Stereo regions-of-interest selection for pedestrian protection: A survey

FERNANDEZ-LLORCA, DAVID; SOTELO, M.A.; HELLIN, A.M.; ORELLANA, ADRIÁN

INFORMATION PROCESSING LETTERS

ELSEVIER SCIENCE BV

Lugar: Amsterdam; Año: 2012 vol. I p. 226 - 237

0020-0190

Vision-based pedestrian detection for intelligent vehicles applications is a crucial andactive research area due to the essential benefits in terms of reducing the number of acci-dents involving pedestrians and vehicles. During the last decade a considerable amount ofresearch studies have been proposed, filling the gap between prototypes and commercialimplementations. Pedestrian detection systems can be roughly divided into three main dif-ferent sub-parts: Region Of Interest ? ROI ? selection, classification and tracking. Previous sur-veys have covered the literature in a holistic way. An example would be, analyzing all thesolutions proposed for all the stages and including higher level analysis, but in most casesthey give more emphasis to the classification stage. Due to the difficulty of this detectiontask, the variety of solutions, sensor configurations (monocular/stereo; visible/infrared)available in the literature, we propose to break down the variability of the problem by pro-viding exhaustive review of one specific stage: stereo-based ROI selection. ROI selection is akey component that has to be designed to provide generic obstacles at lowest false nega-tive rate and maintain a low number of false positives. The number of missed pedestrianshas to be approximately equal to 0 since a pedestrian missed by the ROI selection stagewould not be detected in further stages. In addition, the number of non-pedestrians obsta-cles should be as low as possible to reduce both the number of false alarms and the com-putational costs of further stages. In contrast to monocular approaches, stereo ROI selectiondetermines the relative distance between the pedestrian and the vehicle, assuring that thereported candidates are related with real physical objects. The stereo-based ROI selectionstep is also divided into different components that are independently analyzed, increasingvisibility for future proposals and developments. Discussion is finally presented highlight-ing the current problems for obtaining a global overview of the actual performance of thedifferent approaches and analyzing future trends.