关键词：传感器；探测器；网络中心；热电性；回归分析
摘 要：The University of Memphis conducted basic research into techniques for advancing network-centric sensors for eventual deployment in Department of Defense (DoD) applications. This basic research included the following focus sub-areas: (i) feature fusion/feature-based sensor system design techniques; (ii) sensor ontologies for problemsolving architectures; (iii) profiling sensor improvement through the use of innovative classification algorithms and data visualization techniques; (iv) alternative sensing modalities; (v) turbulence mitigation techniques; and (vi) development of a feature sensing laboratory. Under the topic of feature fusion/feature-based sensor system design, techniques known as Lasso, Group Lasso, and Sparse Multiple Kernel Learning were applied to break beam profiling sensor design. The results indicate that the Group Lasso technique is effective for feature quality maximizing sensor design because of its ability to provide both intergroup and intra-group feature sparsity. Under the topic of sensor ontologies for problem-solving architectures, a framework that matches sensors to compatible algorithms to fonn synthesized systems was developed and applied to improved forms of the beam-break profiling sensor. This work resulted in several publications. Under the topic of profiling sensor improvement, various algorithms for improving the classification performance of a pyro-electric based profiling sensor were investigated and tested using data from field collections. Results indicated that Logistic regression with a simple height to width ratio provide good performance.