A New Paradigm for Persistent Wide Area Surveillance Lloyd L. Coulter, Douglas A. Stow, Yu Hsin Tsai, Christopher M. Chavis Department of Geography San Diego State University San Diego, CA, USA lcoulter@projects.sdsu.edu stow@mail.sdsu.edu cindyxtsai@gmail.com cchavis10@ucsbalum.com Richard W. McCreight NEOS Ltd. Prescott, AZ, USA neos500@gmail.com Christopher D. Lippitt Grant W. Fraley TerraPan Labs, LLC. La Mesa, CA, USA lippitt@terrapanlabs.com fraley@terrapanlabs.com Abstract— A novel and patent pending approach representing a new paradigm for persistent wide area surveillance is presented. As part of the Department of Homeland Security (DHS) National Center for Border Security and Immigration (BORDERS), San Diego State University (SDSU) researchers have developed a method for accurate and automated detection of people and vehicles moving through remote border regions or other uninhabited areas. Instead of imaging small areas at video frame rates (as with traditional surveillance), the approach uses a repeat pass, location-based image capture approach that trades time for space and enables repetitive imaging of large areas at lower repetition rates. Multiple targets may be detected and tracked over large areas, compared to video monitoring which focuses on small areas and often on individual targets previously detected using other means. The approach utilizes high frequency, repeat pass image collection (e.g., same imaging stations every 15 minutes) with frame array cameras to monitor large areas with a single aircraft and detect and locate objects moving through uninhabited landscapes. High frequency imaging from low-cost light aircraft is utilized to characterize the expected brightness response of each patch of ground corresponding to the ground resolution element of a pixel (3-inch ground sampling distance for this study), and an anomaly detection algorithm is used to detect subtle deviations from this expected response. Once anomalies (i.e., objects that have moved) are detected, small image chips (i.e., subsets) may be transmitted wirelessly to command and control stations so that the detection results may be visually verified in near real-time. The detection algorithm utilizes unique change detection thresholds per pixel, making the approach highly sensitive. Initial test results indicate that 98% of people and 100% of vehicles were correctly detected, with virtually no false detection (only 12 pixels within 19 images 21 megapixel in size). SDSU researchers, NEOS Ltd., and commercial partners are working to build a prototype system to further test and demonstrate this near real-time detection approach. This work is developed by the National Center for Border Security and Immigration: A Department of Homeland Security Science and Technology Center of Excellence. Keywords- wide area surveillance, change detection, airborne, video, real-time, pattern-of-life, UAV, UAS, border security I. INTRODUCTION The U.S. Customs and Border Protection (CBP) agency is responsible for securing the borders of the United States, and the Border Patrol specifically is responsible for patrolling the 10,000 kilometers of Mexican and Canadian international borders. Their general mission is to detect and prevent illegal entry of people and/or goods into the United States. The Border Patrol also performs a humanitarian mission, by rescuing people lost in remote locations and exposed to harsh environmental conditions. Since the terrorist attacks of September 11, 2001, the focus of the Border Patrol has expanded to include detection, apprehension and/or deterrence of terrorists and terrorist weapons. It is not practical, however, to closely monitor the tens of thousands of square kilometers of open land within close proximity of the border using agents and ground-based sensors alone. Airborne remote sensing offers the potential to monitor expansive areas within the border region, and identify activity of people/vehicles that has not been detected by agents patrolling the border or by ground-based sensors [1]. As part of the National Center for Border Security and Immigration, researchers with the Center for Earth Systems Analysis Research (CESAR) within the Department of Geography at SDSU have developed a novel approach for accurate and automated detection of people and vehicles moving through remote border regions or other uninhabited areas. This approach builds upon several years of experience with precise multitemporal image registration and detailed change detection. The people and vehicle detection approach utilizes time series imagery for characterizing expected scene/background response on a per-pixel basis, and then looks for subtle brightness changes that deviate from the expected response of each individual pixel. The general approach and initial results were first described in Coulter et al. (2012) [1]. The objective of this paper is to provide further insights into the characteristics and benefits of this time series change detection approach, and to discuss the next steps in the development and testing of the methodology. 978-1-4673-2709-1/12/$31.00 ©2012 IEEE 51