5. Conclusion
This paper has presented a Kalman filtering method for weld defect tracking in the radiographic inspection of spiral pipes. The study aimed to distinguish true defects from false alarms by the use of a potential defect tracking approach. An algorithm was developed and applied to typical sequences of radiographic images. Experimental results showed that the algorithm was capable of eliminating the false alarms by tracking the motion of the true defects. The method is proved insensitive to the change of the segmentation threshold of potential defect regions. Experiment also infers that the performance of the detection sustains even when the contrast level of the radiographic image sequence is instable. The proposed method also shows robustness to the variation of inspection condition and the instability of the detection velocity.