure reference spectral.

Whats the motivation behind Object based image analysis (OBIA)

was created as a alternative to per pixel classification

all classifiers so far consider the spectral info of a single pixel regardless of its neighbors

what are the two sets of characteristics used to classify objects

1) the characteristics of the object itself (spectral, texture, shapes)

2) the relationship between those objects ( connectivity, proximity, etc)

why is it called supervised classification

because it requires interaction with the analyst

(analyst identities the training sets and develops a summary statistics for each category)

Minimum Distance Classifier

the means in the band space

for each category in the training set, the classifier calculates the mean spectral value in each band.

the distance between a pixel and the mean of a category in the space

for a pixel with a unknown identity the classifier measures the distance from the pixels spectral values to the mean of each category in each band space

assigning the pixel to the category of the highest probability

the pixel is assigned to the category whose mean is the closest to the pixel in the band space

if the pixel falls beyond the distance defined by the analyst, it can be labeled as “unknown”

Gaussian Maximum Likelihood Classifier

what is the probability distribution of the DN values in the band space.

the probability distribution of the training set is normal

what is the membership of a pixel in all categories

its described in probability terms. Probability computed with probability functions.

is the pixel assigned to the category of the highest probability

the pixel is assigned to the category with the highest probability

Anderson Classification System

what is the difference between land use and land cover

Land Use = reflects peoples relation to the environment

Land Cover = natural / artificial compositions at the surface

Classification Accuracy

how to calculate overall accuracy

overall accuracy = total correctly classified pixels / total pixels evaluated

how to calculate producers accuracy

correctly classified in each category / the total pixels in the category (column total)

how to calculate Omission Error

Omission Error = 1 - Producers Accuracy

how to calculate Users Accuracy

correctly classified in each category / total pixels in the category (ROW TOTAL)

how to calculate Commission Error

Commission Error = 1- Users Accuracy

whats KHAT statistics design principles

accuracy measure based on the error metric that accounts for chance agreement.

Aerial Photography

what is feature height

feature height is used to measure the height of objects and is determined by vertical displacement

Aircraft Instability : Roll, Crab, Pitch

Roll= Side to side motion of the aircraft

Crab (Yaw) = rotation of the aircraft caused by compensating for drift

Pitch= foward and backward (head and tail) motion

Aircraft Instability : Relief Displacement

Relief Displacement is caused by vertical features on aerial photographs. The farther the feature is from the principal point, the taler the feature is.

Thermal Remote Sensing (Nathan Dubinin lecture)

what is heat capacity

controls how much a materials temperature changes through time

what is specific heat

the amount of heat required to raise the temperature of a material

Thermal Image

Landscape Factors

surface material : Land Vs Water

topography : Sunlight Vs Shadowed

vegetation cover : Land, Grass, Forest, Water

moisture

Thermal Image

Timing

Early Afternoon : High thermal contrast, strong differences between materials, effects of slope orientation and thermal shadows present.

Water is cool; bare soil; meadow and forest are warmer

Before Dawn: Lower thermal contrast but little slope orientation effects on thermal snow

water is warm; then forest; open meadow and bare soil are cool

Why Nathan’s group waited until afternoon to image artifacts using thermal sensors

They waited until the afternoon because thermal sensors works best after the area has absorbed electro magnetic radiation. When they waited they returned to different materials had absorbed heat at different rates. This increased thermal contrast ( made thermal anomalies easier to point out)

from Rich Quodomine’s lecture, whats the big event that the LA Metro is preparing for AND the sensor system involved

The 2028 Olympic Games ; the sensors that are being used are UAV drone data collection systems

Microwave Remote Sensing

describe active

the sensor transmits microwave energy towards the surface.

measures the returned signal

Example: SLAR, SAR

Microwave Remote Sensing

describe passive

records naturally emitted microwave energy

does not transmit energy

Polarization

Describe HH

HH: radar transmits horizontally polarized signals and receives the horizontally polarized return signals

Describe HV

HV: radar transmits horizontally polarized signals but receives vertically polarized signals

Depolarization

the horizontally polarized outgoing microwave energy that is returned as vertically polarized energy

depolarizers on the ground appear brighter on HV image
rough surface and inhomogeneous subsurface are depolarizers

Doppler Effect

1) objects experience frequency shifts in relation to their distance from the aircraft

2) objects at the leading edge of the radar beam reflect pulses with a higher frequency than objects at the trailing edge

3) the frequency shift allows the SAR system to assign reflections to features at their correct positions along the flight direction

Define Specular Reflectance

1) when the surface is smooth relative to the wavelength

2) incident angle = reflection angle

Diffuse Reflection

1) when surface is rough relative to the wavelength

2) the signal will be scattered equally all directions

Corner Reflection

1) a double reflection caused by adjacent smooth surface

2) high reflectance appears as sparkles on the image

3) tends to be proportionately larger than its real size

Radar image brightness depends on: Slope facing, Surface roughness define them

Slope Facing = determines signal returns

Surface Roughness= determines the type and amount of returns

SRTM (Shuttle Radar Topography Mission)

Used two radar antennas mounted on the same space shuttle which collected data simultaneously. The goal of the mission was to produce near global elevation data

LIDAR - principles of operation

1) LIDAR operates by sending rapid pulses of laser light towards the ground

2) the system measures distance by calculating the return time of the reflected pulses

3) operates in the visible and near infrared (NIR) range.

Google Earth Engine

Ready to use Datasets

Multi-source remote sensing data already available for analysis

Includes optical imagery, LiDAR data, SAR data, Google Earth imagery, and NAIP imagery

Used to study land change, human-induced environmental change, air quality, deforestation-related fire, and offshore infrastructure

Web-based code editor & powerful API

Google Earth Engine enables analysis of large remote sensing datasets

Supports automated extraction and analysis of features such as offshore infrastructure

Integrates multiple remote sensing sources within one platform for land change science and Remote Sensing of Society studies

Real-world applications

Understanding land changes primarily induced by human activities

Monitoring deforestation-related fires and air quality (aerosol optical depth, particulate matter)

Observing environmental impacts on society