Classification Matrices 

This project was completed for the Applied Remote Sensing course at Utah State University. Seen to the left is a classified land cover image creates using training points collected by Professor Doug Ramsey. Using a smile random forest classification tree pixels were categorized into the landcover types represented in the training points based on their spectral characteristics. 

To test the reliability of the classification a confusion matrix was generated, a confusion matrix shows how training points were classified through the tree compared to their true landcover types. By analyzing the confusion matrix it was discovered that the category 'Deciduous Tree' was the most likely to be wrongly categorized and was almost always mistaken for 'Gambel Oak' this is likely because of their very similar spectral characteristics. Using the Confusion Matrix the overall accuracy of the classification was calculated at 93%. While the Confusion Matrix is visible in the Console tab of the Google Earth Engine Code associated with this project, a more readable version is pictured below.