Image Classification

As a GIS Analyst for Black Water National Wildlife Refuge, I was asked to perform supervised image classification using an aerial photograph of the refuge to obtain land cover information. Developing a methodology to be applied to the entire refuge aerial photography will help refuge managers develop a management plan. The aerial photograph was taken in June 2017 and is a 4-band (true-color and colorinfrared) image. As part of this assignment, I was required to perform supervised classification using at least two sample polygons per each (seven) Land Cover Class. Then, use a redefined training sample set by adding additional polygons to the training sample set from the first round for the land cover classes I wished to improve to improve the classification output. The learning objectives were to gain an understanding of the general procedures for image classification using the unsupervised image analysis process to classify land use and land cover

Strategies: For this problem, I used ESRI's ArcGIS Pro 3,0 along with tools such as Build raster attribute table after creating a training sample and using the classify tools. All data required for the assignment are provided in a zip file (m_3807640cr.zip) through the theme block and includes an aerial photograph named m_3807640cr.tif (file size about 22 MB). Supervised classification requires analyst-specified training data (i.e., samples based on land cover types) to develop information classes.

Methods: To complete this assignment first I created a schema, then began making training samples (i.e., sample polygons that represent each land cover class). With the subset image displayed in a viewer, the ArcGIS Pro AOI tools was used to create polygons and delineate homogeneous areas representing the categories of interest. This is accomplished by using the Polygon tool to create polygons for homogeneous areas. Once I added multiple training samples for my first class (i.e., Water), next I clicked on another class (i.e., Urban) to begin added training samples that class. Next, I clicked the Imagery Tab, Classification Tools within the Image Classification group I selected Classify. The Image Classification pane opens, Classifier was set to Support Vector Machine. For Training Samples, I added the training samples created previously Output Classified Dataset and Output Classifier Definition File (.ecd) were left default. The final step of this assignment was to use build raster attribute table tool to build an attribute table for our classified data

Process diagram

In this assignment I gained an understanding of the general procedures for image classification using the supervised image analysis process to classify land use and land cover. I can use the skills I learned in completing this assignment to classify Raleigh, NC based on 6 land cover classes such as water, developed, forest, herbaceous etc.…

New problem statement: Classify Raleigh, NC based on 6 land cover classes such as water, developed, forest, herbaceous etc..

Data Needed: Orthophoto of Raleigh from USGS website.

Analysis procedure: create a training sample using Aoi tools, Save training sample to my files then finally use classify tool with my saved training sample as an input.