期刊名称：International Journal of Grid and Distributed Computing
印刷版ISSN：2005-4262
出版年度：2016
卷号：9
期号：3
页码：91-98
DOI：10.14257/ijgdc.2016.9.3.11
出版社：SERSC
摘要：Due to the problem of low imaging accuracy and slow imaging speed when applying SVM image reconstruction algorithm in ECT system to dealing with a large amount of sample data set, the method of combining feature dimension reduction with SVM algorithm is proposed. This method classifies the sample data by using the way of clustering and extracts the feature parameter, finds out the connection between each sample and the feature, and deals the sample data with dimension reduction, thus finally getting the high-quality training sample. Then it trains the simplified sample data by applying SVM algorithm and obtains decision function, then the decision function is used to predict and image. The experimental results of image reconstruction show that this method greatly reduces the running time and improves the accuracy of imaging compared to using the SVM algorithm alone.
