Robust online detection of pipeline corrosion from range data

We present the Finite-Window Robust Sequential Estimator for the detection and analysis of corrosion in range images of gas pipelines. This statistica lly robust, real-time technique estimates the pipeline surface range functi on in the presence of noise, surface deviations, and changes in the underly ing model. Deviations from the robust surface fit, corresponding to statist ical outliers, represent potential areas of corrosion. Because the algorith m estimates surface parameters over a finite, sliding window of data, it ca n track moderately high-order surfaces using lower order models. The system is consistent, objective, and non-destructive and can be used with the pip eline in service.