Robust system identification using the Hough transform

The Hough transform provides a basis for robust extraction of shapes and continues to attract the interest of the image processing community. It has robustness properties desirable in many applications of pattern recognition, including parameter estimation. These properties are robustness to impulsive noise, insensitivity to partial occlusion of patterns, graceful degradation of performance in the presence of Gaussian and impulsive noise, as well as the ability to respond to concurrent patterns coexisting in the same data source. This technique has been applied to the problem of parameter estimation for linear and non-linear models. A critical comparison is made with the more traditional least-squares-based parameter estimators and it is argued that transform-based techniques are, in certain circumstances, more suitable for real-time intelligent control than those currently in use.