Activity Level Change Detection for Persistent Surveillance
Abstract
A new approach to GMTI data exploitation for large area persistent surveillance is presented. Instead of traditional target tracking, this approach utilizes GMTI data as moving spots on the ground to estimate the level of activities and detect unusual activities such as military deployments. A multilayer hierarchical exploitation scheme is proposed. This computational framework has clean interfaces between layers consisting of multiple processing modules. Various data processing, machine learning, and reasoning algorithms can be implemented in these modules. This system is easily extendable and can be tested using a generalized test bed. The development of two processing modules, vehicular volume and convoy detector, is described. For the vehicular volume module, US highway data were used as a surrogate of long-term GMTI surveillance data. The relationship between the activity level of Norfolk Naval Base and the traffic pattern on a road leading to the Base is studied. The convoy detection module, developed using real GMTI data, contains an algorithm that detects convoys without explicit target tracking. An end-to-end testing facility was also developed. Using this test bed, the system can he tested at different levels: as an individual processing module, as multiple cooperating processing modules across layers, or as the entire system.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 02, 2004
- Accession Number
- ADA457106
Entities
People
- Fangli Liu
- L. A. Bush
Organizations
- Massachusetts Institute of Technology