In this paper we present the automatic real time segmentation algorithm we devised to be consistent with human visual perception for a highly contrasted scene, like the one generated by the projection of the luminous profiles from high power sources on a uniform untextured pattern. An accurate identification of shadow-light profiles is required, for example, from industrial diagnostics of light sources, in compliance with regulations for their employment by human users. Off-the-shelf CCD technology, though it could not be able to cover the wide dynamic range of such scenes, could be successfully employed for the geometric characterization of these profiles. A locally adaptive segmentation algorithm based on low-level visual perception mechanisms has been devised and tested in a very representative case study, i.e the geometrical characterization of beam profiles of high power headlamps. The evaluation of our method has been carried out by comparing (according to a curve metric) the extracted profiles with the ones pointed out by five human operators. The experiments prove that our approach is capable of adapting to a wide range of luminous power, mimicking visual perception correctly even in presence of low SNR for the acquired images.
|Name||Lecture Notes in Computer Science|
|Publisher||Springer Berlin Heidelberg|
|Conference||6th International Conference on Image Analysis and Recognition 2009|
|Abbreviated title||ICIAR 2009|
|Period||6/07/09 → 8/07/09|