Image Classification

I used ArcGIS Pro 3.0.2 to solve this problem. The main geoprocessing and image classification tools that I used include “Training Samples Manager”, “Classify”, “Build Raster Attributes Table”, “Add Field”, and “Calculate Field”. The data to solve this problem include an aerial photograph of the Black Water National Wildlife Refuge. The data are provided by the class.

Part I. I first added the raster data of the aerial photo into ArcGIS. Then I used the “Training Samples Manager” to draw polygons on the map as training samples. I created two training samples for each land cover type. Then I used “Classify” under image classification to perform supervised classification. Then I used the “Build Raster Attributes Table”, “Add Field”, and “Calculate Field” tools to calculate the total area of each land cover type. Then I compared the output raster with the aerial photo to identify any areas of misclassification. Then I changed the symbology of the sample polygons and output raster data to create maps.

Part II. After identifying areas that need improvement, such as Water being misidentified as Developed and Developed misidentified as Wetlands. I created additional training samples for land cover classes that need improvement in classification. I added 4 additional samples for Water, 1 for Developed, 2 for Forest, 12 for Cultivated, and 5 for Wetlands. I then performed supervised classification again with more training samples using “Classify”. Then I used the “Build Raster Attributes Table”, “Add Field”, and “Calculate Field” tools to calculate the total area of each land cover type. Then I compared the output raster with the aerial photo and the previous raster output to check for improvements in classification and misclassified areas. Then I updated the symbology of the new training samples and raster output to make maps.

Image classification is a very useful tool for processing aerial photo data, and the process can be improved by adding representative training samples. A possible scenario could be that an urban planner in Raleigh is trying to decide where to add new green space in South Raleigh, but he only has an aerial photo of South Raleigh.

Problem description: As a GIS specialist, I am asked to perform an Image Classification to identify suitable land for developing open space.

Data needed: Aerial photo of South Raleigh.

Analysis procedures: I will perform supervised classification to transfer the aerial photo into raster data with each land cover type identified. I will create training samples for Forests, Developed, and Cultivated land cover. Then I will use “Classify” to perform image classification. The output raster can be used to identify potential new green space.