In this approach, a new method for classification and
clustering in the rosette pattern is introduced. In the IR (Infrared)
rosette pattern seeker, the obtained IR signal would be sampled and samples
in the rosette pattern space would be reconstructed and mapped into a
new space. Then samples in the new space are clustered and classified
and target class will be detected. Finally, target position would be determined
and central gravity of target is computed in order to track the target.
The advantage of mapping to this new space which we are named BRMS (Binary
Rosette Mapping Space) is that the data in BRMS can be clustered, classified
and central gravity computed easily. Process time which is an important
parameter in RSIS (Rosette Scanning Infrared Seeker) will be improved
compared with previous methods and the effect of the rosette pattern nonlinearity
will be decreased. In the next step, features of clusters are extracted
by PCA (Principle Component Analysis), ICA (Independent Component Analysis)
or LDA (Linear discriminant Analysis), instead of clusters size or intensity
in the rosette pattern and then features would be classified by MLP (Multy
Layer Perceptron), RBF (Radial Basis Function) and ART (Adaptice Resonance
Theory) neural networks and the results will be compared.
Hadi Soltanizadeh and Shahriar Baradaran Shokouhi, 2008. Feature Extraction and Classification of Objects in the Rosette Pattern Using Component Analysis and Neural Network. Journal of Applied Sciences, 8: 4088-4096.