Computer Engineering and Applications ›› 2017, Vol. 53 ›› Issue (14): 167-171.DOI: 10.3778/j.issn.1002-8331.1512-0378

Previous Articles     Next Articles

Improved integral channel features for fast multiscale pedestrian detection

HUANG Peng, YU Fengqin, CHEN Ying   

  1. School of Internet of Things Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China
  • Online:2017-07-15 Published:2017-08-01


黄  鹏,于凤芹,陈  莹   

  1. 江南大学 物联网工程学院,江苏 无锡 214122

Abstract: In order to overcome the problem that integral channel features have redundant information and slow detection speed in multiscale pedestrian detection, an improved integral channel features for fast multiscale pedestrian detection is proposed. Firstly, fast feature pyramids are used to compute multiscale channel features, which avoid computing the same features at multiple locations and scales. Then the detection windows are divided into cells and blocks to make the overall description of the image and reduce the redundancy, replacing the original method of random location and size. Finally, it obtains the sums of pixels in cells and blocks as the pedestrian features and then the features are classified by soft cascade Adaboost. Simulation experiment results show that the proposed method gives improved accuracy and 15.4 times speedup compared with the original algorithm and runs at 18.5 f/s on 640×480 images.

Key words: pedestrian detection, integral channel features, multi-scale, fast feature pyramids

摘要: 针对积分通道特征冗余信息多,在多尺度行人检测中检测速度较慢的问题,提出了改进积分通道特征的快速多尺度行人检测算法。该方法首先采用快速特征金字塔计算图像不同尺度下的特征通道,避免对图像重复缩放计算特征通道;然后将检测窗口分为单元和块来对图像进行整体描述,代替原始方法的随机位置和大小矩形来减少冗余特征,最后计算单元和块内的像素和作为特征向量送入软级联Adaboost分类器进行分类。仿真实验结果表明,该算法检测精度优于积分通道特征算法,同时检测速度提高了15.4倍,在640×480大小图像上检测速度达到18.5 f/s。

关键词: 行人检测, 积分通道特征, 多尺度, 快速特征金字塔