Ultrasound anomaly signal identification method based on local similarity feature
The recognition of ultrasound abnormal signals plays a crucial role in reducing defect detections,improving defect detection and inversion accuracy.However,due to the complex and diverse patterns of ultrasound abnormal signals,sample labeling cannot exhaust their patterns,resulting in low accuracy of current abnormal signal recognition methods.To further improve the recognition accuracy of ultrasound abnormal signals,a local similarity based ultrasound abnormal signal recognition method is proposed.Firstly,a local similarity feature extraction method based on dynamic time warping was designed,which can still extract effective and separable features without using ultrasound abnormal signal samples.Then,based on the local similarity feature vectors,a single class support vector description algorithm is used for anomaly recognition.The experimental results show that the area under the working curve of the subjects using this method is 0.995,which is 0.18-0.61 higher than traditional unsupervised anomaly recognition algorithms.It has strong application value in practical engineering.
万方数据
维普
