Authors: Rhett Mohler, Professor at Saginaw Valley State University, Jarod Morse*, Saginaw Valley State University
Topics: UAS / UAV, Environment, Remote Sensing
Keywords: phragmites, invasive, wetland, uav
Session Type: Virtual Paper
Start / End Time: 11:10 AM / 12:25 PM
Room: Virtual 18
Presentation File: No File Uploaded
Unmanned Aerial Vehicles (UAVs) have become crucial and effective tools for mapping, identifying, and controlling invasive plant species across the globe. One wetland plant that can be pinpointed with the use of UAV imagery is the invasive reed, Phragmites australis. The UAV imagery for this study was collected on 2 different dates, August 11th and September 20th, using an RGB camera with a pixel size of 5.04 centimeters and a flight altitude of 375 feet. The UAV images were preprocessed to produce orthorectified images and digital surface models (DSM). This study evaluated the detection of P. australis using 3 different classifiers; Maximum Likelihood Classifier (MLC), Support Vector Machine (SVM), Neural Network. We classified two different images, a 4-band image with the RGB and DSM from the August data and a 7-band image with the RGB and DSM from August and the RGB bands from the September data. The accuracy analysis was achieved by an error matrix. The two classifiers with the highest overall accuracy were the SVM and MLC with values of 0.7955 and 0.7444 respectively. The true focus of the study was of the classification of P. australis. The errors of omission and commission for the SVM were 0.1412 and 0.1098 and the MLC values were 0.0810 and 0.1707. Even though MLC is considered a basic and widely available classification, it has the ability to effectively classify P. australis and can be useful to managers who may not have the software or experience in remote sensing.