Geomorphological Analysis Using Unpiloted Aircraft Systems, Structure from Motion, and Deep Learning

Chen, Zhiang; Scott, Tyler R.; Bearman, Sarah Kate; Anand, Harish; Keating, Devin; Scott, Chelsea; Arrowsmith, J Ramón; Das, Jnaneshwar

doi:10.1109/iros45743.2020.9341354

Cited by 15 publications

(18 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This spatially continuous dataset can help NWS WFOs and emergency managers with detailed damage assessments. While our work has investigated DL approaches for tornado damage estimation using 2D visible orthomosaics, the Mask R-CNN architecture (based on 2D CNNs) can be easily extended to other applications with additional information (e.g., multispectral orthomosaics, digital elevation models), as demonstrated by Chen et al [35].…”

Section: Discussionmentioning

confidence: 99%

“…Following the data preparation process in the UAS-SfM-DL pipeline [35], orthomosaics were split into 2000 × 2000 pixel image tiles such that the average residential building coverage (the ratio of a residential building area to a tile area) was about 10.3%, considering the average house size in Kansas of 165.6 square meters. The size of 2000 × 2000 pixels, however, still required large neural network models (parameter number) and GPU global memory.…”

Section: Data Preparationmentioning

confidence: 99%

“…This study adopted transfer learning because of its advances demonstrated in the previous work [35]. The weights of the Mask R-CNN were initialized with a model welltrained from COCO 2017 [42], except that the last layer was modified to accommodate the class numbers in this study.…”

Section: Neural Network Trainingmentioning

confidence: 99%

See 2 more Smart Citations

Data-Driven Approaches for Tornado Damage Estimation with Unpiloted Aerial Systems

et al. 2021

Self Cite

View full text Add to dashboard Cite

Tornado damage estimation is important for providing insights into tornado studies and assisting rapid disaster response. However, it is challenging to precisely estimate tornado damage because of the large volumes of perishable data. This study presents data-driven approaches to tornado damage estimation using imagery collected from Unpiloted Aerial Systems (UASs) following the 26 June 2018 Eureka Kansas tornado. High-resolution orthomosaics were generated from Structure from Motion (SfM). We applied deep neural networks (DNNs) on the orthomosaics to estimate tornado damage and assessed their performance in four scenarios: (1) object detection with binary categories, (2) object detection with multiple categories, (3) image classification with binary categories, and (4) image classification with multiple categories. Additionally, two types of tornado damage heatmaps were generated. By directly stitching the resulting image tiles from the DNN inference, we produced the first type of tornado damage heatmaps where damage estimates are accurately georeferenced. We also presented a Gaussian process (GP) regression model to build the second type of tornado damage heatmap (a spatially continuous tornado damage heatmap) by merging the first type of object detection and image classification heatmaps. The GP regression results were assessed with ground-truth annotations and National Weather Service (NWS) ground surveys. This detailed information can help NWS Weather Forecast Offices and emergency managers with their damage assessments and better inform disaster response and recovery.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Data Preparationmentioning

confidence: 99%

See 1 more Smart Citation

Data-Driven Approaches for Tornado Damage Estimation with Unpiloted Aerial Systems

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…The last option would be to utilize recent advancements in deep-learning object detection algorithms like convolutional neural networks (CNNs). CNNs have been proven as valuable object detectors for a variety of applications, including detecting sea scallops from benthic imagery [32]; detecting object signatures from ground penetrating radar [33]; detecting archaeological sites from LiDAR DEMs [34]; detecting ice-wedge polygons from aerial imagery [35]; detecting rocks from aerial imagery [36]; detecting mining-related valley fill faces from LiDAR DEMs [37]; and detecting airplanes, tennis courts, basketball courts, baseball diamonds and vehicles from aerial imagery [38]. CNNs are also fast, with models like Yolo and Faster R-CNN that can For the selection of the right detection method for Carolina Bays, one might first try some traditional image processing techniques, including Hough transforms, blob detectors like the Laplacian of Gaussian or difference of Gaussians, or feature detectors like the scale-invariant feature transform (SIFT).…”

Section: Traditional Computer Vision Pixel-based Classification and Object Detectionmentioning

confidence: 99%

“…CNNs have been proven as valuable object detectors for a variety of applications, including detecting sea scallops from benthic imagery [32]; detecting object signatures from ground penetrating radar [33]; detecting archaeological sites from LiDAR DEMs [34]; detecting ice-wedge polygons from aerial imagery [35]; detecting rocks from aerial imagery [36]; detecting mining-related valley fill faces from LiDAR DEMs [37]; and detecting airplanes, tennis courts, basketball courts, baseball diamonds and vehicles from aerial imagery [38]. CNNs are also fast, with models like Yolo and Faster R-CNN that can run through and detect objects in several images per second, up to 65 frames per second for Yolov4 on a Tesla V100 graphical processing unit (GPU) [39].…”

Section: Traditional Computer Vision Pixel-based Classification and Object Detectionmentioning

confidence: 99%

Using Convolutional Neural Networks for Detection and Morphometric Analysis of Carolina Bays from Publicly Available Digital Elevation Models

Lundine

Trembanis

2021

Remote Sensing

View full text Add to dashboard Cite

Carolina Bays are oriented and sandy-rimmed depressions that are ubiquitous throughout the Atlantic Coastal Plain (ACP). Their origin has been a highly debated topic since the 1800s and remains unsolved. Past population estimates of Carolina Bays have varied vastly, ranging between as few as 10,000 to as many as 500,000. With such a large uncertainty around the actual population size, mapping these enigmatic features is a problem that requires an automated detection scheme. Using publicly available LiDAR-derived digital elevation models (DEMs) of the ACP as training images, various types of convolutional neural networks (CNNs) were trained to detect Carolina bays. The detection results were assessed for accuracy and scalability, as well as analyzed for various morphologic, land-use and land cover, and hydrologic characteristics. Overall, the detector found over 23,000 Carolina Bays from southern New Jersey to northern Florida, with highest densities along interfluves. Carolina Bays in Delmarva were found to be smaller and shallower than Bays in the southeastern ACP. At least a third of Carolina Bays have been converted to agricultural lands and almost half of all Carolina Bays are forested. Few Carolina Bays are classified as open water basins, yet almost all of the detected Bays were within 2 km of a water body. In addition, field investigations based upon detection results were performed to describe the sedimentology of Carolina Bays. Sedimentological investigations showed that Bays typically have 1.5 m to 2.5 m thick sand rims that show a gradient in texture, with coarser sand at the bottom and finer sand and silt towards the top. Their basins were found to be 0.5 m to 2 m thick and showed a mix of clayey, silty, and sandy deposits. Last, the results compiled during this study were compared to similar depressional features (i.e., playa-lunette systems) to pinpoint any similarities in origin processes. Altogether, this study shows that CNNs are valuable tools for automated geomorphic feature detection and can lead to new insights when coupled with various forms of remotely sensed and field-based datasets.

show abstract

Quantifying and analysing rock trait distributions of rocky fault scarps using deep learning

Chen

Scott

Keating

et al. 2023

Earth Surf Processes Landf

Self Cite

View full text Add to dashboard Cite

We apply a deep learning model to segment and identify rock characteristics based on a Structure‐from‐Motion orthomap and digital elevation model of a rocky fault scarp in the Volcanic Tablelands, Eastern California, USA. By post‐processing the deep learning results, we build a semantic rock map and analyse the rock trait distributions. The resulting semantic map contains nearly 230 000 rocks with effective diameters ranging from 2 to 250 cm. Rock trait distributions provide a new perspective on rocky fault scarp development and extend past research on scarp geometry including slope, height and length. Heatmaps indicate rock size spatial distributions on the fault scarp and surrounding topographic flats. Median grain size changes perpendicular to the fault scarp trace, with the largest rocks on the downslope proximal to the scarp footwall. Correlation analyses illustrate the relationship between rock trait statistics and fault scarp geomorphology. Local fault scarp height correlates with median grain size ( R2 = 0.6), the mean grain size of the largest rocks ( R2 = 0.8) and the ratio of the number of small to large rocks ( R2 = 0.4). The positive correlation ( R2 = 0.8) between local fault scarp height and standard deviation of grain size suggests that rocks on a higher fault scarp are less well sorted. The correlation analysis between fault scarp height and rock orientation statistics supports a particle transportation model in which locally higher fault scarps have relatively more rocks with long axes parallel to fault scarp trace because rocks have a larger distance to roll and orient the long axes. Our work demonstrates a data‐driven approach to geomorphology based on rock trait distributions, promising a greater understanding of fault scarp formation and tectonic activity, as well as many other applications for which granulometry is an indicator of process.

show abstract

Geomorphological Analysis Using Unpiloted Aircraft Systems, Structure from Motion, and Deep Learning

Cited by 15 publications

References 25 publications

Data-Driven Approaches for Tornado Damage Estimation with Unpiloted Aerial Systems

Data-Driven Approaches for Tornado Damage Estimation with Unpiloted Aerial Systems

Using Convolutional Neural Networks for Detection and Morphometric Analysis of Carolina Bays from Publicly Available Digital Elevation Models

Quantifying and analysing rock trait distributions of rocky fault scarps using deep learning

Contact Info

Product

Resources

About