NRI: Considerate Co-robot Intelligence through Ubiquitous Human State Awareness, National Science Foundation, National Robotics Initiative (NRI), 2014-2017, PI: Weihua Sheng and Co-PI: Guoliang Fan.

Summary: The objective of this project is to develop a new theoretical/algorithmic framework and an open hardware/software platform for considerate co-robot intelligence, enabling a co-robot to assist humans in their daily lives in a proactive way while still having the freedom to do its routine work. Such considerate intelligence is developed through ubiquitous human state awareness-knowing human’s activity and location in an indoor environment without constantly following and watching the human. This capability is realized through wearable sensing and computing from a human-based perspective. The major research efforts consist of four parts: co-robot semantic mapping through human-environment interaction; human activity and location inference using minimal motion sensor data; activity prediction and behavioral anomaly detection based on human state awareness; experimental evaluation using open hardware/software platforms and a case study evaluating the effectiveness of considerate co-robot intelligence in elderly fall prevention, detection and intervention.

Go to quick links

A Tool for Posture Assessment and Personalized Training, OCAST, Health Research Program, 2012-2015, PI: Guoliang Fan.

Summary: Virtually all adults have experienced one or more brief episodes of musculoskeletal pain associated with injury, overuse, and more often, poor posture and gait. Recurrent or chronic musculoskeletal pain problems are also common (10-40%). The estimated total annual cost of productive time lost due to back pain, arthritis and other musculoskeletal pain problems was US$ 41.7 billion in 2002. Dynamic posture/gait assessment and personalized training/rehabilitation are the best remedies for detecting the early sign of poor posture/gait, preventing the deterioration of various musculoskeletal pains and improving posture and gait through targeted intervention. However, existing clinic tools are mainly used for static posture assessment or qualitative gait analysis, and have limited capability for dynamic and quantitative posture/gait assessment that is more useful and informative for diagnosis, treatment and prevention. This project seeks an integrated multi-sensor approach to develop a comprehensive yet economic clinical tool for dynamic posture/gait assessment, which is expected to surpass all existing ones by providing accurate and quantitative functional motion analysis as well as targeted intervention and personalized training via serious health games.

Go to quick links

Vision-based Clinical Markerless Gait Stability Analysis, OCAST, Health Research Program, 2009-2012, PI: Guoliang Fan, Collaborator: Dr. Li-Shan Chou, University of Oregon.

Summary: Falls among the elderly population are prevalent, dangerous, and costly. Early fall-risk detection would greatly enhance our ability to design interventions for preventing fall injuries in community-dwelling elderly. This project seeks a vision-based prototype system that is able to assess gait instability in clinical settings with normal video cameras. This project seeks an integrated vision-based approach for gait instability analysis, and our goal is to promote the feasibility and applicability of markerless motion capture for detecting early fall-risk in the elderly in an uncontrolled environment. The fundamental assumption is that a new human gait motion can be extrapolated from a set of representative human motions with sufficient accuracy and robustness for gait instability assessment. The long term goal will be a feasible solution to clinical markerless gait stability analysis that has tremendous potential value for monitoring safe ambulation and recognizing accidental falls in a video surveillance environment.

Go to quick links

Video-based Assessment of Crew Vibration during Rocket Launch, Oklahoma NASA EPSCoR, 2009, PI: Guoliang Fan.

Summary: This project aims to develop a practical prototype system for video-based human motion analysis that is used for crew vibration assessment during rocket launch. This topic emerges as an important issue when NASA is planning to bring a new crew exploration vehicle into serve after shuttle retirement for which the new rocket lunch system has to be much more powerful and stronger, and consequentially could be more dangerous to the safety of crew due to the significant vibration created during rocket lunch. It is critical to understand the effects of such vibrations on the crew members to ensure that they are safe and able to perform their tasks effectively. In this project, we will extend our existing human motion analysis research effort for real-time vibration assessment.

Go to quick links

Multiple Domain Particle Filters with Dual-Band Sensing for Multi-Contextual Tracking and Recognition, Army Research Office (ARO), 2008-2011, PI: Joseph Havlicek (OU) and Co-PI: Guoliang Fan.

Summary: This research focuses on two important ATR-related research issues that extend our previous ATR research. The first one is how to incorporate additional and relevant contextual information to improve the ATR performance by better handling the uncertainty and incompleteness in the observed infrared imagery, which includes a geometric camera model, a background  representation, as well as a 3D target representation and motion models. The second one is how to develop a generative graphical model to integrate tracking, recognition and learning in one computational flow where these ATR tasks are systematically and jointly formulated as one inference problem involving multi-contextual information. Also, we will collaborate with the OU team led by Dr. Joseph Havlicek to validate and test our algorithms in an experimental platform that involves dual-band sensing and multiple-domain target modeling.

Go to quick links

CAREER: Advanced Statistical Modeling Approaches for Structured Video Representation and Research Oriented Multidisciplinary Education, National Science Foundation, 2004-2009, PI: Guoliang Fan.

Summary: This research aims to develop structured video representations via advanced statistical modeling approaches. The goal is to enhance interpretability and manipulability of visual data at both the object-level and the scene-level. Specifically, at the objective level, we are interested in developing new human representations for integrated pose recognition, localization, segmentation, and tracking as well as human motion estimation. At the scene level, we are focused on exploring and learning semantic structures from sports or commercial videos. These two studies are complementary and inter-related that address fundamental issues in computer vision, with widespread applications, including multimedia, digital library, visual surveillance, entertainment, and human-computer interface, etc. Moreover, the proposed research would have significant contribution to industry standards. 

Go to quick links

Integrated Target Detection, Tracking, Classification, and Learning for Dual-band Infrared Imagery, Army Research Office (ARO), 2004-2007, PI: Joseph Havlicek (OU) and Co-PI: Guoliang Fan.

Summary: Summary: Infrared (IR) automatic target recognition (ATR) systems are employed by all three services in variety of critical military system to detect, track, and classify targets based their thermal signatures. In this project, we want to enhance the overall ATR performance by addressing three major issues, appearance representation, dynamic modeling, and integrated algorithm implementation. The research outcomes will lead to an unified ATR system which can incorporate both motion and visual cues into a Bayesian estimation framework for integrated detection, tracking, recognition and learning. 

Go to quick links

Advanced Retinal Imaging for Non-invasive Disease Study, OCAST (Oklahoma Center for Advancement and Science Technology) Health Research Program, 2003-2006, PI: Guoliang Fan, and Co-PI: Gary Yen. 

Summary: If detected early, ninety-five percent of the severe vision loss from diabetic retinopathy is preventable; yet, 40,000 people still go blind each year from this disease. Failure to undergo universally recommended annual eye examinations is the primary cause of this continued loss of sight. Digital retinal imaging systems, conveniently located in the primary care environment and connected to expert graders via computer networks, provide "old-standard" quality evaluations. To efficiently scale these systems, software tools are needed to certify image quality before the patient leaves the imaging center, facilitate grading, and support the grader training processes. In this project, we aim to develop and validate advanced retinal imaging approaches to (1) assess and control the quality of retinal imaging, (2) promote disease detection, staging and monitoring by novel 2-D/3-D image annotation methods, and (3) invoke a hierarchical grading process to optimize the involvement of graders of different levels. 

Go to quick links

Using a Multi-Resolution GIS-Modeling Approach to Evaluate the Carbon Sequestration Potential in Texas County, Oklahoma, Oklahoma NASA EPSCoR, Feb. 2005-July 2005, PI: Mahesh Rao (Geography), Co-PI: Guoliang Fan and Johnson Thomas (CS)

Summary: This project aims to evaluate the carbon sequestration potential in Oklahoma panhandle using a multi-resolution GIS-based modeling approach. Owing to USDA's largest program, conservation Reserve Program (CRP), a majority of the areas in the region are sustainable, in spite of intensive agriculture. Our project seeks to evaluate the long-term carbon sequestering potential of CRP in Texas County , Oklahoma . Specifically, we will develop a multi-resolution remote sensing approach that involves MODIS and Landsat data as well as two GIS models from NASA and USDA, i.e., CASA (Carnegie Ames Stanford Approach) and SWAT (Soil and Water Assessment Tool). Moreover, interactive querying tools will be developed and fused with the ArcGIS interface using .NET programming modules. Our long-term goal is to develop an integrated query-based Web-GIS Decision Support System (DSS) to help evaluate the environmental benefits of CRP and manage future enrollments.

Go to quick links

Development of a Web-GIS Decision Support System for Environmental Water Quality and Resource Management, OSU Environmental Institute's Water Research Center, 2003-2004, PI: Johnson Thomas (Computer Science), Co-PI: Guoliang Fan, and Co-PI: Mahesh Rao (Geography).

Summary: Conservation programs such as the USDA Conservation Reserve Program (CRP) address environmental concerns including the decline of riparian areas on private lands. The CRP Program encourages farmers to plant long-term resource-conserving covers to improve soil, water and wildlife resources. On Oct. 1st, 2002 , USDA announced payments of nearly $1.6 billion for CRP. This calls for sophisticated, accurate, and timely decision-support aids and research tool to evaluate and justify the environmental and water benefits of the CRP program. The overall goal of this proposal is to design and develop a prototype web-GIS Decision Support System (DSS) aimed at aiding USDA/NRCS to better manage and plan CRP enrollments. Our proposed DSS will be based on the emerging industry-standard ArcIMS GIS platform and will integrate a mapping component AFIRS (Automated Feature Information Retrieval System) and a modeling component SWAT (Soil and Water Assessment Tool).

Go to quick links

Toward an Integrated Web-GIS Decision Support System for Evaluating USDA's Conservation Reserve Program (CRP), Oklahoma NASA EPSCoR Research Initiation Grant, 2003, PI: Guoliang Fan, Co-PI: Johnson Thomas (Computer Science), and Co-PI: Mahesh Rao (Geography).

Summary: This project aims to develop a prototype of an integrated Web-GIS Decision Support System (DSS) for USDA Conservation Reserve Program (CRP), i.e., CRP-DSS. CRP is a voluntary program to provide incentives for farmers and ranchers to strengthen environmental stewardship of their lands, and gives producers additional resources to reduce topsoil erosion, increase wildlife habitat and improve air and water quality on these lands. However, CRP has been criticized for administrative shortcomings and failure to achieve ancillary environmental objectives, such as improving wildlife habitat and promoting water quality. Key to the inefficiency of current CRP procedures is the lack of automated decision-support tools. The objective of this project is to develop accurate and timely decision-support aids and research tools to delineate and evaluate CRP. The evaluation of CRP is based on a GIS-based environmental modeling approach. More importantly, we will propose a prototype CRP-DSS that will be interfaced with the Internet and be capable of accessing databases in a distributed environment. Our long-term goal is to develop an integrated Web-GIS DSS to help USDA manage, plan, and prioritize CRP enrollments, leading to  maximum environmental benefit within the budget constraints.

Go to quick links

Developing a GIS-based Tool for Automated Feature Information Retrieval from Multisource Geospatial Data: Application to CRP Mapping at Texas County, Oklahoma, OSU Environmental Institute's Water Research Center, July 2002-June 2003, PI: Guoliang Fan and Co-PI: Mahesh Rao (Geography).

Summary: This project aims at developing a GIS-based tool, Automated Feature Information Retrieval System, to delineate USDA's Conservation Reserve Program (CRP) tracts. In addition to the satellite imagery, AFIRS will involve multisource geospatial data to achieve accurate and robust feature extractions. Specifically, a GIS database consisting of multisource data, including Landsat TM imagery, soils, elevation, and slope data, etc., will be developed for Texas County, OK, which ranks first in the state for CRP enrollments. Based on this database and reference data, AFIRS will be trained, developed, and verified. Two machine learning approaches, i.e., decision tree classifier (DTC) and support vector machine (SVM), are applied to develop AFIRS in this work. Experimental results show that significant improvements can be obtained by incorporating GIS ancillary data and other derived features. This work validates the applicability of machine learning approaches to the real-world remote sensing applications.  

Go to quick links