Detecting pedestrians is still a challenging task for automotive vision system due the extreme variability of targets, lighting conditions, occlusions, and high speed vehicle motion. A lot of research has been focused on this problem in the last 10 years and detectors based on classifiers has gained a special place among the different approaches presented. This work presents a state-of-the-art pedestrian detection system based on a two stages classifier. Candidates are extracted with a Haar cascade classifier trained with the DaimlerDB dataset and then validated through part-based HOG classifier with the aim of lowering the number of false positives. The surviving candidates are then filtered with a feature-based tracking to enhance the recognition robustness and improve the results' stability. The system has been implemented on a prototype vehicle and offers high performance in terms of several metrics, such as detection rate, false positives per hour, and frame rate. The novelty of this system rely in the combination of HOG part-based approach, tracking based on specific optimized feature and porting on a real prototype.
15th International Conference on Intelligent Transportation Systems, 2012, p. 73-77
Main Research Area:
Intelligent Transportation Systems Conference, 2012