Please click on the poster thumbnail above to enlarge the image.
Audio Presentation
Link will open in a new tab.
External Poster Media Link
Link will open in a new tab.
Ultra-High-Resolution Mapping of River Habitat and Geomorphic Features using UAVs, Structure from Motion, and Machine Learning
Abstract:
<p>For more than a century, the Los Angeles River has faced extensive environmental degradation due to urbanization, flood control measures, and other human activities. Over the last couple of decades, public interest in restoring some aspects of the river, while maintaining flood protection for adjacent communities, has grown steadily. The success of future ecosystem restoration and management efforts will hinge on our understanding of how alterations in flow can impact habitat resilience, which is constrained by the costs and time required to collect high quality data. To address these challenges and align with the growing public interest in its revitalization, we used a compact RGB camera, mounted on a consumer-grade UAV, to capture ultra-high-resolution (GSD = 1.6 cm) images of a 2-mile, soft-bottom reach of the LA River. Leveraging SfM, focal statistics and multiple RGB vegetation indices, we generated detailed topographic and vegetation height data, image texture information, and greenness measures. Further, we utilized principal component analysis (PCA) for data dimension reduction and employed machine learning, specifically support vector machines (SVM), to classify land cover based on these multi-modal data sources. Our approach achieved an overall classification accuracy of 83% (kappa = 80%). The generated land cover classification maps provide a comprehensive view of the Los Angeles River's habitat and geomorphic features, improving on previous efforts in terms of accuracy and detail. In the future, products generated from this approach could be instrumental in informing and guiding future management decisions that balance ecological restoration with flood risk management needs.</p>
Keywords: remote sensing, GIS, UAVs, drones, habitat mapping, urban river, ecosystem restoration and management, machine learning
Authors:
Michael Beland, Department of Geography, Geology, and Environment, California State University, Los Angeles; Submitting Author / Primary Presenter
Chris Guo, Department of Earth System Science, University of California, Irvine; Co-Author (this author will not present)
Alex Purdom, Department of Geosciences, Middle Tennessee State University; Co-Author (this author will not present)
Alireza Farahmand, Department of Geography, Geology, and Environment, California State University, Los Angeles; Co-Author (this author will not present)
, {SubmissionForm.presentersAuthors/affiliation#4};
, ;
, ;
, ;
, ;
, ;
, ;
Ultra-High-Resolution Mapping of River Habitat and Geomorphic Features using UAVs, Structure from Motion, and Machine Learning
Category
In-Person Paper Abstract