摘要
针对在处理大规模样本集的ECT系统数据时,SVM算法存在的图像重建的成像精度不高及速度慢的问题,采用了轮换对称分块支持向量机CSPSVM算法.算法利用ECT系统模型的轮换对称性,将大样本矩阵按照成像单元某一层按轮换对称性进行简化,并选择性分块,形成多个小样本矩阵;然后分别采用SVM算法进行训练,用得出的决策函数进行样本预测;最后将各成像单元组合成像.图像重建实验结果表明使用CSPSVM改进算法要比单独使用SVM算法重建图像具有更高的分类精度和更短的成像时间.
According to support vector machine （SVM） has low accuracy and low training speed to deal with large scale sample matrix in ECT system, a new algorithm that combined SVM with the cyclic symmetrical partition （CSPSVM） is presented. By the cyclic symmetry of ECT system model, a large sample matrix is simplified accord- ing to a layer of the imaging unit, and segments block selectively into multiple smaller sample matrixes. Then they are trained by SVM respectively, and the decision function obtained can be used to classify the prediction sample. Finally, all prediction units are combined for imaging. Experimental results show that the image reconstruction u- sing CSPSVM algorithm has higher classification accuracy and shorter imaging time than using SVM alone.
出处
《哈尔滨理工大学学报》
CAS
北大核心
2015年第3期40-44,共5页
Journal of Harbin University of Science and Technology
基金
黑龙江省教育厅科学技术研究项目(12521100)
黑龙江省自然科学基金(F2015038)
哈尔滨市优秀学科带头人基金(2013RFXXJ034)
关键词
电容层析成像
支持向量机
轮换对称性
选择分块
图像重建
electrical capacitance tomography
support vector machine
cyclic symmetry
choice and segmen-tation
image reconstruction
