基于超声相控阵的聚乙烯管道缺陷信号重构方法研究

在对聚乙烯管道缺陷进行超声检测的过程中,由于聚乙烯材料中传播的声速小,散射噪声强,信噪比极低,并且仪器设备本身会受到电信号干扰,从而影响缺陷成像的结果。因此针对A扫信号进行数据处理以提高检测图像的质量尤为重要。另一方面,采用阵元数较多的超声相控阵探头进行不同类型的聚乙烯管道缺陷的数据采集时,将会得到大量的缺陷数据,对存储、传输和处理带来各种困难。而针对传统方法进行压缩感知时,如果信号的信噪比较低而重构均方误差较大,则很难保留信号中重要信息,在低码率下更容易产生细节丢失的问题。所以本文提出一种基于K-SVD超完备字典学习的稀疏表示缺陷信号压缩重构方法,借助该学习算法训练过完备字典,并选择高斯随机矩阵为观测矩阵和正交匹配追踪算法(OMP)为重构算法对聚乙烯管道缺陷回波信号进行压缩感知,同时分析字典元素个数与迭代次数等参数变化对重构信号与成像效果的影响。

Research on Reconstruction Method of Polyethylene Pipeline Defect Signal Based on Ultrasonic Phased Array

In the process of ultrasonic testing of polyethylene pipe defects, due to the low speed of sound propagation in polyethylene materials, loud scattering noise, extremely low signal-to-noise ratio, and the electrical signal interference of the instrument and equipment, the result of defect imaging can be affected. Therefore, it is particularly important to process the A-scan signal to improve the quality of the detected image. On the other hand, when ultrasonic phased array probes with more array elements are used to collect data of different types of polyethylene pipe defects, a large number of defect data will be obtained, which brings various difficulties to storage, transmission and processing. However, when using traditional compression sensing methods, if the signal has a low signal-to-noise ratio and a large reconstruction mean square error, it is difficult to retain important information in the signal, and it is more likely to cause the problem of loss of details at low bit-rate. Therefore, this paper proposes a sparse representation defect signal compression and reconstruction method based on K-SVD super complete dictionary learning. With the help of this learning algorithm, the over complete dictionary is trained, and the Gaussian random matrix is selected as the observation matrix and the orthogonal matching pursuit algorithm (OMP) is selected as the reconstruction algorithm to process the compressedSsensing of the polyethylene pipe defect echo signal. At the same time, the influence of the number of dictionary elements and the number of iterations on the reconstructed signal and imaging effect is analyzed.