[Legacy Report] Performance of object classification for detection of pedestrians

#night #vision #video #processing #object #detection #multi-branch #classifier #pedestrian #Adaboost #histogram #of #oriented #gradients

An analysis of object classification during the process of detection of pedestrians in night vision infrared images was presented. The classical solution with one classifier was tested and some methods for increasing the performance are proposed. A modified multi-branch classifier is presented to increase performance of the pedestrian detection system. The solution is accurate, fast, and especially good suited for all real-time applications where pedestrians may appear in many distances to the camera. The obtained results show an increased efficiency of the classification process (by up to 3%) with similar processing time in comparison to a single classifier. All tests were conducted using the Adaboost classifier, but generally, the results should be consistent also for other types of classifiers.

• Add Event to Calendar
  iCal
  Google Calendar

• Poznan, Wielkopolskie
• Poland

---

• Contact Event Host