Visual Sensing and Analytics for Construction and Infrastructure Management Academic Committee Annual Conference Speaker Moderator: Burcu Akinci, Carnegie Mellon University 2015 CII Annual Conference August 3 5 Boston, Massachusetts
General research question What are different types of visual sensing technologies? Current and coming in the horizon How do they perform for construction/infrastructure data collection/situation awareness needs? What types of analytics can be performed on this data? How do they help in supporting a variety of construction and infrastructure management decisions?
Panel Mani Image Golparvar-Fard, based sensing Assist. and Prof analytics of Civil for & Env. Eng.& construction Computer Scien., University of Illinois Daniel Aerial Huber, Robots for Senior Infrastructure Systems Scientist at Robotics Management Institute, Carnegie Mellon Univ. Burcu Terrestrial Akinci, scanners Paul and Christiano analytics Prof. for of construction and Civil & Env. Eng., infrastructure Carnegie Mellon Univ. management Lincoln Visual Wood, Sensing Manager, and analytics Virtual Design current & usages Construction, Turner Construction
Dr. Mani Golparvar-Fard, University of Illinois at Urbana-Champaign IMAGE BASED SENSING AND ANALYTICS FOR CONSTRUCTION
Smooth flow of production in construction Identifying different forms of waste 25-50% waste in coordination labor and equipment and in managing, moving, and installation material. http://ih.constantcontact.com/fs119/1111590816962/img/136.png?a=1114613873918 Cost overrun and delays 90% of projects exhibit average 28% higher cost than their forecasted cost. What do I need for minimizing waste? Continuous downstream feedback. Awareness on who does what task, and where
Opportunity- 5D BIM for progress analysis Turner Construction PB Operation-Level 4D (3D + Schedule) 5D (3D + Schedule + Cost) Extend the application of BIM/CIM primarily used for clash prevention and constructability review as a basis for monitoring work in progress
Opportunities for Reality capture (Images & Videos) Time-lapse photography and videos Unordered construction pictures Drones equipped with cameras
Autonomous Image Data Collection Automatic creation of flight path Icarusaerials U of Illinois Collaboration, 2015, FL
Opportunity- 4D As-built Models + 4D BIM BIM and as-built point cloud are jointly registered in this environment $500M Sacramento King s Stadium Sacramento, California Turner Construction
Automated Detection of Progress Deviations Color-coding deviations BIM elements based on traffic light color metaphor Components ahead of schedule Components behind schedule Weekly Work Plan Updates
Video-based Activity Analysis via Crowdsourcing Worker with Role A is conducting Activity B with Tool C at Body Posture D
Dr. Burcu Akinci, Carnegie Mellon University LIDAR FOR CONSTRUCTION
Opportunity LIDAR for reality capture Increased precision and density Hand-held laser scanners Terrestrial scanning Increased spatial Mobile ground scanning Scanning with Drones coverage Airborne and efficiency lidar http://cenews.com/article/8332/mobile_laser_scanning http://www.geoconnexion.com/news/optech-to-deliver-state-of-the-art-airborne-lidar-and-thermal-imaging-solut/_
Opportunity LIDAR for reality capture
Opportunity Integration of Virtual with Reality Context Reality
Assessing the capabilities of 3D imaging technologies Surface 2 Surface 1 0.888 m 0.914 m Surface 3 0.859 m 0.893 m 0.920 m 0.926 m Surface flatness 0.858 m 0.931 m Edge 1 Edge 2 Edge detection and Boundary effects Crack Detection
Points to BIM Sensor / Data Collection Errors Registration Errors Modeling Errors
Construction decision support Construction quality control Virtual inspection
Automated up-date of BIM
Dr. Daniel Huber, Carnegie Mellon University AERIAL ROBOTS FOR INFRASTRUCTURE MANAG.
The Advent of Drones for Inspection
ARIA The Aerial Robotic Infrastructure Analyst Benefits Go in difficult to reach places Comprehensive monitoring Reduced footprint Offline access Challenges Difficult to get hands on FAA, regulation, safety, privacy
ARIA Research Objectives Rapid infrastructure modeling and analysis Robotic inspection assistant Immersive inspection and assessment Algorithms transform 3D and imagery from the MAV into a high-level semantic model, and finally a finite element model. The robot acts as an inspector s apprentice, learning to accomplish inspection tasks with various levels of autonomy. A visualization environment provides an immersive virtual infrastructure representation to aid in inspection and assessment tasks.
Laser and Visual Odometry and Mapping
Exploration Planning
Reverse Engineering Our Infrastructure
Virtual Inspection
Lincoln Wood, Turner Construction VISUAL SENSING CURRENT USAGES
Problem We are great planners but we are poor at adjusting as we go Results in large delta between as-planned vs.as-built data
Switch primary focus from Office to field to field to office As-planned vs. as-built comparison and documentation requires images/videos to be located within BIM environment http://lanmarservices.com/wp-content/uploads/2014/12/pipes-comparison.jpg
Data Requirements Reliable/Trust Relevant- that impacts 3 week look-ahead plan Speed (must be at the pace of the conversation/meeting) Task Readiness Level = R Location Entropy = TL Highlighting as-risk locations with Integrated point cloud and BIM Location-based monitoring for tasks not associated with physical elements
Social Patchwork or small batch is OK (crowd sourced with "infill" capture) Ownership of the builder / data collector
Conclusion Visual Sensing and Analytics can Reduce the gap between as-built and as-planned data Minimize the challenges for data collection, synthesis, and analytics Enable root-cause assessment on potential/actual deviations Detect and communicate at-risk locations In case of videos, provided detailed information on work efficiency During operation, provides up-to-date information on condition and location of assets
Future directions Autonomous data collection: Robotic vehicles path-planning and data collection with engineered precision Data processing and management: Data cleaning without removing artifacts, data fusion, big data management Analytics: Real-time awareness of work status and availability of resources, more reliable weekly work plans, enhanced QA/QC and safety monitoring, accurate and up-to-date as-is modeling Communication and Visualization: Real-time project controls, virtual analyses, complete situation awareness