Use of sUAS imagery for surveying waterfowls in a managed wetland in Colusa County

Authors: Kaitlyn Hernandez*, Humboldt State University, Judson Fisher, Humboldt State University, Katherine Marlin, Humboldt State University, Alex Pickering, Humboldt State University, Luke Scaroni, Humboldt State University, Ariel Weisgrau, Humboldt State University, Buddhika Madurapperuma, Humboldt State University, Sharon Kahara, Humboldt State University, James Lamping, Humboldt State University
Topics: Remote Sensing, Drones, Environmental Science
Keywords: sUAS imagery, waterfowl survey, wetland, OBIA
Session Type: Virtual Poster
Day: 4/10/2021
Start / End Time: 3:05 PM / 4:20 PM
Room: Virtual 52
Presentation File: Download



Recent advances in small unmanned aerial systems (sUAS) have facilitated monitoring and counting waterfowl using object based image analysis (OBIA) in remote sensing. The objective of the study is to use a semi-automated workflow to extract waterfowls from a managed wetland in Colusa County, California. Over 560 sUAS imagery was obtained using a DJI Mavic 2 PRO at an average Ground Sample Distance of 3 cm/px. Nine ground control points were placed across the study area and the coordinates were recorded using a Real-Time Kinematic GPS. An orthomosaic image was created using the Agisoft Photoscan software and the image was smoothed using a low pass filter to prevent over segmentation. Training points of waterfowl were manually created and then ENVI Segmentation only workflow was used to extract waterfowl objects, using the Edge algorithm at a scale of 75% and merge algorithm at a level of 95%. Two subsets of waterfowl present (6.8 ha) and waterfowl absent (1.4 ha) were used for OBIA. Rule-based feature extraction workflow in ENVI was used to classify two data subsets. The total automated waterfowl count was 2,259. The overall classification accuracy for identifying birds was 57.3%. The user's accuracy for birds and non-birds was 93.9% and 51.5% and producer’s accuracy for birds and non-birds was 23.6% and 98.1% respectively. The greatest misclassification had visually similar grass patches in shallow water, or areas without birds. Conducting automated and manual counts in defined habitat may overcome the challenge.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login