Lab 6 Flashcards
What is the relationship between information classes and spectral classes?
Information classes are categories of interest that the analyst is actually trying to identify in the imagery (e.g. crops, forest type, tree species). Spectral classes are groups of pixels that are uniform in brightness values in the different spectral channels of data. Through identifying spectral sub classes that identify the targeted feature we can relate spectral classes to their corresponding informational classes.
Provide an example of the following types of data integration:
a. Multitemporal
Data integration is the combining or merging of data from multiple sources into an effort to extract better and more information.
Multitemporal- Imagery collected at different times is integrated to identify areas of
change. An example is looking at changes in the landscape before and after a wildfire.
Provide an example of the following types of data integration: Multiresolution
Multiresolution- Mergeing of data of different resolutions, the merging of lower resolution
data with higher spatial resolution can sharpen the spatial detal in an image and enhance
the discrimation of features. One example is SPOT data.
Provide an example of the following types of data integration: Multisensor
Multisensor- Merging data from different sensors, one example is combining multispectral
optical data with radar imagery.
Provide an example of the following types of data integration: Multi-data type
Multi-data type- Combining different types of data and from different sources. One
example is DEM/DTM, or including maps of soils, forest types into the GIS.
What are the data requirements for crop type mapping? What data from class exercises would be suitable for this purpose?
It is important to collect crop types and delineate their extent. Traditionally methods such as information census and ground surveying were used. Now with remote sensing other data can be collected such as health, heights, moisture content.
Multitemporal imagery, multisenser data and multispectral sensors are needed.
Landcover types, lidar (to tell heights, distinguish between crop types). Using multispectral different bands to classify moisture content, chlorophyll (NDVI), etc.
What are the data requirements for clear cut mapping & deforestation mapping? What data from class exercises would be suitable for this purpose?
Data used for clear cut mapping includes multitemporal data (change detection), high resolution data, and radar (view through clouds). Clear cut mapping also requires large area coverage and data continuity. Regional scale images with mid to high resolution is a must. Lidar data for mapping changes in heights so clear-cut areas in forest can be identified, also NDVI could be useful for identifying forest health around those clear-cut areas.
What are the data requirements for land cover / biomass mapping? What data from class exercises would be suitable for this purpose?
The use of moderate resolution data is appropriate for large scale projects. The use of high resolution might be needed for some applications such as wetland delineation. Data is time sensitive so knowing the time frames and having access to multi-temporal data is important. Thus, not only one season is being analyzed. Use of NDVI data can be used for this analysis.
What are the data requirements for ocean color & phytoplankton concentration mapping? What data from class exercises would be suitable for this purpose?
Ocean colour is used to analyze the health of the ocean. Multispectral data is required for colour measurements. Additionally, there needs to be a wide spatial coverage of the data. Hyperspectral data and multitemporal data are benefitial for capturing narrow ranges of wavelengths and seeing the seasonality of the changes in colour and phytoplankton for modeling trends. Use of different band data (in equations) for identifying specific wavelengths used in identifying phytoplankton is suitable.
What is the difference between supervised and unsupervised classification?
Image classification is extracting information classes from a multiband raster image. In arc the classification is done though the image classification toolbar. Supervised classification used signatures obtained through training samples to classify an image whereas unsupervised classification finds clusters without training and analyst intervention, the clusters are created from statistical properties of the pixels.
What is a cluster?
A group of pixels that is distinguishable in a multidimensional attribute space. The ground object that a cluster represents is unknown when the analysis is performed.
As compared with unsupervised classification, what steps are unique to the supervised
classification workflow?
In the supervised work flow, the training sample manager is used to identify features (through a series of the user creating polygons around desired features) and then using the Maximus likelihood classification tool to classify the input image.
In your own words, describe how Iso Cluster works.
The Iso cluster separates all cells in to specified groups. It is often used for unsupervised classification. It is an iterative process that assigns cells to a cluster based on Euclidean distance.
In your own words, describe how Maximum Likelihood Classification works.
Using statistical probability of 0.005 the number of cells is calculated that have the likelihood of being correlated to a particular value. The tool is based on two principles, cells being normally distributed and Bayes theorem of decision making. The tool uses the mean and covariance (how far way two variables are from each other and the mean, or the grouping of variables).
