相控阵超声缺陷检测数据压缩感知

为了解决超声相控阵信号采集、存储和传输中数据量大的问题,研究了压缩感知在相控阵无损检测信号和图像压缩重构中应用的可行性。首先使用5种贪婪算法对相控阵仿真信号进行压缩重构,根据百分比均方误差选取最优算法并考虑了噪声对精度的影响,结果表明压缩感知可以用低于奈奎斯特极限的测量点数准确重构原始图像;其次用人工缺陷回波信号进行实验验证,通过稀疏度计算选择适用相控阵信号的最优稀疏基,并通过5种传感矩阵的优化选择进一步提高了重构精度。实验结果尽管达不到仿真中的理想效果,但是能以少量测量值准确恢复图像,并能保证缺陷的识别,说明压缩感知算法可以有效提高相控阵缺陷检测效率。此外,在保持测量点数相同的情况下,仿真和实验都研究了不同采样率对重构精度的影响,当测量点数超过一定值时,证实了压缩感知实际与采样率无关。 

Ultrasonic phased array signal compressed sensing in defect detection

A Compressed Sensing(CS)based procedure specifically dedicated to phased array signals and images in ultrasonic nondestructive detection is proposed,aiming to reduce a huge amount of data collection,storage,and in some instances data transmission.First of all,five greedy algorithms were employed to reconstruct simulated signals.The optimal one was selected according to the percent residual difference,and the impact of noise level was also considered.An excellent image recovery was reached using measurement points less than the Nyquist sampling limitation required.The experimental verification was presented on echo signals reflected from artificial defects.Through a quantitative sparsity calculation,the most suitable sparse basis was chosen,and the reconstruction performance was further improved via five sensing matrices selection.Although not as impressive as the simulated results,CS algorithm verified its efficiency and practicability in ultrasonic array image reconstruction using a small amount of samples,with defects fully recognizable.In addition,both in simulation and experiment,we studied the influence of Sampling Rate(SR),keeping the measurement points same.The results imply that,when measurement points increase to some extent,CS is independent with SR.