基于多特征融合的行人检测方法

doi:10.16180/j.cnki.issn1007-7820.2021.05.006

摘要/Abstract

摘要：

作为一个典型的目标检测问题,行人检测问题已成为近年来的研究热点。行人检测技术虽被广泛应用于智能交通、自动驾驶、视频监控以及行为分析等领域,但仍存在着需要解决的问题。文中在多特征融合的基础上提出了一个多通道特征模型,多通道特征模型由非深度学习分支、整体分支以及肢体分支组成。文中通过非深度学习分支提取出数量少、质量高的行人候选区域,减轻了滑动窗口穷举搜索带来的计算负担,提高了计算效率。该方法由多层卷积通道特征得到的整体分支以及肢体分支,分别通过人体整体信息和人体部位的语义信息来检测行人;使用多通道特征模型分别在Caltech行人数据集和INRIA行人数据集中进行训练和检测。实验结果表明,结合各分支的输出,文中提出的行人检测器具有较低的漏检率,在INRIA行人数据集和Caltech行人数据集中漏检率分别为8.24%和19.78%。

关键词: 行人检测, 卷积神经网络, 多通道特征, 多层卷积通道特征, 局部去相关通道特征, 方向梯度直方图

Abstract:

As a typical problem of object detection, pedestrian detection has been an active research topic in recent years. Pedestrian detection is widely used in intelligent transportation, autonomous driving, video surveillance, behavior analysis and other fields, but there are many problems to be solved. In this study, a multi-channel feature model based on multiple feature fusion which consists of a non-deep learning branch, a body branch and a limb branch is proposed. A small number of high-quality pedestrian candidate areas are extracted through the non-deep learning branch, thus reducing the computational cost cause by exhaustive search by sliding window and improving the computational efficiency. The body branch and limb branch obtained from the feature of multi-layer convolutional channel are applied to detect pedestrians through the overall human body information and the semantic information of human body parts, respectively. Caltech pedestrian dataset and INRIA pedestrian dataset are adopted to train and test the proposed model. Experimental results show that combine with the output of each branch, the proposed pedestrian detector have a lower miss rate. The miss rates on INRIA pedestrian dataset and Caltech pedestrian dataset are 8.24% and 19.78%, respectively.

Key words: pedestrian detection, convolutional neural network, multi-channel feature, multi-layer convolution channel feature, locally decorrelated channel features, histogram of oriented gradient

中图分类号:

TP391.41

顾伟,李菲菲,陈虬. 基于多特征融合的行人检测方法[J]. 电子科技, 2021, 34(5): 29-34.

GU Wei,LI Feifei,CHEN Qiu. Pedestrian Detection Algorithm Based on Multiple Feature Fusion[J]. Electronic Science and Technology, 2021, 34(5): 29-34.

图/表 9

图1

表1

图2

图3

图4

图5

图6

图7

图8

参考文献 27

[1]	Dalal N, Triggs B. Histograms of oriented gradients for human detection[C]. San Diego:IEEE International Conference on Computer Vision and Pattern Recognition, 2005.
[2]	Papageorgiou C, Poggio T. A trainable system for object detection[J]. International Journal of Computer Vision, 2000,38(1):15-33. doi: 10.1023/A:1008162616689
[3]	Sabzmeydani P, Mori G. Detecting pedestrians by learning shapelet features[C]. Minneapolis:IEEE International Conference on Computer Vision and Pattern Recognition, 2007.
[4]	Ojala T, Harwood I. A comparative study of texture measures with classification based on feature distributions[J]. Pattern Recognition, 1996,29(1):51-59. doi: 10.1016/0031-3203(95)00067-4
[5]	Walk S, Majer N, Schindler K, et al. New features and insights for pedestrian detection[C]. San Francisco: International Conference on Computer Vision and Pattern Recognition, 2010.
[6]	Doll'ar P, Tu Z, Perona P, et al. Integral channel features[C]. London:British Machine Vision Conference, 2009.
[7]	Felzenszwalb P F, Girshick R B, Mcallester D, et al. Object detection with discriminatively trained part-based models[J]. IEEE Transactions on Software Engineering, 2010,32(9):1627-1645.
[8]	Tan X, Triggs B. Enhanced local texture feature sets for face recognition under difficult lighting conditions[J]. IEEE Transactions on Image Processing, 2010,19(6):1635-1650. doi: 10.1109/TIP.2010.2042645
[9]	Feigang T. SLBP:an improved texture feature for pedestrian detection[C]. Changsha:International Conference on Smart City and Systems Engineering, 2017.
[10]	Dollar 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. doi: 10.1109/TPAMI.2014.2300479
[11]	Nam W, Dollar P, Han J H. Local decorrelation for improved pedestrian detection[C]. Montreal:Neural Information Processing Systems, 2014.
[12]	Zhang S, Benenson R, Schiele B. Filtered channel features for pedestrian detection[C]. Boston:International Conference on Computer Vision and Pattern Recognition, 2015.
[13]	Li H, Wu Z, Zhang J. Pedestrian detection based on deep learning model[C]. Shanghai:IEEE International Congress on Image & Signal Processing, 2017.
[14]	Hu Q, Wang P, Shen C, et al. Pushing the limits of deep CNNs for pedestrian detection[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2018,28(6):1358-1368. doi: 10.1109/TCSVT.2017.2648850
[15]	Ribeiro D, Nascimento J C, Bernardino A, et al. Improving the performance of pedestrian detectors using convolutional learning[J]. Pattern Recognition, 2016,61(1):641-649. doi: 10.1016/j.patcog.2016.05.027
[16]	Bui H M, Lech M, Cheng E, et al. Object recognition using deep convolutional features transformed by a recursive network structure[J]. IEEE Access, 2016,4(1):10059-10066. doi: 10.1109/ACCESS.2016.2639543
[17]	Lecun Y L, Bottou L, Bengio Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998,86(11):2278-2324. doi: 10.1109/5.726791
[18]	Krizhevsky A, Sutskever I, Hinton G. ImageNet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012,25(2):1-9.
[19]	Hosang J, Omran M, Benenson R, et al. Taking a deeper look at pedestrians[C]. Boston:International Conference on Computer Vision and Pattern Recognition, 2015.
[20]	Zitnick C L, Dollar P. Edge Boxes:locating object proposals from edges[C]. Zurich:The Thirteenth European Conference on Computer Vision, 2014.
[21]	Ren S, He K, Girshick R, et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015,39(6):1137-1149. doi: 10.1109/TPAMI.2016.2577031
[22]	Yang B, Yan J, Lei Z, et al. Convolutional channel features[C]. Santiago:IEEE International Conference on Computer Vision, 2015.
[23]	Hochreiter S, Schmidhuber, Jürgen. Long short-term memory[J]. Neural Computation, 1997,9(8):1735-1780. pmid: 9377276
[24]	Viola P, Jones M J, Snow D. Detecting pedestrians using patterns of motion and appearance[J]. International Journal of Computer Vision, 2005,63(2):153-161. doi: 10.1007/s11263-005-6644-8
[25]	Paisitkriangkrai S, Shen C, Hengel A V D. Strengthening the effectiveness of pedestrian detection with spatially pooled features[C]. Zurich:The Thirteenth European Conference on Computer Vision, 2014.
[26]	Wu F C, Wang Z H, Yang Y. Exploring prior knowledge for pedestrian detection[C]. Swansea:British Machine Vision Conference, 2015.
[27]	巨志勇, 黄凯. 基于共生概率特征量的行人检测[J]. 电子科技, 2015,28(11):139-142.
	Ju Zhiyong, Huang Kai. Pedestrian detection based on symbiosis probability feature[J]. Electronic Science and Technology, 2015,28(11):139-142.

卷积特征图	误检率/%
Conv3-4	14.14
Conv4-4	10.94
Conv5-4	10.10
Conv4-4, Conv5-4	9.09
Conv3-4, Conv4-4,Conv5-4	9.76