HANDFIT: An Algorithm for Automatic Fitting of Continuous Piecewise Regression, with Application to Feature Extraction from Remote Sensing Time Series Data

摘要

This work presents the algorithm HANDFIT, an iterative method for continuous piecewise regression with automatic change-points estimation. From an initial guess about the number and positions of the changepoints or hinges, they are iteratively adjusted by Newton-like displacements, with very fast convergence in most cases. The algorithm allows for sufficiently close hinges to be identified, thus reducing the number of changepoints. Examples of applications to feature extraction from remote sensing vegetation indices time series data are presented.