Underwater Target Fusion Recognition Based on Non-linear Dimensionality Reduction
Yue Zhou · Microcomputer applications · 2008
Aiming at the difficulty of handling high dimension data of underwater target feature extracted by tradition methods of spectrum and model parameters,this paper presents a novel underwater targets recognition method by introducing non-linear reduction method,which can effectively eliminate the correlation between vectors of eigenvector,and therefore decrease the computation complexity of post-phase recognition. Additionally,we designed a fusion decision classifier called DS-SVM classifier to classify feature vectors gained from non-linear dimensionality reduction.The experimental results illustrated the superior performance of the proposed method compared with traditional SVM classifier.