DOI: 10.7763/IJCCE.2012.V1.54
Fast Pedestrian Detection Based on Haar Pre-Detection
Abstract—We proposed a simple but efficient way to accelerate the traditional pedestrian process without decrease the detection accuracy. Haar/adaboost is one of the fastest ways in pattern recognition, and HOG/libSVM is an accurate way to deal with pedestrian detection. Haar/adaboost can work fast using low resolution images, but HOG/libSVM, which was proposed by Dalal in 2005, is much slower due to the high resolution detection window and complicated feature extraction. Inspired by the above results, we combine these two most popular ways to develop a better pedestrian detection system. And due to the fast small scale Haar/adaboost pre-detection, experiment result shows that the time our method consumed is about one third of the original HOG/LibSVM method, while maintaining almost the same detection rate.
Index Terms—Small scale, HOG, HAAR, fast pedestrian detection
The authors are with School of Computer and Information Engineering, Peking University Shen Zhen Graduate School, China (e-mail: springs@126.com)
Cite: Wenfeng Xing, Yong Zhao, Ruzhong Cheng, Jiaoyao Xu, Shaoting Lv, and Xinan Wang, "Fast Pedestrian Detection Based on Haar Pre-Detection," International Journal of Computer and Communication Engineering vol. 1, no. 3, pp. 207-209 , 2012.
General Information
-
Dec 29, 2021 News!
IJCCE Vol. 10, No. 1 - Vol. 10, No. 2 have been indexed by Inspec, created by the Institution of Engineering and Tech.! [Click]
-
Mar 17, 2022 News!
IJCCE Vol.11, No.2 is published with online version! [Click]
-
Dec 29, 2021 News!
The dois of published papers in Vol. 9, No. 3 - Vol. 10, No. 4 have been validated by Crossref.
-
Dec 29, 2021 News!
IJCCE Vol.11, No.1 is published with online version! [Click]
-
Sep 16, 2021 News!
IJCCE Vol.10, No.4 is published with online version! [Click]
- Read more>>