压缩感知在电容层析成像流型辨识中的应用

Application of compressed sensing to flow pattern identification of ECT

针对传统电容层析成像(ECT)流型辨识方法识别率较低的问题,提出一种基于压缩感知理论的ECT流型辨识方法。首先,将ECT系统获得的测量电容向量归一化,并表示为训练样本集的过完备字典稀疏线性组合;然后,将随机高斯矩阵作为测量矩阵对测试样本和标准样本分别进行采样,并利用压缩感知信号重构算法求解L0范数下的最优化问题,从而得到各样本在训练样本集上的稀疏表示;根据待测样本和标准样本稀疏解之间的线性相关程度来确定归属流型。对典型流型的仿真实验结果显示,在无噪声、40dB、20dB信噪比的情况下,流型辨识准确率分别为100%、99.25%和98.12%,表明本文方法抗噪声干扰能力强,是一种有效、准确率较高的ECT流型辨识方法,为ECT流型辨识技术的研究提供了一种新的手段。

In view of lower recognition rates of traditional methods in the flow pattern identification of Electrical Capacitance Tomography （ECT）, a identification method for the ECT based on Compressed Sensing （CS） was put forward. Firstly, measurement capacitance vectors obtained by an ECT system were normalized and represented as a sparse linear combination of training sample set in an over com- plete dictionary. Then,the random Gaussian matrix was taken as the measurement matrix to sample from the test and standard samples respectively, and the signal reconstruction algorithm based on the CS was used to solve the optimization problem of L0 norm for the sparse representation of each sample on the training sample set. The linear correlation coefficient between the sparse solutions of samples to be tested and the standard samples are calculated to determine the classification of flow pattern. The simulation experiment results of typical flow patterns indicate that the flow pattern identification rates under absence of noise, and signal to and 98.12% respectively. It concluds that Noise Ratios（SNRs） of 40 db and 20 db are 100%, 99.25% the flow pattern identification method proposed has high efficiency and accuracy and a good noise immunity, which also provides a new method for the flow pat tern identification of ECT.