Pedestrian detection using quaternion histograms of oriented gradients
Guoyun Lian · 2020 IEEE International Conference on Power, Intelligent Computing and Systems (ICPICS) · 2020
In recent years, pedestrian detection has attracted more attention in many practical applications. In this paper, a novel pedestrian detection method using quaternion histograms of oriented gradients (QHOG) was proposed, which is integrated the advantages of both the quaternionic representation and HOG feature of a color image. Firstly, the quaternionic representation was performed on the color image in the sliding window, and then the histograms of oriented gradients (HOG) feature was extracted over the quaternionic representation map. After that, the QHOG histogram was constructed to represent the sliding window. Two groups of experiments are performed on two popular pedestrian datasets, INRIA dataset and Daimler Chrysler (DC) dataset, respectively. The experimental results show that our proposed QHOG detector performs better than the HOG detector, HOG-LBP detector and MWLD detector.