Adaptive infrared target detection
Automatic Target Recognition (ATR) algorithms are extremely sensitive to differences between the operating conditions under which they are trained and the extended operating conditions (EOCs) in which...
Using image local response for efficient image fusion with the hybrid evolutionary algorithm
Image fusion serves as the basis for automatic target recognition; it maps images of teh same scene received from different sensors into a common reference system. A novel fusion method is described ...
Selectionof the kernel parameters is critical to the performance ofSupport Vector Machines (SVMs), directly impacting the generalization and classificationefficacy of the SVM. An automated procedure for parameterselection is clearly desirable given the intractable problem of exhaustivesearch methods. The authors' previous work in this areainvolved analyzing the SVM training data margin distributions for aGaussian kernel in order to guide the kernel parameter selectionprocess. The approach entailed several iterations of training theSVM in order to minimize the number of support vectors. Our continued investigation of unsupervised kernel parameter selection hasled to a scheme employing selection of the parameters beforetraining occurs. Statistical methods are applied to the Grammatrix to determine kernel optimization in an unsupervised fashion. This preprocessing framework removes the requirement for iterative SVM training. Empirical results will be presented for the "toy" checkerboardand quadboard problems.
By Year
You are not logged in to this journal. 
Connotea
CiteULike
del.icio.us
BibSonomy
