摘要
针对ECT系统SVM图像重建算法在处理大规模样本数据集时,成像精度不高及速度慢的问题,提出轮换对称分块支持向量机RSPSVM算法。算法对ECT系统模型进行具有轮换对称性的等面积剖分,使得通过一个单元即可得到同层其他所有单元的敏感度值;再选择性分块,形成可分别应用SVM算法进行训练的小样本矩阵,用得出的决策函数进行样本预测;并采用FPGA硬件实现RSPSVM算法。图像重建实验结果表明,通过硬件实现的RSPSVM算法大大减少了执行时间,并提高了成像的精度。
According to Support Vector Machine（SVM）has low accuracy and low training speed to deal with large scale data in ECT system, a new algorithm that combined SVM with the Rotation Symmetric Partition（RSPSVM）is presented.Model of ECT system is partitioned into equal area units with rotation symmetry, sensitivities of all units in the same layer can be obtained through one unit. Sample matrix is segmented selectively into smaller sample matrixes for training, and RSPSVM algorithm is implemented with FPGA hardware. Experimental results show that the algorithm implemented with FPGA hardware can not only improve classification accuracy of reconstruction image, but also have a shorter time in imaging.
出处
《计算机工程与应用》
CSCD
北大核心
2016年第11期204-208,共5页
Computer Engineering and Applications
基金
黑龙江省教育厅基金(No.12521100)
黑龙江省自然科学基金(No.F2015038)
哈尔滨市优秀学科带头人基金(No.2013RFXXJ034)
关键词
电容层析成像
支持向量机
轮换对称
选择分块
现场可编程门阵列
electrical capacitance tomography
support vector machine
rotation symmetry
choice and segmentation
field programmable gate array
