Search

This paper assesses trending AI foundation models, especially emerging computer vision foundation models and their performance in natural landscape feature segmentation. While the term foundation model has quickly garnered interest from the geospatial domain, its definition remains vague. Hence, this paper will first introduce AI foundation models and their defining characteristics. Built upon the tremendous success achieved by Large Language Models (LLMs) as the foundation models for language tasks, this paper discusses the challenges of building foundation models for geospatial artificial intelligence (GeoAI) vision tasks. To evaluate the performance of large AI vision models, especially Meta's Segment Anything Model (SAM), we implemented different instance segmentation pipelines that minimize the changes to SAM to leverage its power as a foundation model. A series of prompt strategies was developed to test SAM's performance regarding its theoretical upper bound of predictive accuracy, zero-shot performance, and domain adaptability through fine-tuning. The analysis used two permafrost feature datasets, ice-wedge polygons and retrogressive thaw slumps because (1) these landform features are more challenging to segment than manmade features due to their complicated formation mechanisms, diverse forms, and vague boundaries; (2) their presence and changes are important indicators for Arctic warming and climate change. The results show that although promising, SAM still has room for improvement to support AI-augmented terrain mapping. The spatial and domain generalizability of this finding is further validated using a more general dataset EuroCrop for agricultural field mapping. Finally, we discuss future research directions that strengthen SAM's applicability in challenging geospatial domains.

Permafrost thaw has been observed at several locations across the pan-Arctic in recent decades, yet the pan-Arctic extent and potential spatial-temporal variations in thaw are poorly constrained. Thawing of ice-rich permafrost can be inferred and quantified with satellite imagery due to the subsequent differential ground subsidence and erosion that in turn affects land surface cover. Information contained within existing and rapidly growing collections of high-resolution satellite imagery (Big Imagery) is here extracted across the Arctic region through a collaboration between software engineers, computer- and earth scientists. More specifically, we are a) developing geospatial data down to sub-meter resolution, and also b) enabling discovery and knowledge-generation through visualization tools. This cyberinfrastructure platform, the Permafrost Discovery Gateway (PDG), is being designed with input from users of the PDG, e.g. primarily the Arctic earth science community but also the general public. The PDG builds upon other NSF supported data management resources (Arctic Data Center and Clowder) and the Fluid Earth Viewer. The Fluid Earth Viewer, which is the first visualization tool implemented into the PDG, was initially created for the public to explore atmospheric and oceanographic visualizations and is here modified to support permafrost geospatial products, and a number of community built analytic tools to identify permafrost artifacts within satellite imagery. The effort also includes workflow optimization of remote sensing code for pan-Arctic sub-meter scale mapping of ice-wedge polygons from optical imagery. We are additionally actively engaging with the user-community to ensure that the PDG becomes useful, both in terms of the type of data contained within the PDG and the design of the visualization tools. The PDG has the potential to fill key Arctic science gaps, such as bridging plot to pan-Arctic scale findings, while also serving as a resource informing

Permafrost thaw has been observed at several locations across the pan-Arctic in recent decades, yet the pan-Arctic extent and potential spatial-temporal variations in thaw are poorly constrained. Thawing of ice-rich permafrost can be inferred and quantified with satellite imagery due to the subsequent differential ground subsidence and erosion that also affects land surface cover, storage and flow of water, sediment, and nutrients. However, a lack of supporting cyberinfrastructure necessary to harness information from the existing and rapidly growing collection of high-resolution satellite imagery (Big Imagery) has limited our advances in understanding the nature of pan-Arctic permafrost degradation. In the coming four years, we will empower the broader Arctic community with a cyberinfrastructure platform, the Permafrost Discovery Gateway (PDG), aimed at making Big Imagery permafrost information accessible and discoverable through novel visualization and analysis tools designed with input from users of the PDG, e.g. the diverse peoples living, working, and/or studying in the Arctic. From the start of the project, we will engage the user-community through in-person and online meetings to ensure effective development of permafrost Big Imagery products for archiving, processing, analyzing, and visualizing. The framework will utilize existing resources, such as the (1) NSF supported data management resources the Arctic Data Center and Clowder, (2) web application visualization tools (Fluid Earth Viewer, Google Earth, and Gapminder Foundation), (3) high performance computing resources (XSEDE, Google Earth Engine etc.), and (4) and satellite imagery (Polar Geospatial Center, Landsat, Sentinel, and Planet). The PDG will include the management of ingesting remote sensing big data into machine and deep learning models. We welcome collaborations with national and international Native, industry, and academic organizations and individuals to ensure broad community engagement an

Traditional methods of detecting and mapping utility poles are inefficient and costly because of the demand for visual interpretation with quality data sources or intense field inspection. The advent of deep learning for object detection provides an opportunity for detecting utility poles from side-view optical images. In this study, we proposed using a deep learning-based method for automatically mapping roadside utility poles with crossarms (UPCs) from Google Street View (GSV) images. The method combines the state-of-the-art DL object detection algorithm (i.e., the RetinaNet object detection algorithm) and a modified brute-force-based line-of-bearing (LOB, a LOB stands for the ray towards the location of the target [UPC at here] from the original location of the sensor [GSV mobile platform]) measurement method to estimate the locations of detected roadside UPCs from GSV. Experimental results indicate that: (1) both the average precision (AP) and the overall accuracy (OA) are around 0.78 when the intersection-over-union (IoU) threshold is greater than 0.3, based on the testing of 500 GSV images with a total number of 937 objects; and (2) around 2.6%, 47%, and 79% of estimated locations of utility poles are within 1 m, 5 m, and 10 m buffer zones, respectively, around the referenced locations of utility poles. In general, this study indicates that even in a complex background, most utility poles can be detected with the use of DL, and the LOB measurement method can estimate the locations of most UPCs., National Science Foundation (U.S.) (grant No. 1414108), Eversource Energy