Detection of ultrasonic NDE signals using time-frequency analysis

This paper proposes a detection procedure based on the time-frequency analy sis of the real defect signals. The multicomponent nature of NDE signals mo tivates the use of better visual as well as numerical techniques to detect the defects. In particular, the detection becomes extremely difficult in th e presence of noise or if the defects are very closely located. The signals used in this paper correspond to artificially produced defects of various geometries in solid aluminium blocks. An appropriate time-frequency represe ntation (TFR) is used to represent the signals in the time-frequency plane. After the TFR is obtained, various image processing morphological techniqu es are applied to the TFR, for example region of interest (ROI) masking, ed ge detection, and profile separation. The resulting binary image can be eff ectively used to detect and locate the defects.