Module 10: Supervised Classification

This week we continued to look at automated classifications, focusing on Supervised Classifications.  Supervised Classification requires the user to set "training sets" to guide in the separation of the pixels into classes.  The objective of the training sets is to select a homogeneous area for each spectral class.  Selection of multiple training sets helps to identify the many possible spectral classes in each information class of interest.  Training sets can be selected by 1)digitize polygons, that have a high degree of user control but often result in overestimate of spectral class and 2)seed pixel where the user sets thresholds.  Both of these methods were utilized this week in lab.  Training aids help evaluate spectral class separability, for example graphical plots like histograms, coincident spectral mean plots, or scatter plots.  There are also statistical measures of separability like divergence or Mahalanobis distance.  Parametric distance approaches are based on statistical parameters assuming normal distribution of the clusters (mean, std deviation, co-variance).  Examples of parametric methods are mahalanobis distance, minimum distance and maximum likelihood.  Non-Parametric distance approaches are not based on statistics but on discrete objects and simple spectral distance (examples are feature space and parallel-piped).

In lab this week we produced classified images from satellite data.  We created spectral signatures and areas of interest (AOI) features utilizing both digitize polygons and seed pixels.  We evaluated signatures by with histogram plots and mean plots to recognize and limit spectral confusion between spectral signatures.  We classified images utilizing maximum likelihood classification and then merged classes and calculated area.

Module 9: Image Classification

This week we continued with image classification.  Image Classification is a major application of remotely sensed imagery to provide information on land use and land cover.  In a previous weeks we utilized tone, texture, shape, size, pattern, shadow and surroundings to identify and classify.  Then moved on to looking at different spectral bands.  This week we move toward automated procedures.  Spectral Pattern Recognition is a numerical process whereby elements of multispectral image data sets are categorized into a limited number of spectrally separable, discrete classes.  Unsupervised Classification, the focus this week, uses some form of clustering algorithm to decide which land cover type each pixel most looks like.  Supervised Classification, next week, uses training sites from multiple spectral band data to guide the computer's classification.   

Tools like Feature Space Image allows visualization of 2 bands of image data simultaneously through a 2 band scatterplot, one band plotted on the X and the other on the Y axis.  This tool helps consider covariance, correlation and clustering between bands.  Spectral Distance, is one means of predicting the likelihood that a pixel belongs to one class or another.  Spectral Distance can be measured as simple euclidean distance in unsupervised or a statistical distance in supervised.  These tools help to differentiate spectral classes, clusters of spectrally similar pixels, that can then be used to form information classes, meaningful groups.  

This week in lab we performed unsupervised classification in both ArcMap and ERDAS.  We attempted to accurately classify images of different spatial and spectral resolutions.  And we manually reclassified and recoded images to simplify the data.  We utilized ISODATA (Iterative Self-Organizing Data Analysis Technique) Clustering Algorithm.  

The map this week is of the UWF campus.  The main map was created by utilizing ISODATA classification in ERDAS.  The tool was set to 50 classes with 25 iterations with a convergence threshold of 0.95 and a skip factor of 2 for both X & Y.  Reclassification of the 50 classes to 5:  trees, grass, building&roads, shadows and mix.  The reclassification was then merged into the 5 classes.  The lab instructions were to calculate an estimate of porous and non-porous surfaces in the image.  I decided to run ISODATA with the same specifications and reclass based on porous, non-porous, and mixed.  Both processes proved difficult primarily due to pixels categorized in the same groups of 50 that were different information classes.  Pixels for roads, parking lots and roof tops sharing groupings with grass or bare ground.   With the original 50 classes to ulizmarly get 5 classes for the main map and 3 classes for the insert the error ratio, based only on my opinion of the visual inspection, is still high.

Module 8: Thermal Infrared

This week in Photo interpretation we learned about thermal energy.  Thermal energy is heat emitted from objects.  Thermal energy waves are much larger than the waves we have previously discussed.  Due to the size of the waves there is very little scattering, but the range of measurable waves is limited by the atmosphere, by elements like water.  Stefan-Bolzmann law states radiation increases with temperature. 

In lab this week we worked between ArcMap (my ArcPro was not cooperating) and ERDAS.  We learned to create composite multispectral images in both ERDAS with layer stack and ArcMap with the composite band tool.

The map deliverable shows the difference in imagery displaying different bands in different colors.  I chose to look for three features recognizable by me: Interstate 10, Fairhope Pier, and the Grand Hotel.  Fairhope Pier is really only discernible in the True Color display and then only when zoomed very close.  The interstate gets lost under what I think are clouds in the Mobile Bay Estuary, but at least a part can be seen in the thermal layer (I think).  The False Color shows clearly the areas of high vegetation in dark red wich are also the lighter areas on the Thermal display as those are cooler than the urban areas.  The Grand Hotel grounds showed up very well in the False Color.

Multi Spectral Analysis

This week in lab we gain more information about spectral bands, band ratios, spectral properties of vegetation.  We explored some vegetation indices: NDVI, SAVI and EVI.  And we discussed spectral enhancements in the form of Tasselled cap and Principle Components Analysis.

In lab we continue to utilize ERDAS to explore Image Histograms.  We gained experience operating the inquire cursor, interpret histogram data, and identifying features by interpreting digital data.

I must admit that I am unsure of my results this week.  I worked through the lab instructions and made it to the deliverable assignment and read it and thought I do not know how to do that!  But I kept playing around in ERDAS.  I examined the histogram, and looked at the image in gray scale and as multispectral changing the band combination.  I looked at the histogram again, still not sure what I am supposed to be getting.  Then I remembered the histogram X axis is the brightness and the Y axis the frequency of that brightness.So then I had some clues directing to bright objects or dark objects and how frequent.  Then in grey scale I looked for changes to the imagery in different bands.  And I repeated that some with the multispectral and adjusted the bands (not so much time here because I could keep trying different combinations for days).  Then I utilized the Inquire Cursor to see if I had put the clues together in an area the meet the pixel values provided.  I could be totally wrong on this one.  But I tried to follow the clues as best I could and I came up with answers, and SOME COOL (may be correct) MAPS.

Module 6: Image Enhancement

This week we covered Radiometric Correction or Enhancements to account for effects of sensor-detector-and platform, atmospheric and illumination effects, and terrain effects.  We also looked at Spatial Enhancements with Pan-Sharpening (merging a panchromatic image at high resolution with an image at lower resolution) and spatial filters.  We were presented with:  types of radiometric effects that can impact remotely sensed data, differences between absolute and relative atmospheric correction, pan sharpening, convolution filter (high and low pass filters), and Fourier transform.

This week in lab we walked through the steps of downloading and importing satellite imagery from https://glovis.usgs.gov/  (which wasn't loading properly at the time), performed spatial enhancements in ArcMap and ERDAS and utilized Fourier Transform function.  The goal of the above image was to restrict the striping impact of the image while keeping the detail of the image.  I don't think mine is a very good example.  This was much more difficult and felt more of an art than a science.  The lab instructions included several enhancement techniques, a wedge mask, a low pass filter and then a sharpening.  From there the direction was our own.  I applied another low pass filter of 5X5 in addition to the 3X3 we used earlier and then sharpened again.  The second sharpening did not seem to enhance as much as the earlier.  And although the additional low pass toned down the striping I feel I lost significant detail in the process.  Previous to this exercise I would have though that image correction would involve specific calculations and then transformations to correct.  This was not my experience with this lab.  It was more of trial and error.  Maybe this changes with more experience or maybe the experiences shape the decisions of how to correct.

Module 5a - Intro to ERDAS Imagine & Digital Data

This week in lecture we learned about Electromagnetic Radiation (EMR), models of EMR focusing on the Wave model, Electromagnetic Spectrum, EMR interactions of refraction, scattering, absorption, reflectance and transmission.

In lab this week we were exposed to ERDAS Imagine for the first time.  We used basic tools to add data, fit to frame, adjust vector symbology, navigate the image, set default data directory and default output directory, and changed band combinations to examine enhancements.  The project we added an area column in the attribute table in ERDAS, selected a small area of image utilizing the inquire box and created a subset of the image from the inquire box and saved to output as .img.  The .img was opened in Arc Pro and the above map created to show the classes within my subset and the area of each of classes.

Module 4: Ground Truthing and Accuracy Assessment

This week in lab we took our LULC map from last week and performed ground truth via Google Map.  I randomly selected 30 points in a grid fashion and then adjusted to account for each land type from the original map.  Each point was located and verified in google map for the location land use.  Here is a summary table of those results.

Of the 30 sample points 24 of them were classified correctly for an overall accuracy of 80%.  The smaller table above shows the accuracy broken out by classifications.  Mixed Urban had 0% correct.  Since I utilized this class for areas that I could not completely identify when classifying it is reasonable that with additional data the need for this classification would be eliminated.

Land Use Land Cover Classification

This week in lab we practiced our skills of recognizing features on the ground using a natural color aerial photograph.  We digitized an area of Pascagoula, MS, creating a land use/land cover map.  We identified the ground based on size, shape, color, pattern, shadows and association.  We utilized the USGS Standard Land Use / Land Cover Classification System.  The assignment was to categorize to level two (two digit classification).  Classifications of three or higher are typically utilized for local and some regional planning. 

My specific categories for this project 

Part of my employment background was working for 10 years in a building department in a municipality.  We worked very closely with the planning department and GIS guy.  Identifying urban land use from aerial perspectives I have done before.  It is a lot easier when you are familiar with the area.  Correctly identifying natural elements just at a level 2 classification was difficult because I am not familiar with the difference in kinds of trees or when a stream or canal becomes part of an estuary or bay. 

Technically doing the polygons was new and still somewhat challenging in the new to me GIS Pro.  My first attempt I did not have the Edge snapping or vertex snapping turned on, so I utilized this attempt as practice.  The second attempt I did turn on the snapping features while creating the polygons.  First creating a feature class file in my gdb.  Then in the edit tab clicking the create button and choosing polygon.  It was challenging to be able to navigate the picture while creating the polygon.  I settled on zooming in and out with the mouse scroll to move the image without having to click.  I did not master the clip portion to separate smaller interior polygons from larger surrounding areas.  Instead I relied on more of a lasso method with the larger polygon.  I started with larger easily identified areas keeping my attribute table open and adding the code as I drew each polygon (realizing that you must click off the cell in the attribute table that you are making changes to before you save the changes or that cell will not be recorded).  As I progressed from larger isolated polygons then I started working right and left back and forth to fill in the surrounding areas and smaller isolations.  Distinguishing between commercial service and commercial industrial was harder without knowledge of the practice of the location.  Ultimately, I settled on cleaner sites without outside materials or industrial roof or ground mechanisms and Commercial Service and those with as Commercial Industrial.

Visual Interpretation

This week in Photo Interpretation and Remote Sensing we  learned more about the types of and techniques for interpreting aerial photography. We were provided background on the types of aerial cameras, the different types of images they capture (oblique, vertical, stereo), the types of film used by traditional cameras, and understand how resolution (spatial, spectral, temporal) applies to aerial photography. We also learned concepts, techniques, and application of visually interpreting aerial photos. We learned about the methods and techniques used to visually interpret aerial photos (i.e. recognition elements). These techniques form the basis for deriving geographic features and/or land use land cover types from aerial photos and that are used in a wide variety of real world applications. 

In the laboratory exercise, we learned some basic principles of interpreting features found on aerial photographs.  The first map above illustrates ranges of tone and texture for this photo.  The second map above demonstrates examples of elements identified, at least partially, by shape and size, pattern, shadow, and association of the surroundings.  We also examined a true color image (blue, green, red), identifying 5 areas of color and compared those elements in the same image provided in false color (green, red, and infrared).

We have moved from ArcGIS desktop to ArcGIS Pro and it has been a disorienting week.  The changes are vast! location of symbology, labeling, and properties are different, but even the environment has changed to a project oriented system instead of .mdx maps.  I am sure I will catch up as soon as I figure out how to control drive location.