The map below is a Land Use/ Land Cover map of Germantown, Maryland. The stratification of classes was performed in ERDAS Imagine using a supervised classification. I was presented with a false color image of the study area and used the Inquiry (Legacy) tool to located given metric coordinates to attempt to classify the similar pixels. I added an AOI layer so there would be a data table to ascribe to classification to on the image. There were some issues in using the Inquiry (Legacy) as it would not allow me to input coordinates unless the actual Inquiry tool was open. Once it was able to function, I was able to carry on by closing out of the actual Inquiry pane.
I maneuvered around the map with the Inquiry (Legacy) tool and would select a group of pixels using the 'at Inquiry' function on my Region Growing Properties function. This would present a polygon around a small and similar group of pixels. I ensured that as I selected the appropriate Spectral Euclidean Distance that I was not making the polygon so large that I was selecting pixel values I had already classified. For the larger regions of classification, it was useful to use the 8-way neighborhood and for the smaller areas the 4-way as it would add values far more conservatively allowing for a more accurate allocation.
As I completed each polygon, I added them to my Signature Editor and ascribed the appropriate classification nomenclature. I was quite pleased with my completed image, pre-merger. In the Signature Editor I then looked at the Display Mean Plot window to observe my values in graph form. The lines (values) with the greater distance between them over each spectral layer meant that there was the least amount of spectral confusion so I selected them (layers 5, 4, and 3) to assign my true colors to. This part offered some difficulty as, without looking in detail at the values, I could have argued that layers 4, 5, and 3 would have been a better order but I believed the gaps to be greater in descending order with the one used. I assigned these in the Set Signature Colors window.
After this I went to the Classification tab and clicked the Supervised button. This classified the entirety of my image. In addition I created a Distance file to illustrate the accuracy of my classification.
In the Thematic button I selected Recode and merged similar classifications to have the same unique 'New" classification numerical value. This is called merging. After recoding, I was then presented with a map with now only 8 classifications from the greater original amount.
Selecting the appropriate colors was rather daunting. There was an eclectic organization of information being displayed. I settled on the shown color scheme as it was aggressive enough to illustrate the separation of classes, but similar enough to not appear garish and unreadable. I brought the attribute table up and added an Area field set to acres to illustrate the total areas of the classified polygons.
I moved my two images to ArcGIS Pro and created the map you see below:
Comments
Post a Comment