From Data to Situational Awareness
Semantic Segmentation of 3D Data: Semantic Terrain Points Labeling
It is well known that deep-learning algorithms are data-hungry, especially in the 3D domain. Acquiring and annotating real-world 3D data is a labor-intensive process that requires many staff hours and corresponding supervision to ensure that the resulting annotations are accurate. To this end, we have designed and developed a synthetic 3D data generation pipeline to create synthetic training data that can augment or even replace real-world training data. We aim to provide a large database of annotated ground truth point clouds reconstructed using aerial photogrammetry for training and validating 3D semantic and instance segmentation algorithms. Please refer to our project website for more details: www.stpls3d.com
A Data Acquisition-Planning Framework for Hybrid Data Collection Techniques
planning algorithms with a single data acquisition technique. However, for buildings that have complex building structure and architectural elements, data collection process with a single data acquisition technique is not sufficient or effective. The hypothesis behind this research study is that image-based technique (photogrammetry) and range-based technique (laser scanning) are complementary to each other and that the combination of the two techniques can improve the quality of the derived as-is 3D point cloud in terms of completeness and accuracy. As such, this project develops a framework that will provide an improved data acquisition process to support the creation of complete and accurate 3D models of existing buildings, while reducing the total cost by eliminating the need for site revisits and rework.
As-is 3D building models are valuable in many ways, such as urban planning, historical building information storage, building renovation, facility management, building energy simulation and so on. Data acquisition for complete and accurate as-is 3D building reconstruction is a time consuming and labor intensive process. Establishing a data acquisition plan before or during the data acquisition process is necessary. As such, there has been extensive research on developing/advancing data acquisition
In-situ Quality Control for Scan to BIM
Laser scanners capture geometric information of a building in the order of minutes and with millimeters accuracy, which makes them a valuable asset, especially when capturing the details of the geometry is required. However, to date, challenges remain with accurate detection and extraction of building primitives from the scan data. These challenges often stem from the early data acquisition stage. Existence of data quality issues such as
missing data or low point cloud resolution (point density) result in inaccurate 3D BIM models and often raises the need for redoing the scan process. This project focuses on analyzing the scan data of buildings, and specifically investigates the data quality requirements for modeling the architectural elements on building exteriors within the Scan-to-BIM context. The knowledge about scan data acquisition means and methods (e.g., scan planning, adjusting resolution settings) is integrated with the as-built modeling techniques to realize their synergy in the context of the Scan-to-BIM process. The integration bridges the existing gap in accurate as-built modeling by taking into consideration the requirements for generating a high quality 3D model in the very first steps of data collection. The research objectives are: (1) Identification of the scan data quality requirements for the Scan-to-BIM process; (2) In-situ identification and classification of scan data quality issues at building and feature (e.g. doors, windows) level; (3) Recognition of causes for different data quality issues; and (4) Improving the scan quality based on the knowledge of inherent quality issues.
Indoor Localization for In-Building Emergencies
Building emergencies especially structure fires are big threats to the safety of building occupants and first responders. When emergencies occur, unfamiliar environments are difficult and dangerous for first responders to search and rescue, sometimes leading to secondary casualties. One way to reduce such hazards is to provide first responders with timely access to accurate location information. This research assesses the value of location information through a card game, and identifies a set of requirements for indoor localization through a survey. The most important five requirements are: accuracy, ease of on-scene deployment, resistance to damages, computational speed, and device size and weight. This research introduces a radio frequency (RF) based indoor localization framework to satisfy these requirements. Two algorithms are designed to infer people's indoor locations based on RF signal data collected from existing sensing infrastructure in buildings or ad-hoc networks. Moreover, building information models are integrated to both algorithms. Building information plays an important role in mitigating multipath and fading effects in iterative location computation, enabling the metaheuristic based search for building-specific satisfactory beacon deployment plans, and providing a graphical interface for user interaction. The framework is validated in extensive simulation and field tests, and has reported promising results in satisfying the aforementioned requirements for indoor localization at building emergencies.
As-Built Model Generation, Verification, Maintenance Using Images
As-built models and drawings are essential documents used during the operations and maintenance (O&M) of buildings for a variety of purposes including the management of facility spaces, equipment, and energy systems. These documents undergo continuous verification and updating procedures as buildings are changed and renovated overtime resulting in much time spent taking manual surveys of existing conditions and building dimensions. In some cases, as-built models and drawings do not exist and need to be regenerated from as-built conditions. This project attempts to improve the efficiency of the as-built model creation, verification, and maintenance processes by introducing image capture technology to the facility management as-built workflow. As part of this research effort, building images are captured by ordinary digital cameras and then stitched together to create 3D point clouds. A case study of a university building is used to compare the time and costs associated with models generated from the image capture technology point clouds and technologies procedures currently employed by facilities management. The point clouds are also compared to existing as-built BIM models to assess their accuracy and their value for developing 3D as-built models for other buildings.
RFID-Based Indoor Location Sensing Solutions for the Construction Industry
Indoor location information is of great value to the construction industry laying the foundation of various context aware information services such as equipment maintenance and energy consumption monitoring. Among a number of competing technologies, Radio Frequency Identification
(RFID) demonstrates an advantageous balance between system cost and accuracy, which makes it a promising solution for Indoor Location Sensing (ILS) in built environments. This project aims at developing an RFID based ILS system that is adaptable to the construction industry. An innovative localization algorithm has been developed and tested in room, floor and building levels. Its performance will be evaluated against the following criteria: accuracy, cost efficiency, robustness, scalability, and ease to use. Potential areas of application include facilities management, on-site construction inspection, building energy conservation and emergency response. In addition, the team is currently working on calibration and correction methods for algorithm improvement.
Sensor-based Motorized Autonomous Responsive Target (SMART)
3D laser scanning technology is now widely and increasingly used to create as-built 3D CAD or Building Information Models (BIM) of existing facilities
and new construction. Errors occurred during “Scan-to-BIM” process have different sources of error, including instrument calibration, environmental conditions, point cloud registration, fitting algorithms, and manual modeling. In addition, each of these processes is manual and time consuming. In the first phase of the SMART project, Sensor-based Motorized Autonomous Responsive Target (SMART) is devised to replace paddle targets to reduce registration errors and inefficiencies incurred by manual re-orientation of paddle targets. The second phase of the project focuses on automated meta-data transmittal between the SMART and point cloud engine so that no placement of target IDs by scan crew is necessary and registration process is automated.