A Novel Pedestrian Detection Method Based on Histogram of Oriented gradient and Support Vector Data Description
- DOI
- 10.2991/jimec-18.2018.68How to use a DOI?
- Keywords
- histogram of oriented gradient; support vector data description; pedestrian detection; statistics learning
- Abstract
Pedestrian detection is an important research field of computer vision, and is a key technology of automatic drive. One category of pedestrian detection methods is statistics learning methods. But these methods cannot guarantee that the negative samples include all real scenes in practice. And big negative data lead to a complex classifier. Compared to negative samples, positive samples are more easily to be complete. To solve the above problem, we propose a novel pedestrian detection method based on Histogram of Oriented Gradient (HOG) and support vector data description (SVDD). First, create multi-scale samples by shift window. Second, samples are transformed by HOG method to provide features for classifiers. Third, a single classifier is trained with complete positive samples. The single classifier is based on SVDD. A series of experiments show that the proposed method needs more positive samples to train the classifier compared to previous works, but the number of total samples is less than that of previous works. And the false detection of the proposed method is less than that of classical method.
- Copyright
- © 2019, the Authors. Published by Atlantis Press.
- Open Access
- This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).
Cite this article
TY - CONF AU - Lei Ke AU - Yanping Dai PY - 2018/12 DA - 2018/12 TI - A Novel Pedestrian Detection Method Based on Histogram of Oriented gradient and Support Vector Data Description BT - Proceedings of the 2018 3rd Joint International Information Technology，Mechanical and Electronic Engineering Conference (JIMEC 2018) PB - Atlantis Press SP - 320 EP - 323 SN - 2589-4943 UR - https://doi.org/10.2991/jimec-18.2018.68 DO - 10.2991/jimec-18.2018.68 ID - Ke2018/12 ER -