Song Yongfeng, Turner Joseph A, Peng Zuoxiang, Chen Chao, Li Xiongbing
IEEE Trans Ultrason Ferroelectr Freq Control. 2018 Jul;65(7):1214-1225. doi: 10.1109/TUFFC.2018.2827464.
In an attempt to improve the ultrasonic testing capability of a conventional C-scan system, a flaw detection method using an ultrahigh gain is developed in this paper. A time-dependent threshold for image segmentation is applied to identify automatically material anomalies present in the sample. A singly scattered response model is used with extreme value statistics to calculate the confidence bounds of grain noise. The result is a time-dependent threshold associated with the grain noise that can be used for segmentation. Ultrasonic imaging experiments show that the presented method has advantages over a traditional fixed threshold approach with respect to false positives and missed flaws. The results also show that a low gain is adverse to the detection of microflaws with subwavelength dimensions. The forward model is expected to serve as an effective tool for the probability of detection of flaws and the inspection of coarse-grained materials in the future.
为了提高传统C扫描系统的超声检测能力，本文开发了一种使用超高增益的探伤方法。应用随时间变化的阈值进行图像分割，以自动识别样品中存在的材料异常。使用单散射响应模型和极值统计来计算颗粒噪声的置信界限。结果是得到一个与颗粒噪声相关的随时间变化的阈值，可用于分割。超声成像实验表明，该方法在误报和漏检方面优于传统的固定阈值方法。结果还表明，低增益不利于检测亚波长尺寸的微缺陷。该正向模型有望在未来成为检测缺陷概率和检测粗晶材料的有效工具。
