I/UCRC Phase II: Collaborative Research: I/UCRC Center for Surveillance Research, Phase II

Ohio State University, The, Columbus OH

Investigators

Abstract

The Center for Surveillance Research (CSR) was founded in 2010 as a Phase I NSF I/UCRC. This award enables the Center to become a Phase 2 I/UCRC. Both founding sites of the Center, at Wright State University and at Ohio State University, will move to Phase 2. The main objective of CSR is to develop the theory and practice for modern surveillance. It has a healthy membership profile that has a mix of federal and industry members. In Phase 2, Wright State University seeks to become the Lead Site of the Center, as opposed to Phase 1 lead site of Ohio State. The fundamental intellectual issues addressed by the Center's continuing research portfolio have immediate application in application domains including medical imaging, transportation, and law enforcement. Further, the center will continue to host summer internship programs, in collaboration with the Air Force Minority Leaders Program and others, to attract participation by a diverse population of students. While use of sensor technology is pervasive in modern society, fundamental questions about the performance of sensor systems remain unaddressed. The inability to answer these questions, or even properly phrase the questions, impacts not only the development of sensors themselves, but also the development of data exploitation algorithms. Further, these unanswered questions are critical for eventually providing human users the ability to understanding the reliability of sensor performance under new and changing operating conditions. Specifically, sensors are developed under an assumed mode of use, and the exploitation algorithms are then left to extract whatever information the resultant sensor may afford, without that prior development having been properly informed by a realistic performance model. Moreover, the exploitation algorithms themselves are developed under presumed statistical models that may have varying degrees of agreement with actual sensor phenomenology, and these algorithms are then trained (e.g., to set operating parameters) based on a representative set of training measurements. However, the sensor community lacks any means to propagate mismatch between statistical assumptions and sensor phenomenology, or mismatch between training data and operational data, into the sensor algorithm or to even reliably predict a sensor's ultimate performance. Under Phase II, the Center for Surveillance Research will continue to address these shortcomings in existing sensor technology.