Image Classification

Problem:

As a GIS Analyst for Black Water National Wildlife Refuge, I would like to develop a management plan that has to be applied to the entire refuge aerial photography. That is why I need to perform a supervised classification using one aerial photograph from Black Water National Refuge. The aerial photograph is a 4-band(true color and color-infrared) 1 foot resolution image flown August 2010.

There are 6 main land cover classes in this aerial photograph like Water, Cultivated land, Forest, Developed area, Barren land and Wetland.

Analysis Procedure:

A true color composite image like this image is the combination of different bands like red, green and blue. Imagery is made up of Grid of numbers that are arranged into rows and columns. Image Classification refers to the task of extracting information classes from a multiband raster image. The resulting raster from image classification can be used to make a thematic map. In order to get better looking vegetation feature in the image, infrared band provides increased information. That’s why Infrared band is substituted for red band to get better information. Therefore, 4 bands have been provided through moodle of NCSU GIS 520 course. Band 4 represents as Red, Band 3 represents as green and band 2 represents as blue color. After changing the color as red, green and blue in symbology, 3 bands had to be combined in a composite band. There are two types of image classification, supervised and unsupervised classification. In this case, supervised classification had to be performed for this image classification. Arc GIS 10.4 has been used for solving the problem.

Image classification from toolbar from Customize menu and spatial analyst Extension from Customize menu should have been checked. In order to classify different classes like forest, water, developed, cultivated, barren and wetland, at least 2 -3 polygons per class had to be created on the composite image. As an analyst making new polygons of known pixels to generate representative parameters for each class of interest which step is called training. By using Training sample manager, at least 2-3 polygons per class had to draw first, then by using merge training sample icon, all 2-3 polygons had to be merged into one same class by using same color as referred to the National Land cover classification legend. For all 6 classes same procedure had to be used. The first sample polygon map is shown in map no 1.

After training, the classification method is then used to attach labels to all the image according to the trained parameters. Next by saving the first training sample as a shape file, interactive supervised classification from classification menu of Image classification toolbar had to be used and run to get the image classification result. Then all classes had been shown in the map according to land cover classification legend. The classified image is shown in map no. 2.

                              Workflow Diagram

Results:

Application and Reflection:

This image classification method is very useful for any Landsat or any remote sensing raster image to inquire the land use, metrology, wildlife habitat as well crops, forestry analysis or any natural calamities’ analysis. As California forest fire prevention act analyst, I will use Landsat map to visualize active fire detection image, possible cause of forest fire and how to prevent forest fire from that area. In order to act as a preventive measure, I need to study latest image of infrared Landsat image of forest fire zone of California, then by making training sample polygons on the map by using Arc GIS, using supervised image classification method, it can be quantitatively classified as what kind of forest or vegetation are there that are easily burned down. Is there any fire station or developed area nearby that facilitates the fire prevention act very quickly.

Back to top