Katherine Avery: Automated Location of Bird Roosts using NEXRAD Data and Image Segmentation

Abstract

Weather surveillance radars can effectively detect flying animals, such as groups of birds, bats, and insects. Further, these radars are demonstrably useful for detecting the existence of certain bird roosting locations, particularly those of many species of swallows. The roosts appear on radar as distinctive rings of high reflectivity. Detecting and locating bird roosts have a variety of applications from ecological conservation to wind farm placement and air traffic control.

In this thesis, I first detect the presence of a roost in NEXt Generation Weather RADar (NEXRAD) images of purple martin and tree swallow roosts and improve upon Chilson et al. (2019)'s work by altering the network architecture and data preprocessing. Because determining the exact locations of bird roosts is more useful than detection, I also use image segmentation to attempt to locate roosts. I apply a standard U-Net, which shows promising results for determining roost location, achieving a true positive rate of around 0.80 and true negative rate of around 0.85. To increase performance, the dataset is augmented 32 times its original size through a variety of image transformations.

Thesis

Katherine Avery (2020):Automated Location of Bird Roosts using NEXRAD Data and Image Segmentation. Master's Thesis, School of Computer Science, University of Oklahoma.

• Link to ShareOK copy
• Thesis (local copy)

Related publications and presentations

• Chilson, Carmen; Avery, Katherine; McGovern, Amy; Bridge, Eli; Sheldon, Daniel and Kelly, Jeffrey (2018) Automated Detection of Bird Roosts using NEXRAD Radar Data and Convolutional Neural Networks. Remote Sensing in Ecology and Conservation. [https://doi.org/10.1002/rse2.92]

Code

The code for the thesis can primarily be found in https://github.com/djgagne/hagelslag