Automatic Muck Pile Characterization from UAV Images

Fabian Schenk, Alexander Tscharf, Gerhard Mayer, Friedrich Fraundorfer

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

In open pit mining it is essential for processing and production scheduling to receive fast and accurate information about the fragmentation of a muck pile after a blast. In this work, we propose a novel machine-learning method that characterizes the muck pile directly from UAV images. In contrast to state-of-the-art approaches, that require heavy user interaction, expert knowledge and careful threshold settings, our method works fully automatically. We compute segmentation masks, bounding boxes and confidence values for each individual fragment in the muck pile on multiple scales to generate a globally consistent segmentation. Additionally, we recorded lab and real-world images to generate our own dataset for training the network. Our method shows very promising quantitative and qualitative results in all our experiments. Further, the results clearly indicate that our method generalizes to previously unseen data.
LanguageEnglish
Title of host publicationSPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Number of pages8
StatusSubmitted - 10 Jun 2019
EventISPRS Geospatial Week: Unmanned Aerial Vehicles in Geomatics (UAVg) 2019 - University of Twente, Enschede, Netherlands
Duration: 10 Jun 201914 Jun 2019
http://www.uav-g.com/

Conference

ConferenceISPRS Geospatial Week
CountryNetherlands
CityEnschede
Period10/06/1914/06/19
Internet address

Fingerprint

Unmanned aerial vehicles (UAV)
Piles
Open pit mining
Learning systems
Masks
Scheduling
Processing
Experiments

Keywords

  • UAVs
  • Semantic Segmentation
  • Mining
  • Machine learning
  • Convolutional neural networks

Cite this

Schenk, F., Tscharf, A., Mayer, G., & Fraundorfer, F. (2019). Automatic Muck Pile Characterization from UAV Images. Manuscript submitted for publication. In SPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences

Automatic Muck Pile Characterization from UAV Images. / Schenk, Fabian; Tscharf, Alexander; Mayer, Gerhard; Fraundorfer, Friedrich.

SPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences. 2019.

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Schenk, F, Tscharf, A, Mayer, G & Fraundorfer, F 2019, Automatic Muck Pile Characterization from UAV Images. in SPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences. ISPRS Geospatial Week, Enschede, Netherlands, 10/06/19.
Schenk F, Tscharf A, Mayer G, Fraundorfer F. Automatic Muck Pile Characterization from UAV Images. In SPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences. 2019
Schenk, Fabian ; Tscharf, Alexander ; Mayer, Gerhard ; Fraundorfer, Friedrich. / Automatic Muck Pile Characterization from UAV Images. SPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences. 2019.
@inproceedings{87d1dd2c7d624b5693d58454fc58f66c,
title = "Automatic Muck Pile Characterization from UAV Images",
abstract = "In open pit mining it is essential for processing and production scheduling to receive fast and accurate information about the fragmentation of a muck pile after a blast. In this work, we propose a novel machine-learning method that characterizes the muck pile directly from UAV images. In contrast to state-of-the-art approaches, that require heavy user interaction, expert knowledge and careful threshold settings, our method works fully automatically. We compute segmentation masks, bounding boxes and confidence values for each individual fragment in the muck pile on multiple scales to generate a globally consistent segmentation. Additionally, we recorded lab and real-world images to generate our own dataset for training the network. Our method shows very promising quantitative and qualitative results in all our experiments. Further, the results clearly indicate that our method generalizes to previously unseen data.",
keywords = "UAVs, Semantic Segmentation, Mining, Machine learning, Convolutional neural networks",
author = "Fabian Schenk and Alexander Tscharf and Gerhard Mayer and Friedrich Fraundorfer",
year = "2019",
month = "6",
day = "10",
language = "English",
booktitle = "SPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences",

}

TY - GEN

T1 - Automatic Muck Pile Characterization from UAV Images

AU - Schenk, Fabian

AU - Tscharf, Alexander

AU - Mayer, Gerhard

AU - Fraundorfer, Friedrich

PY - 2019/6/10

Y1 - 2019/6/10

N2 - In open pit mining it is essential for processing and production scheduling to receive fast and accurate information about the fragmentation of a muck pile after a blast. In this work, we propose a novel machine-learning method that characterizes the muck pile directly from UAV images. In contrast to state-of-the-art approaches, that require heavy user interaction, expert knowledge and careful threshold settings, our method works fully automatically. We compute segmentation masks, bounding boxes and confidence values for each individual fragment in the muck pile on multiple scales to generate a globally consistent segmentation. Additionally, we recorded lab and real-world images to generate our own dataset for training the network. Our method shows very promising quantitative and qualitative results in all our experiments. Further, the results clearly indicate that our method generalizes to previously unseen data.

AB - In open pit mining it is essential for processing and production scheduling to receive fast and accurate information about the fragmentation of a muck pile after a blast. In this work, we propose a novel machine-learning method that characterizes the muck pile directly from UAV images. In contrast to state-of-the-art approaches, that require heavy user interaction, expert knowledge and careful threshold settings, our method works fully automatically. We compute segmentation masks, bounding boxes and confidence values for each individual fragment in the muck pile on multiple scales to generate a globally consistent segmentation. Additionally, we recorded lab and real-world images to generate our own dataset for training the network. Our method shows very promising quantitative and qualitative results in all our experiments. Further, the results clearly indicate that our method generalizes to previously unseen data.

KW - UAVs

KW - Semantic Segmentation

KW - Mining

KW - Machine learning

KW - Convolutional neural networks

M3 - Conference contribution

BT - SPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences

ER -