Search USGS⌕ Search
Home/Archive/USGS

USGS · 70258364

Automated deep learning-based point cloud classification on USGS 3DEP lidar data using transformer

Jung-Kuan Liu·Rongjun Qin·Shuang Song
View original source ↗Export metadata ↓

Also available from

Abstract

The goal of the U.S. Geological Survey’s (USGS) 3D Elevation Program (3DEP) is to facilitate the acquisition of nationwide lidar data. Although data meet USGS lidar specifications, some point cloud tiles include noisy and incorrectly classified points. The enhanced accuracy of classified point clouds can improve support for many downstream applications such as hydrologic analysis, urban planning, and forest management. Despite noisy and incorrectly classified points, the current 3DEP classification specifications result in data that can be useful for Digital Terrain Model (DTM) extraction; however, the quality of the classification application can be improved to match state-of-the-art capabilities. Deep Learning (DL)-based approaches have been developed with outstanding performance for point cloud classification. This study will utilize the proven DL technologies to prepare for developing a user-friendly open-source toolkit that would automate classification to refine and enrich the results of existing and future 3DEP data.

Explore related subjects

Conference Paper

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Jung-Kuan Liu, Rongjun Qin, Shuang Song. 2024. Automated deep learning-based point cloud classification on USGS 3DEP lidar data using transformer. https://doi.org/10.1109/igarss53475.2024.10641055

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

Authoritative knowledge infrastructure for GeoAI: Implications for scientific spatial data providers

Geospatial Artificial Intelligence (GeoAI) increasingly relies on automated analysis of authoritative geospatial data, exposing limitations in data infrastructures originally designed for simple dissemination rather than for knowledge-driven use. From the perspective of scientific spatial data providers, the challenge is enabling GeoAI responsibly while retaining semantic clarity, provenance, reproducibility, and governance. This short paper (i) examines how GeoAI exposed limitations within existing geospatial data infrastructures, and (ii) identifies provider-side requirements for an authoritative knowledge infrastructure. Drawing on GIScience, scientific workflow, and operational AI literature, the paper outlines implications for national mapping organizations that seek to support advanced analytics without jeopardizing scientific rigor or stewardship responsibilities.

Conference Paper↗
USGS2026

Springs of the Catskill Mountains: A conduit flow system in predominantly siliciclastic strata

The Catskill Mountains of New York state are composed of thick sequences of interbedded mudstone, sandstone, and siltstone that are intercalated with thin beds of paludal limestone and calcareous mudstone. Sandstones often contain rip-up clasts of calcareous material at their bases and contain minor amounts of carbonate cement throughout their mass. This minor amount of calcareous material weathers out to produce thin zones of solutional porosity and permeability integrated within perched aquifers that feed perennial hillslope springs. Solutional features are evident within thin carbonate lenses and along fractures in sandstones, occasionally resulting in explorable cave-sized conduits. This study focuses on the behavior of bedrock springs and seeps in the headwaters of the Neversink River basin above the Neversink Reservoir. Bedrock springs are numerous throughout the Catskill Mountains and serve to provide baseflow to the headwater streams that tends to be more persistent during dry times than other baseflow sources. We observed dissolution-influenced conduits within the bedrock, and suffosional conduits within unconsolidated materials. Continuous water temperature at six springs was monitored at hourly intervals from May to October 2022 to investigate the thermal response of groundwater to daily and seasonal changes in air temperature along with event-based meteorological events. Temperature monitoring was augmented at some springs with time-lapse photography to evaluate persistence of flow. The thermal responses of springs to precipitation events were highly variable across sites, with some springs exhibiting flashy responses indicating a strong connection to the surface, while other springs showed muted or no thermal response to events. The seasonal thermal response also varied across sites having lag times of 23-48 days between peak air temperature and peak spring water temperature. The water temperatures and geochemical compositions of these bedrock springs support a conceptual model of multiple residence times, showing both event and seasonal groundwater storage within the conduit flow system that feeds the springs of the Catskill Mountains.

New York↗
USGS2026

Stability changes after weak layer collapse: Insights from an unexpected dataset

While conducting avalanche mitigation or traveling through the backcountry, avalanche professionals sometimes experience a "whumpf": a collapse-mode crack accompanied by surface cracking that does not produce an avalanche. These whumpfs are a clear warning sign of instability in the surrounding terrain. But assessing the collapsed slope itself, if it is steep enough to avalanche, carries much more uncertainty. Can slopes that whumpf and are steep enough to slide still avalanche? The answer to this question supports avalanche mitigators and ski guides in assessing whether to open such slopes. Previous research on collapsed slopes is limited to one study conducted before tests that estimate crack propagation propensity, such as the Propagation Saw Test (PST), were developed. This work showed that weak layer shear strength decreased immediately after collapsing and then increased over time. However, no research has examined crack propagation propensity after a whumpf. Studying propagation propensity is key for assessing how whumpfs immediately affect slope stability. To improve our understanding of changes in propagation propensity following weak layer collapse, we conducted PSTs on a test slope in the hours and days after an unexpected collapse occurred on a layer of surface hoar. Two PSTs at our study site three days prior to this collapse provided an estimate of weak layer propagation propensity. PST cut lengths decreased immediately following the collapse, suggesting that the collapse increased the propagation propensity. However, over the next 4-5 hours, PST cut lengths steadily increased at our site, likely due to ongoing sintering and strengthening in the collapsed weak layer. By the end of day two, we were no longer able to initiate propagation, suggesting substantial stabilization of the weak layer. Our results support prior findings and observations suggesting that slopes may remain capable of producing avalanches immediately after a collapse but likely stabilize over the next day or two. Although this research does not provide answers about precisely when a slope becomes stable after a collapse, it does highlight uncertainties in our understanding and interpretation of the stability of slopes that have whumped.

Montana↗