Non-metric lens distortion correction using modified particle swarm optimisation

摘要: Lens distortion correction is crucial to obtain the best performance for vision system where rigorous accuracy demanded. This paper presents a robust non-metric method which relies on fact that straight lines in real world must always project image under pinhole camera model. Neither calibration objects nor any known structure are required. Two types of measurement defined as objective function, and then optimisation algorithms adopted find parameters. The comparison analysed between traditional algorithm, standard particle swarm (PSO) algorithm two other modified based different typical strategies. Experiment results show opposition learning-based (OPSO) with mutation operator has strong robustness, case low-level noise, better, effectiveness verified by degrees lens distortion.