Research on the Multi-sensors Information Fusion Technique Based on the Neural Networks and its Application

• Published in: 2009 Second International Workshop on Knowledge Discovery and Data Mining pp. 93 - 96
• Main Authors: Minghu Zhang, Hui Song, Shijun Lv, Youfeng Li, Xiaojuan Yu, Jingyang Bao
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.01.2009
• Subjects: Acoustic sensors; Chemical analysis; Chemical sensors; Chemical technology; Design of the system; Feature extraction; Gas detectors; Intelligent sensors; MSIF; Neural networks; NN model; Sensor phenomena and characterization; Sensor systems; Wavelet analysis
• ISBN: 9780769535432; 0769535437
• DOI: 10.1109/WKDD.2009.137

Multi-sensors information fusion (MSIF) technology, is being widely applied to various fields, particularly, the modern military field. In order to enhance the capability of ship chemical defense support in the future informationalization sea warfare, a review of the study and application about MSIF technology in naval ships chemical detection (NSCD) field is researched, the model of NSCD system based on MSIF is built. Its measure system based on multi-sensors fusion could catch the wide information of the chemical agents, carry the feature extraction and selection of the chemical agents through wavelet analysis, then make the best of the Neural Networks to manage the data from multi-sensors system, thus the neural networks distinguishing chemical agents (NNDCA) model is built. The realization idea of the NNDCA system is put forward, and the hardware accomplishment and the software structure of the NNDCA system are discussed. The experimental and emulational results show that: it is entirely feasible that using the NNDCA model put up the qualitative and quantitative analysis of the chemical agents; the NNDCA model is capable of, in a great measure, playing down the impact factor of the disturber, concentration and condition, and so on, to measure the chemical agents, as a result, remarkably, the veracity and creditability of measure effect is heightened.