Real-time road segmentation using a morphological multi-pipeline processor

This paper presents the image processing sub-module, part of the in-car embedded system, designed by the CMM in the framework of the European PROMETHEUS project. In this project, the authors´ task is to extract road information in dynamic scenes. In particular, the authors have to perform road/lane segmentation and obstacle detection in real-time. To reach this goal, the authors have studied and developed dedicated algorithms and a multi-pipeline processor especially designed for their implementation. The morphological transformations are supported by two pipeline processors built around powerful morphological ASICs: PIMM1. The road/lane segmentation algorithm runs in background on the main pipeline processor. The second pipeline processor is dedicated either to obstacle detection or to on-request processing such as indicator detection. A very important item of the morphological sub-module, the multi-access image memory board, supplies the two pipeline processors with image data. The intelligent board working under the multi-tasking VxWorks operating system configures via a VSB bus the morphological sub-module. Such a multi-pipeline processor allows very high speed image processing compatible with real-time performances.