首页 > 论文 > 激光与光电子学进展 > 55卷 > 9期(pp:91504--1)

基于改进全卷积网络的多尺度感知行人检测算法

Multi-Scale Aware Pedestrian Detection Algorithm Based on Improved Full Convolutional Network

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

当前行人检测的一个主要挑战是在复杂的场景中检测出不同尺度的行人, 尤其是远尺度行人。由于不同尺度的行人会表现出不同的视觉外观特征, 鉴于此提出了一种多尺度感知的行人检测算法。在全卷积网络结构中引进可形变卷积层, 扩大特征图的感受野; 通过级联区域建议网络提取多尺度行人建议区域, 引入多尺度判别策略, 定义尺度判别层, 判别行人建议区域的尺度类别; 构建一个多尺度感知网络, 引进软非极大值抑制(Soft-NMS)检测算法, 融合每个网络输出的分类值和回归值, 获取最终的行人检测结果。实验表明, 本文提出的检测算法在基准数据集Caltech和ETH上的检测误差较低, 检测精度优于当前其他检测算法, 适用于检测远尺度行人。

Abstract

A major challenge of pedestrian detection is to detect different-scale pedestrians in complicated scenarios, especially for far-scale pedestrians. Motivated by the experiment that pedestrians with different scales exhibit dramatically different visual features, we propose in this paper a multi-scale aware pedestrian detection algorithm. Firstly, we introduce deformable convolutional layers in full convolutional network structure to expand the receptive field of feature maps. Secondly, we use cascade-region proposal network to extract multi-scale pedestrian proposals and introduce discriminant strategy, and define a multi-scale discriminant layer to distinguish pedestrian proposals category. Finally, we construct a multi-scale aware network, use the soft non-maximum suppression algorithm to fuse the output of classification score and regression offsets by each sensing network to generate final pedestrian detection regions. The experiments show that there is low detection error on the datasets Caltech and ETH, and the proposed algorithm is better than the current detection algorithms in terms of detection accuracy and works particularly well with far-scale pedestrians.

Newport宣传-MKS新实验室计划
补充资料

中图分类号:TP391.4

DOI:10.3788/lop55.091504

所属栏目:机器视觉

基金项目:教育部—中国移动科研基金(MCM20182019)

收稿日期:2018-03-23

修改稿日期:2018-04-11

网络出版日期:2018-04-17

作者单位    点击查看

刘辉:江南大学物联网工程学院, 物联网应用技术教育部工程中心, 江苏 无锡 214122
彭力:江南大学物联网工程学院, 物联网应用技术教育部工程中心, 江苏 无锡 214122
闻继伟:江南大学物联网工程学院, 物联网应用技术教育部工程中心, 江苏 无锡 214122

联系人作者:闻继伟(wjw8143@aliyun.com); 刘辉(1391570995@qq.com);

【1】Dollár P, Tu Z, Perona P, et al. Integral channel features[C]∥British Machine Vision Conference, 2009: 7-10.

【2】Nam W, Dollár P, Han J H. Local decorrelation for improved pedestrian detection[C]∥Advances in Neural Information Processing Systems, 2014: 424-432.

【3】Dollár P, Appel R, Belongie S, et al. Fast feature pyramids for object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 36(8): 1532-1545.

【4】Dalal N, Triggs B. Histograms of oriented gradients for human detection[C]∥IEEE Conference on Computer Vision and Pattern Recognition, 2005: 886-893.

【5】Qin J, Wang M H. Fast pedestrian proposal generation algorithm using online Gaussian model[J]. Acta Optica Sinica, 2016, 36(11): 1115001.
覃剑, 王美华. 采用在线高斯模型的行人检测候选框快速生成方法[J]. 光学学报, 2016, 36(11): 1115001.

【6】Tian Y, Luo P, Wang X, et al. Pedestrian detection aided by deep learning semantic tasks[C]∥IEEE Conference on Computer Vision and Pattern Recognition, 2015: 5079-5087.

【7】Tian Y, Luo P, Wang X, et al. Deep learning strong parts for pedestrian detection[C]∥IEEE International Conference on Computer Vision, 2015: 1904-1912.

【8】Ye G L, Sun S Y, Gao K J, et al. Nighttime pedestrian detection based on faster region convolution neural network[J]. Laser & Optoelectronics Progress, 2017, 54(8): 081003.
叶国林, 孙韶媛, 高凯珺, 等. 基于加速区域卷积神经网络的夜间行人检测研究[J]. 激光与光电子学进展, 2017, 54(8): 081003.

【9】Ren S, He K, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[C]∥Advances in Neural Information Processing Systems, 2015: 91-99.

【10】Liu W, Anguelov D, Erhan D, et al. SSD: single shot multibox detector[C]∥European Conference on Computer Vision, 2016: 21-37.

【11】Li J, Liang X, Shen S M, et al. Scale-aware fast R-CNN for pedestrian detection[J]. IEEE Transactions on Multimedia, 2017, 20(4): 985-996.

【12】Cai Z, Fan Q, Feris R S, et al. A unified multi-scale deep convolutional neural network for fast object detection[C]∥European Conference on Computer Vision, 2016: 354-370.

【13】Hou C C, He Y Q, Jiang X H, et al. Deep convolutional neural network based on two-stream convolutional unit[J]. Laser & Optoelectronics Progress, 2018, 55(2): 021005.
侯聪聪, 何宇清, 姜晓恒, 等. 基于二分支卷积单元的深度卷积神经网络[J]. 激光与光电子学进展, 2018, 55(2): 021005.

【14】Girshick R. Fast R-CNN[C]∥IEEE International Conference on Computer Vision, 2015: 1440-1448.

【15】Dai J, Li Y, He K, et al. R-FCN: object detection via region-based fully convolutional networks[C]∥Advances in Neural Information Processing Systems, 2016: 379-387.

【16】Bodla N, Singh B, Chellappa R, et al. Soft-NMS-improving object detection with one line of code[C]∥IEEE International Conference on Computer Vision, 2017.

【17】Luo P, Tian Y, Wang X, et al. Switchable deep network for pedestrian detection[C]∥IEEE Conference on Computer Vision and Pattern Recognition, 2014: 899-906.

【18】Ouyang W, Wang X. Joint deep learning for pedestrian detection[C]∥IEEE International Conference on Computer Vision, 2014: 2056-2063.

【19】Zhang L, Lin L, Liang X, et al. Is faster R-CNN doing well for pedestrian detection?[C]∥European Conference on Computer Vision, 2016: 443-457.

【20】Yang B, Yan J, Lei Z, et al. Convolutional channel features[C]∥IEEE International Conference on Computer Vision, 2015: 82-90.

【21】Du X, El-Khamy M, Lee J, et al. Fused DNN: a deep neural network fusion approach to fast and robust pedestrian detection[C]∥IEEE Winter Conference on Applications of Computer Vision, 2017: 953-961.

引用该论文

Liu Hui,Peng Li,Wen Jiwei. Multi-Scale Aware Pedestrian Detection Algorithm Based on Improved Full Convolutional Network[J]. Laser & Optoelectronics Progress, 2018, 55(9): 091504

刘辉,彭力,闻继伟. 基于改进全卷积网络的多尺度感知行人检测算法[J]. 激光与光电子学进展, 2018, 55(9): 091504

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF