A neural network based approach for classifying critical mobile targets from multisensor data /
A methodology was developed to evaluate the feasibility of neural network identity fusion techniques for classifying vehicles from multisensor data collected from unattended ground sensors. A high fidelity computer simulation called Sensor Evaluation Model (SENSEM), was used to create a battlefield...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
1996.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=739363351&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | A methodology was developed to evaluate the feasibility of neural network identity fusion techniques for classifying vehicles from multisensor data collected from unattended ground sensors. A high fidelity computer simulation called Sensor Evaluation Model (SENSEM), was used to create a battlefield scenario realistically simulating sensorvehicle interactions consisting of three sensor types and four vehicle types. The sensors used multiple sensing techniques and simulated state-of-the-art, unclassified operational characteristics. The vehicles consisted of three non-target vehicle types and one target vehicle type representing a Scud launching vehicle. Data generated from sensor-vehicle interactions was preprocessed and used to train and test three types of neural networks. Backpropagation, Probabilistic, and Radial Basis Function networks were evaluated to determine differential performance effectiveness under various training conditions (i.e., five training sets containing different amounts of category exemplars). Optimal network designs were experimentally determined for each network type for each training set. An overall comparative analysis of classification accuracy and computational efficiency was then conducted to evaluate the performance between the network types' optimal network designs for each training set. Results showed extremely high accuracy and rapid training and test times for all optimal network designs. Results also showed optimal network performance varied as a function of training conditions facilitating specification of the most effective neural network paradigm under certain conditions. The results of this research support the conclusion that neural network techniques can be applied successfully as part of an analysis subsystem for classifying vehicles from multisensor data obtained from unattended ground based sensors. The extremely high classification accuracy results demonstrate the ability to successfully classify vehicles from sensor responses when the data contains both imbalances in the number of category training exemplars as well as small numbers of target exemplars. The rapid training times support the use of these neural networks in an environment critically dependent upon the ability to rapidly update the analysis subsystem based on new battlefield information. Finally, the notably rapid test set evaluation time supports the use of these networks in an environment requiring rapid classification of vehicles from sensor responses. |
|---|---|
| Item Description: | Vita. "Major Subject: Industrial Engineering". |
| Physical Description: | xii, 320 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilms Inc. |
| Bibliography: | Includes bibliographical references: pages 213-225. |