The study of self-organizing clustering neural networks and applications in data fusion
Dong Qiu, Longshan Wang, Wenfeng Bai, Jiafu Wang · 2008
The self-organizing clustering neural network, DIGNET, generally exhibits faster learning and better clustering performance. With a simple architecture and straightforward dynamics, DIGNET is more flexible regarding the choice of different metrics as measures of similarity. The system parameters in the DIGNET model are analytically determined from the self-adjusting process. A two-stage parallel multi-sensor data fusion system designed with DIGNET has been applied to the moving target detection. Experimental results on field data have shown that the multi-sensor DIGNET based data fusion systems successfully detect the moving target embedded in clutter. The generic two-stage DIGNET-based parallel fusion architecture can be applied to different one or two dimensional multi-sensor data fusion problems when the feature vectors are properly identified and extracted from the data.