Abstract: In order to improve traffic safety, both the scientific community and the automobile industry have contributed to the development of different types of safety systems. In the last decade, research has also moved towards more intelligent on-board systems that aim to anticipate and try to avoid or mitigate the severity of traffic accidents. These systems are referred to as Advanced Driver Assistance Systems (ADAS). Pedestrian Recognition is currently one of the most interesting ADAS for the automotive industry worldwide. This talk presents advanced techniques for vision-based pedestrian recognition in road images. A comprehensive combination of feature extraction methods is proposed and discussed. The basic components of pedestrians are first located in the image and then combined with SVM-based classifiers. Candidate pedestrians are located using an attention mechanism based on stereo vision. A components-based learning approach is proposed in order to better deal with pedestrians variability, illumination conditions, partial occlusions, and rotations. Extensive comparisons have been carried out using different feature extraction methods, as a key to image understanding in real traffic conditions. A database containing thousands of pedestrian samples extracted from real traffic images has been created for learning purposes, either at daytime and nighttime. The results achieved up to date show interesting conclusions that suggest a combination of feature extraction methods as an essential clue for enhanced detection performance.
 
 

Read More