Kamm, Matthew; Reed, J Michael (2018): Confusion matrices evaluating the accuracy of supervised classification of habitat types in UAV aerial photos around American kestrel nest sites in Massachusetts, USA [dataset]. Tufts University, PANGAEA, https://doi.org/10.1594/PANGAEA.884660,

---

In: Kamm, M; Reed, JM (2018): Assessment of classification accuracy of habitat types in UAV aerial photos around American kestrel nest box sites in Massachusetts, USA [dataset publication series]. PANGAEA, https://doi.org/10.1594/PANGAEA.884669

Confusion matrices generated by program ENVI to evaluate the accuracy of Supervised Classification via a Maximum Likelihood method. Each of 12 sites was photographed at 25m and 50m heights by a Phantom 2 Vision+ quadcopter drone. Each 50m photo was also cropped to the same field of view as the 25m photo in order to examine effects of changes in image resolution with altitude. At 25m and 50m heights, different final image resolutions (kernel sizes, in pixels) were also recorded to compare. Each image was classified into a maximum of five different cover types, and the number of pixels correctly and incorrectly assigned to each cover category is recorded in the confusion matrix.