Advanced Remote Sensing Data Classification Approaches

This session will feature presentations on image classification approaches including AI and machine learning.

10:30 AM – 10:45 AM – Deep Learning Techniques for Building Footprint Creation and Change Detection in Cook County, Illinois, USA

As access to high resolution remote sensing data increases, there is an expectation from end users and stakeholders that actionable information is provided through adjacent analyses performed by curators of these data. However, limitations remain for the processing of these data, including storage and compute requirements, the impact of land use heterogeneity on model performance and standardization of analysis workflows for project use cases. The Cook County Bureau of Technology’s GIS Division has investigated deep learning techniques for building footprint creation and change detection. The processes have been completed with Esri’s deep learning libraries in ArcGIS Pro using multispectral leaf-off imagery and lidar. This presentation will cover end-to-end processes and feasibility of implanting these workflows including necessary data and compute, model development and change detection analysis.

Daniel Bartlett, Cook County Government

Alice Ferruzzi, Cook County Government

10:45 AM – 11:00 AM – Combing UAV LIDAR and imagery data for urban mapping

The focus of this presentation is on integrating optical images and laser point clouds carried on low-cost UAVs to create an automated system capable of generating urban city models. After pre-processing both datasets, we co-registered both datasets using the DLT transformation model. We estimated structure heights from the LiDAR dataset through a progressive morphological filter followed by removing bare ground. Unsupervised and supervised image classification techniques were applied to a six-band image created from the optical and LiDAR datasets. After finding building footprints, we traced their edges, outlined their borderlines, and identified their geometric boundaries through several image processing and rule-based feature identification algorithms. Comparison between manually digitized and automatically extracted buildings showed a detection rate of about 92.3% with an average of 7.4% falsely identified areas with the six-band image in contrast to classifying only the RGB image that detected about 63.2% of the building pixels with 25.3% pixels incorrectly identified. Moreover, our building detection rate with the 6-band image was superior to that attained by performing traditional image segmentation for only the LiDAR DEM. Shifts in the horizontal coordinates between corner points identified by a human operator and those detected by the proposed system were in the range of 10-15 cm. This is an improvement over traditional satellite and manned-aerial large mapping systems that have lower accuracies due to sensor limitations and platform altitude. These findings demonstrate the benefits of fusing multiple UAV remote sensing datasets over utilizing a single dataset for urban area mapping and 3D city modeling. 

Ahmed Elaksher, New Mexico State University

11:00 AM – 11:15 AM – Multimodal Remote Sensing and AI for Real-Time Wildfire Risk Assessment and Decision Support

Wildfires represent one of the most pressing environmental and societal challenges of our time, with increasing frequency, intensity, and destructiveness driven by climate change, prolonged droughts, and rapid urbanization into wildland-urban interface (WUI) zones. In the United States alone, wildfires cause between $394 billion and $893 billion in annual economic losses, including property damage, suppression costs, and long-term health and ecological impacts. Traditional wildfire management systems are limited, and their shortcomings result in delayed responses, inefficient resource allocation, and heightened risks to communities and infrastructure. FireVision, an innovative AI-powered decision-support platform developed by Opal AI, leverages advanced Earth Science data and enables proactive wildfire risk assessment, mitigation planning, and real-time response. At its foundation, FireVision employs a multimodal data fusion approach, integrating diverse remote sensing sources such as Synthetic Aperture Radar (SAR), optical and infrared imagery, alongside topographic and real-time weather forecasts. This integration produces quasi-real-time, high-resolution fuel maps at scales less than 30 meters and with weekly updates that capture dynamic changes in vegetation, moisture levels, and fuel loads. The platform’s core innovation lies in its hybrid analytical-machine learning framework, which computes dynamic Fire Potential Indices (FPIs). These indices forecast wildfire ignition and spread probabilities with up to 90% accuracy over 48-hour horizons. Key geospatial and technological components we offer include multimodal data fusion for superior accuracy, AI transforming Earth observation accessibility, and WUI-focused vulnerability and mitigation tools. FireVision not only enhances operational efficiency for fire agencies, utilities, and emergency responders but also fosters interagency collaboration by providing a shared platform for data-driven insights. By emphasizing AI’s role in transforming raw Earth observation data into decision-ready intelligence and keeping humans in the loop, we will equip geospatial professionals, policymakers, and practitioners with tools to build more resilient communities against wildfire threats.

Ryan Alimo, Opal AI Inc.

11:15 AM – 11:30 AM – Leveraging Advanced Geospatial Technologies to Support Lake Lure Recovery and Resilience After Hurricane Helene

This presentation highlights the strategic use of advanced technologies in the ongoing recovery and maintenance efforts at Lake Lure following Hurricane Helene. A comprehensive data collection initiative included high-resolution shoreline imagery and video acquired via boat-mounted systems, with results hosted in ArcGIS Online Map Viewer for intuitive access and visualization. Aerial lidar and imagery were captured to support a detailed topographic survey extending from the water’s edge to the 995-foot contour line. Conventional surveying was performed for subdivision and easement platting, and a hydrographic survey was completed to assess underwater conditions. By integrating these technologies, the project team achieved a highly accurate, efficient, and reliable assessment of storm impacts. This approach enabled early risk identification, informed mitigation planning, and ultimately helped protect critical infrastructure and ensure public safety.

Robert Crawshaw, McKim & Creed

11:30 AM – 11:45 AM – Benthic Habitat Mapping with Emphasis on SAV on the East Coast of Canada using Topo-bathymetric Lidar

Airborne topo-bathymetric lidar (TB-lidar) has traditionally been used to update nautical charts and map shallow water hazards to navigation. More recently researchers have used TB-lidar data to map the nearshore benthic environment with emphasis on submerged aquatic vegetation (SAV), which is challenging to map using traditional remote sensing methods. Seaweed is recognized as a valuable commodity with uses ranging from blue carbon storage to benefits for plant additives (=fertilizers), pharmaceuticals and animal-human food. The Leica Chiroptera shallow water system has been used to estimate the biomass of rockweed, an intertidal seaweed, as well map seagrass and most recently map subtidal kelp beds off the east coast of Canada in Nova Scotia and Newfoundland. In addition to these airborne surveys, ground-truth collection in the form of underwater photos-videos and echo-sounding have been used to verify the interpretation of the lidar and aerial photo data acquired. The TB-lidar point cloud was classified into water-surface, bathymetry and submerged aquatic vegetation and ground and non-ground features (builds and vegetation, etc.). The lidar intensity of the green laser (515 nm) returns of the bathymetry provides insights into the composition of the seabed. The lidar intensity raster along with bathymetry, seabed roughness and the RGB imagery are used to produce benthic habitat maps including the distribution of wild kelp. The lidar point cloud is able to separate kelp above the seabed to allow an estimate of vegetation thickness which in turn was used to estimate biomass. Details are not sufficient to separate kelp species at this point, however. These surveys provide information on wild kelp distribution, health and habitat and can provide areas for potential harvest or where broodstock could be harvested for to supply seed for local kelp aquaculture operations.

Tim Webster, Applied Geomatics Research Group, Nova Scotia Community College

11:45 AM – 12:00 PM – Detroit Tree Census – A New Perspective on the Urban Forest

Urban trees are essential for the functioning of a modern city. Understanding and managing the tree canopy can be both expensive and time consuming. In most cases only street tree and trees on the local government land are measured but this only makes up part of the overall tree resource. To get both an understanding of the extent, diversity, health and distribution of the City of Detroit’s tree resource, The Nature Conservatory (TNC) funded a project to map all the trees within the City’s boundaries to support planting and management of the urban tree canopy by The Greening of Detroit (TGD) and the General Services Department (GSD) of the City of Detroit. 

Currently the City of Detroit has a tree canopy cover of around 26% and has a goal of increasing this Citywide to around 40%. To achieve these goals planting will be required on both public and private lands, so understanding the tree resource and opportunities for tree planting across all ownerships is important for both the management and the development of tree planting programs for public and private lands within the City boundaries, and across the region. 

The project leverages existing high resolution lidar dataset that was flown in 2021, tree inventory data, and 2025 summer imagery to create a wall-to-wall tree census of the City. This will result in information about every dominant tree within the boundaries of the City. These data will be integrated into enterprise systems for tree management and utilized for tree management and community engagement. 

The presentation will review the approach and results of the projects, and how tree census data will change the way we understand the urban forest and manage it. 

Andrew Brenner, NV5