GIS 4035 - Remote Sensing and Photo Interpretation - Student Spotlight

Module 10: Supervised Classification

The blog postings for Week 10 illustrated that a number of students had a good understanding of how to use Erdas Imagine to perform a supervised classification on multispectral satellite imagery.
However, Ashlee Malone's blog was especially good and stood out from the rest.  This week, we would like to highlight her excellent work!

In this lab, students were instructed to create spectral signatures and AOI features, produce classified images from satellite data, and recognize and eliminate spectral confusion between spectral signatures.  

Ashlee's blog description was well written, easy to read and outlined all the steps . She covered all the crucial steps required to conduct a supervised image classification in Erdas Imagine, including the development of training sites (using AOI layers), evaluating the training sites to limit the amount of spectral confusion, and ultimately the choice of spectral bands to include. She also included  an excellent description of both (a) the spectral distance file, which can be used to evaluate the accuracy of the classified image, and (b) the chosen classification method (maximum likelihood).  In addition to her well written blog description, Ashlee's map was well designed and easy to interpret (we especially liked her color choices for the different LULC classes). She also included the distance image itself as an inset so readers can evaluate the effectiveness of her classification. Her resulting classified image, was also once of the best we have seen. Differentiating "roads" and "urban" from using images with this level of spatial resolution (30 meters) is very difficult but her final map does an excellent job differentiating these often confused classes.  Overall, Ashlee's blog posting was excellent!

GIS 4035 - Remote Sensing & Photo Interpretation - Student Spotlight

Module 3 - Land Use/Land Cover (LULC) Classification 

For Week 4 in GIS 4043, students were instructed to locate and identify features using Google Maps street view, construct an unbiased sampling system, and calculate the accuracy of Land Use/Land Cover classification maps.  

The blog postings for Week 4 illustrated that a number of students had a good understanding of how to conduct an accuracy assessment on a LULC classification. However, this week, we would especially like to highlight Daniela Sabillion!

Her blog post is easy to understand, provides an excellent overview of the accuracy assessment techniques learned in this week's Module, and demonstrates a high level of understanding of the material. Daniela provides a brief review of the two main types of accuracy assessment: in-situ and ex-situ.  Despite not having the highest accuracy (which has a lot to do with the differences in scale of the imagery vs the scale of the ground truthing), her blog post provides a good description on how she developed a stratified random sample using a fishnet grid created in ArcGIS. In addition to the description, she also provided graphics of both (a) different sampling methods and (b) her table of sampling points with accuracy assessment (yes or no) with notes. These additional graphics greatly assist with interpreting the process she used for conducting the accuracy assessment.  

In addition, her map demonstrates a keen design. It is easy to read, the map elements are extremely well balanced, and the colors she chose make it very easy to distinguish the different LULC classes.  Overall, Daniella's blog posting was very good!