Module 5: Unsupervised and Supervised Classification

 Module 5: Unsupervised and Supervised Classification

This weeks lab we learned how to use two different ways to look at Land Use\Land Cover (LULC) by unsupervised and supervised classification. Image classification is the process of giving discrete and unique values to all pixels in a raster image.

First we learned how to do unsupervised classification with ERDAS. The important part of this portion was trying to understand that as the GIS analyst you need make sure the pixels actually make sense. When creating the pixels we were told to do 50 different classes with a 95% reliability of the pixels. We opened the attribute table to change the different pixels to certain colors. While doing this you could see how some of the pixels were showing up in places that did not make sense. An example was when doing the grass class there were some pixels where it showed up on the roof of a building. Also, the grass was considered a shadow class. Even though unsupervised classification is faster it does not always show up in the correct pixels.

Second we learned how to do supervised classification with ERDAS. In this portion of the lab the supervised classification gives the GIS analyst a little more control on what the different colors are going to be. There were two tool methods for creating spectral signature. One of the tool methods you can use is using the inquire tool and put in the coordinates you need for the location in UTM. Then under the Drawing tab go to the Geometry tab and click the polygon button. You then will draw your own boundary. After that you go to the Signature editor and click on Create New Signature(s) from AOI button to make your new class. The second tool method is using the inquire (legacy) tool and once again you put in the coordinates you need for the location in UTM. Then under the Drawing tab go to the Geometry tab and click on the Grow tab and click on Growing Properties. Once in there make sure to click on the At Inquire button to get you to the location you put in the inquire (legacy) box. When looking at the box you can look at the table and see if there are any other colors in the box. You can change the Spectral Euclidean Distance and Neighborhood boxes. Also, you can move the box to smaller if you just want one color. The same as the other method you go to the Signature editor and click on Create New Signature(s) from AOI button to make your new class.

The image below is a supervised classification of Georgetown, Maryland and the current LULC. While doing this image I did the steps for the supervised classification, but there were some hiccups. The major hiccup I had was making sure the signature file I created was big enough that the color was correct on the image. Another hiccup was even though I had everything that was big enough there were two classes that kept getting mixed up in the image. I tried to create the signature file a couple of times, but I was not able to get it correctly. The example for this was how the roads and Urban\residential area kept getting mixed up in the image.

Module 4: Spatial Enhancements, Multispectral Data, and Band Indices

Module 4: Spatial Enhancements, Multispectral Data, and Band Indices

In this weeks lab we learned even more in ERDAS Imagine and how to interpret our data we are given from different websites. First we learned how to take data from Glovis and EarthExplorer on USGS. When looking at websites it is important to know where it is coming from and that it is a credible source. Also, many of the data is already in the source we need it for ERDAS but it is important to make sure they are .img so they can be looked at in ERDAS. Lastly, we need to know what processing they used to make the images and if we need to do anything else such as taking the haze out of the photo. 

The second thing we learned was using spatial enhancements in both ERDAS Imagine and ArcGIS Pro. In ERDAS and ArcGIS Pro you are able to have the image go through different filters, such as low filter and high filter. Both of the filters do important things to the image depending on what you need it to look like in the final version. The low pass filter allows low frequency data, or data that does not change much from pixel to neighboring pixel, to pass through, removing the high frequency data, or data that changes rapidly from pixel to neighboring pixel. The visible result is the image appears blurred or smoothed. High pass filter allows high frequency data to pass through, or data that changes rapidly from pixel to neighboring pixel, suppressing low frequency data, or data that does not change much from pixel to neighboring pixel. High pass filtering can be useful for finding edges, enhancing lines and edges, or sharpening an image. Being able to do these can help with what it is in your image and show it easily in your final version. Lastly, you can sharpen an image to help get the boundaries of rivers and buildings more precise.

Next we learned how to look at the Histogram data to understand what the image statistical data to help us interpret the data. This is important because it helps the GIS analyst represent the data in a manner that everyone can understand. Also, we learned how to look at the spectral characteristics of the different band combinations to show either True Color, False Color IR, and False Natural Color. Being able to know the different band combinations helps because depending on what you are trying to represent you need to change the image. Lastly we learned calculating differences between different spectral bands, an index can be created to further enhance the appearance of certain features. We created a Normalized Differential Vegetation Index (NDVI) to help distinguish clearcut areas. Knowing how to do this can help enhance features within the image.

Lastly, we took all the things we learned from above to make three different images. This was to help us understand what we could do to data and present it to the public. First image is a False Natural Color to help show the water feature but not be too bright like False IR. The second image is a False Color Infrared to distinguish the mountain top compared to the rest of the landscape. The last image is a True Color to show the different areas of the water feature and how they change colors.

Module 3: Intro to ERDAS Imagine and Digital Data

 Module 3: Intro to ERDAS Imagine and Digital Data

In this weeks lab we learned about how to calculate the Electromagnetic Radiation (EMR) and understanding how EMR transmits/travels through the landscape. Also, in this weeks lab we learned how to use ERDAS Imagine and why one would use it. Lastly, in this weeks lab we learned how different passive and active sensors can present their data in ERDAS Imagine.

It is important to note that EMR energy comes from the sun and is a continuous spectrum of energy. Also, EMR has different wavelengths and frequencies that range from a longer wavelength equals a lower frequency to a shorter wavelength equals a higher frequency. Nowadays the equations to know the frequency and wavelengths are done automatically through the remote sensors. We did learn how to do basic equations called Maxwell’s wave theory: C = λν and Planck Relation: Q = hν. The Maxwell's wave theory is the relationship between the wavelength and frequency of the EMR. The Planck Relations is the amount of energy is inversely related to the wavelength of the light. This means that the shorter the wavelength, the greater the energy of each photon of the light. Being able to understand how the remote sensors interpret the data can help with the GIS analyst knowing what EMR energy is being collected.

I have been using remote sensing machines for awhile but I have always used ArcGIS Pro after the data was processed. This weeks lab showed another way to look at data and that is ERDAS Imagine. This was supposed to help us understand how to look at large-data sets from remote sensors. First we learned how to put in Raster data into the data layer. The raster layers can be raw/multiple layer continuous data, single layer panchromatic continuous data, or categorical/single layer thematic data. This is important because the raster data is meant for many different sensors. Also, when presenting the Raster Options and make sure to use the correct Display As tab. They are True Color, Pseudo Color, Grayscale, and Relief. In the lab we used Pseudo Color because it displays thematic images by associated each class value a color or single layer. Another tab we need to click on is Multiple and click on the Multiple Independent Files because it brings in files, individually, as separate layers that can be arranged or deleted separately from the other files in the Viewer. Lastly, click OK button and it will present your data on the page. Many of the other features are the same as ArcGIS Pro when zooming in and out as well as clicking on the different tabs to get information you may need to understand the data.

While being in ERDAS Imagine you can create an image that can be put into ArcGIS Pro. The way this is done is by opening the data imagine you want onto the layer. Since much of this data has a lot of information you will want to only take a portion of the image out. This is done by clicking on the Home tab and hit the button next to Inquire tab until you see Inquire Box. When it opens up you can change the size of the box and have that as the boundary. In the Raster tab, under the Geometry group, you will expand Subset and Chip menu, and select Create Subset Image. Then you will press the From Inquire Box button to set the appropriate area. Lastly, under the Output File navigate to your Output folder and name the file. Also, in this image we needed to get the Area of the different classification categories. This was done by right clicking on the layer I wanted and clicking Display Attribute Table and it will show up on the bottom of the screen. Then you will click on the Table tab and hit the Add Area tab. It will then show up and you can pick different ways of presenting the data but we chose hectares. The image below is how I presented the data in ArcGIS Pro.

Lastly, in this weeks lab was looking at the layer information, spatial resolution, and radiometric resolution. To see the layer information make sure to click on the layer you want and click on the Metadata tab. It will present a lot of data starting with File Info: which shows the basic information about the imagine including file type and file path. Next it is Layer Info: which shows the width and height in pixels, Type (if it continuous or thematic), and Data Type (how the pixel data is stored). Also, this is where you find the radiometric resolution that is usually expressed as a bit number (Unsigned 8-bit). Next tab is Statistics Info: which shows the min and max of the pixel darkness (min) or brightness (max). Also, shows the mean, median, and mode which the values are the measures of the average brightness of all pixels in the layer. Whereas the standard deviation  represents how close most of the pixels fall to the mean value. Next tab shows Map Info (Pixel Center): which shows provides the spatial coordinates of the corners of the image in the projection and the height and width of each pixel in the image. The last tab is Projection Info: tells you how the data is being projected and if it is not geo-rectified it will be blank in this portion. Knowing how to read the metadata can help understand how the data was collected and if it was projected correctly.