Eddy current signal defect detection algorithm based on improved SSD network

The main function of eddy current testing is to compare the results of magnetic flux leakage nondestructive testing. By analyzing the defects, it is possible to analyze whether the position of the defect is inside or outside the pipeline. The noise interference is so severe that the conventional threshold method cannot be adapted to the eddy current signal. This paper proposes a method to convert eddy current signals into pseudo-color images. The pseudo-color image obtained by eddy current signal conversion has high resolution and clear defect features, which is suitable for labeling and network training. This paper also proposes an improved single shot detection (SSD) algorithm to better solve the problem of misclassification of external defects. The method proposed in this paper is more applicable to low-noise pipes. The results show that the method proposed in this paper can improve the positioning accuracy of eddy current data.