{"id":{"repo_id":"uiuc","oai_identifier":"oai:www.ideals.illinois.edu:2142/101504"},"canonical_url":"https://search.dev.ndltd.org/etd/uiuc/oai:www.ideals.illinois.edu:2142/101504","repository":{"repo_id":"uiuc","name":"University of Illinois - Urbana-Champaign","base_url":"https://www.ideals.illinois.edu/oai-pmh"},"display":{"title":"Towards vision based robots for monitoring built environments","abstract":"In construction, projects are typically behind schedule and over budget, largely due to the difficulty of progress monitoring. Once a structure (e.g. a bridge) is built, inspection becomes an important yet dangerous and costly job. We can provide a solution to both problems if we can simplify or automate visual data collection, monitoring, and analysis. In this work, we focus specifically on improving autonomous image collection, building 3D models from the images, and recognizing materials for progress monitoring using the images and 3D models. Image capture can be done manually, but the process is tedious and better suited for autonomous robots. Robots follow a set trajectory to collect data of a site, but it is unclear if 3D reconstruction will be successful using the images captured by following this trajectory. We introduce a simulator that synthesizes feature tracks for 3D reconstruction to predict if images collected from a planned path will result in a successful 3D reconstruction. This can save time, money, and frustration because robot paths can be altered prior to the real image capture. When executing a planned trajectory, the robot needs to understand and navigate the environment autonomously. Robot navigation algorithms struggle in environments with few distinct features. We introduce a new fiducial marker that can be added to these scenes to increase the number of distinct features and a new detection algorithm that detects the marker with negligible computational overhead. Adding markers prior to data collection does not guarantee that the algorithms for 3D model generation will succeed. In fact, out of the box, these algorithms do not take advantage of the unique characteristics of markers. Thus, we introduce an improved structure from motion approach that takes advantage of marker detections when they are present. We also create a dataset of challenging indoor image collections with markers placed throughout and show that previous methods often fail to produce accurate 3D models. However, our approach produces complete, accurate 3D models for all of these new image collections. Recognizing materials on construction sites is useful for monitoring usage and tracking construction progress. However, it is difficult to recognize materials in real world scenes because shape and appearance vary considerably. Our solution is to introduce the first dataset of material patches that include both image data and 3D geometry. We then show that both independent and joint modeling of geometry are useful alongside image features to improve material recognition. Lastly, we use our material recognition with material priors from building plans to accurately identify progress on construction sites.","abstract_html":"In construction, projects are typically behind schedule and over budget, largely due to the difficulty of progress monitoring. Once a structure (e.g. a bridge) is built, inspection becomes an important yet dangerous and costly job. We can provide a solution to both problems if we can simplify or automate visual data collection, monitoring, and analysis. In this work, we focus specifically on improving autonomous image collection, building 3D models from the images, and recognizing materials for progress monitoring using the images and 3D models. Image capture can be done manually, but the process is tedious and better suited for autonomous robots. Robots follow a set trajectory to collect data of a site, but it is unclear if 3D reconstruction will be successful using the images captured by following this trajectory. We introduce a simulator that synthesizes feature tracks for 3D reconstruction to predict if images collected from a planned path will result in a successful 3D reconstruction. This can save time, money, and frustration because robot paths can be altered prior to the real image capture. When executing a planned trajectory, the robot needs to understand and navigate the environment autonomously. Robot navigation algorithms struggle in environments with few distinct features. We introduce a new fiducial marker that can be added to these scenes to increase the number of distinct features and a new detection algorithm that detects the marker with negligible computational overhead. Adding markers prior to data collection does not guarantee that the algorithms for 3D model generation will succeed. In fact, out of the box, these algorithms do not take advantage of the unique characteristics of markers. Thus, we introduce an improved structure from motion approach that takes advantage of marker detections when they are present. We also create a dataset of challenging indoor image collections with markers placed throughout and show that previous methods often fail to produce accurate 3D models. However, our approach produces complete, accurate 3D models for all of these new image collections. Recognizing materials on construction sites is useful for monitoring usage and tracking construction progress. However, it is difficult to recognize materials in real world scenes because shape and appearance vary considerably. Our solution is to introduce the first dataset of material patches that include both image data and 3D geometry. We then show that both independent and joint modeling of geometry are useful alongside image features to improve material recognition. Lastly, we use our material recognition with material priors from building plans to accurately identify progress on construction sites.","abstract_has_math":false,"creators":["Degol, Joseph"],"institution":"University of Illinois at Urbana-Champaign","degree_name":"Ph.D.","degree_level":"Dissertation","degree_discipline":"Computer Science","degree_department":null,"school":null,"contributors":["Hoiem, Derek","Bretl, Timothy","Golparvar-Fard, Mani","Forsyth, David","Sinha, Sudipta"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2018,"date_issued":"2018-09-27T16:17:31Z","date_published":"2018-09-27T16:17:31Z","updated_at":"2026-07-22T22:24:40Z","subjects":["Computer Vision, 3D Reconstruction, Fiducial Markers, Material Recognition, Construction Progress Monitoring, Robotics, Drones, Structure from Motion"],"languages":["en"],"rights":["Copyright 2018 by Joseph DeGol. All rights reserved."],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/2142/101504","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Hoiem, Derek","Bretl, Timothy","Golparvar-Fard, Mani","Forsyth, David","Sinha, Sudipta"]},{"key":"dc:creator","label":"Author","values":["Degol, Joseph"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2018-09-27T16:17:31Z","2018-06-26","2018-08"]},{"key":"dc:type","label":"Dc Type","values":["text"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Science"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Ph.D."]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["University of Illinois at Urbana-Champaign"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computer Vision, 3D Reconstruction, Fiducial Markers, Material Recognition, Construction Progress Monitoring, Robotics, Drones, Structure from Motion"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["Copyright 2018 by Joseph DeGol. All rights reserved."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["http://hdl.handle.net/2142/101504"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["In construction, projects are typically behind schedule and over budget, largely due to the difficulty of progress monitoring. Once a structure (e.g. a bridge) is built, inspection becomes an important yet dangerous and costly job. We can provide a solution to both problems if we can simplify or automate visual data collection, monitoring, and analysis. In this work, we focus specifically on improving autonomous image collection, building 3D models from the images, and recognizing materials for progress monitoring using the images and 3D models. Image capture can be done manually, but the process is tedious and better suited for autonomous robots. Robots follow a set trajectory to collect data of a site, but it is unclear if 3D reconstruction will be successful using the images captured by following this trajectory. We introduce a simulator that synthesizes feature tracks for 3D reconstruction to predict if images collected from a planned path will result in a successful 3D reconstruction. This can save time, money, and frustration because robot paths can be altered prior to the real image capture. When executing a planned trajectory, the robot needs to understand and navigate the environment autonomously. Robot navigation algorithms struggle in environments with few distinct features. We introduce a new fiducial marker that can be added to these scenes to increase the number of distinct features and a new detection algorithm that detects the marker with negligible computational overhead. Adding markers prior to data collection does not guarantee that the algorithms for 3D model generation will succeed. In fact, out of the box, these algorithms do not take advantage of the unique characteristics of markers. Thus, we introduce an improved structure from motion approach that takes advantage of marker detections when they are present. We also create a dataset of challenging indoor image collections with markers placed throughout and show that previous methods often fail to produce accurate 3D models. However, our approach produces complete, accurate 3D models for all of these new image collections. Recognizing materials on construction sites is useful for monitoring usage and tracking construction progress. However, it is difficult to recognize materials in real world scenes because shape and appearance vary considerably. Our solution is to introduce the first dataset of material patches that include both image data and 3D geometry. We then show that both independent and joint modeling of geometry are useful alongside image features to improve material recognition. Lastly, we use our material recognition with material priors from building plans to accurately identify progress on construction sites.","Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2018-09-27 without embargo terms","The student, Joseph Degol, accepted the attached license on 2018-06-25 at 16:01.","The student, Joseph Degol, submitted this Dissertation for approval on 2018-06-25 at 16:25.","This Dissertation was approved for publication on 2018-06-26 at 17:36.","DSpace SAF Submission Ingestion Package generated from Vireo submission #12670 on 2018-09-27 at 10:45:32","Made available in DSpace on 2018-09-27T16:17:31Z (GMT). No. of bitstreams: 3 DEGOL-DISSERTATION-2018.pdf: 65806930 bytes, checksum: 8047d0ac46fac793f04297852792ec31 (MD5) LICENSE.txt: 4209 bytes, checksum: 9e5fa3f23fcc0e874ba93c3204666a27 (MD5) PROQUEST_LICENSE.txt: 4555 bytes, checksum: a8d799c1a77757a015ccd2173823753b (MD5) Previous issue date: 2018-06-26"]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Towards vision based robots for monitoring built environments"]}]}],"canonical_facts":{"dc:contributor":["Hoiem, Derek","Bretl, Timothy","Golparvar-Fard, Mani","Forsyth, David","Sinha, Sudipta"],"dc:creator":["Degol, Joseph"],"dc:date":["2018-09-27T16:17:31Z","2018-06-26","2018-08"],"dc:description":["In construction, projects are typically behind schedule and over budget, largely due to the difficulty of progress monitoring. Once a structure (e.g. a bridge) is built, inspection becomes an important yet dangerous and costly job. We can provide a solution to both problems if we can simplify or automate visual data collection, monitoring, and analysis. In this work, we focus specifically on improving autonomous image collection, building 3D models from the images, and recognizing materials for progress monitoring using the images and 3D models. Image capture can be done manually, but the process is tedious and better suited for autonomous robots. Robots follow a set trajectory to collect data of a site, but it is unclear if 3D reconstruction will be successful using the images captured by following this trajectory. We introduce a simulator that synthesizes feature tracks for 3D reconstruction to predict if images collected from a planned path will result in a successful 3D reconstruction. This can save time, money, and frustration because robot paths can be altered prior to the real image capture. When executing a planned trajectory, the robot needs to understand and navigate the environment autonomously. Robot navigation algorithms struggle in environments with few distinct features. We introduce a new fiducial marker that can be added to these scenes to increase the number of distinct features and a new detection algorithm that detects the marker with negligible computational overhead. Adding markers prior to data collection does not guarantee that the algorithms for 3D model generation will succeed. In fact, out of the box, these algorithms do not take advantage of the unique characteristics of markers. Thus, we introduce an improved structure from motion approach that takes advantage of marker detections when they are present. We also create a dataset of challenging indoor image collections with markers placed throughout and show that previous methods often fail to produce accurate 3D models. However, our approach produces complete, accurate 3D models for all of these new image collections. Recognizing materials on construction sites is useful for monitoring usage and tracking construction progress. However, it is difficult to recognize materials in real world scenes because shape and appearance vary considerably. Our solution is to introduce the first dataset of material patches that include both image data and 3D geometry. We then show that both independent and joint modeling of geometry are useful alongside image features to improve material recognition. Lastly, we use our material recognition with material priors from building plans to accurately identify progress on construction sites.","Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2018-09-27 without embargo terms","The student, Joseph Degol, accepted the attached license on 2018-06-25 at 16:01.","The student, Joseph Degol, submitted this Dissertation for approval on 2018-06-25 at 16:25.","This Dissertation was approved for publication on 2018-06-26 at 17:36.","DSpace SAF Submission Ingestion Package generated from Vireo submission #12670 on 2018-09-27 at 10:45:32","Made available in DSpace on 2018-09-27T16:17:31Z (GMT). No. of bitstreams: 3 DEGOL-DISSERTATION-2018.pdf: 65806930 bytes, checksum: 8047d0ac46fac793f04297852792ec31 (MD5) LICENSE.txt: 4209 bytes, checksum: 9e5fa3f23fcc0e874ba93c3204666a27 (MD5) PROQUEST_LICENSE.txt: 4555 bytes, checksum: a8d799c1a77757a015ccd2173823753b (MD5) Previous issue date: 2018-06-26"],"dc:format":["application/pdf"],"dc:identifier":["http://hdl.handle.net/2142/101504"],"dc:language":["en"],"dc:rights":["Copyright 2018 by Joseph DeGol. All rights reserved."],"dc:subject":["Computer Vision, 3D Reconstruction, Fiducial Markers, Material Recognition, Construction Progress Monitoring, Robotics, Drones, Structure from Motion"],"dc:title":["Towards vision based robots for monitoring built environments"],"dc:type":["text"],"thesis:degree_discipline":["Computer Science"],"thesis:degree_level":["Dissertation"],"thesis:degree_name":["Ph.D."],"thesis:institution_name":["University of Illinois at Urbana-Champaign"]},"updated_at":"2026-07-22T22:24:40Z"}