显著性背景感知的多尺度红外行人检测方法

赵斌; 王春平; 付强

doi:10.11999/JEIT190761

显著性背景感知的多尺度红外行人检测方法

doi: 10.11999/JEIT190761

陆军工程大学石家庄校区电子与光学工程系石家庄 050003

详细信息

作者简介:
赵斌：男，1990年生，博士生，研究方向为深度学习、目标检测

王春平：男，1965年生，博士生导师，研究方向为图像处理、火力控制理论与应用

付强：男，1981年生，讲师，博士，研究方向为计算机视觉、网络化火控与指控技术

通讯作者:
王春平　wang_c_p@163.com

中图分类号: TN215
计量
- 文章访问数: 2384
- HTML全文浏览量: 2045
- PDF下载量: 117
- 被引次数: 0
出版历程
- 收稿日期: 2019-09-30
- 修回日期: 2020-05-13
- 网络出版日期: 2020-05-20
- 刊出日期: 2020-10-13

Multi-scale Pedestrian Detection in Infrared Images with Salient Background-awareness

Department of Electronic and Optical Engineering, Shijiazhuang Campus of Army Engineering University, Shijiazhuang 050003, China

摘要

摘要: 超大视场(U-FOV)红外成像系统探测范围大、不受光照限制，但存在尺度多样、小目标丰富的特点。为此该文提出一种具备背景感知能力的多尺度红外行人检测方法，在提高小目标检测性能的同时，减少冗余计算。首先，构建了4尺度的特征金字塔网络分别独立预测目标，补充高分辨率细节特征。其次，在特征金字塔结构的横向连接中融入注意力模块，产生显著性特征，抑制不相关区域的特征响应、突出图像局部目标特征。最后，在显著性系数的基础上构建了锚框掩膜生成子网络，约束锚框位置，排除平坦背景，提高处理效率。实验结果表明，显著性生成子网络仅增加5.94%的处理时间，具备轻量特性；超大视场(U-FOV)红外行人数据集上的识别准确率达到了93.20%，比YOLOv3高了26.49%；锚框约束策略能节约处理时间18.05%。重构模型具有轻量性和高准确性，适合于检测超大视场中的多尺度红外目标。
- 红外行人检测 /
- 超大视场 /
- 卷积神经网络 /
- 背景感知 /
- 多尺度
Abstract: The infrared imaging system of Ultrawide Field Of View (U-FOV) has large monitoring range and is not limited by illumination, but there are diverse scales and abundant small objects. For accurately detecting them, a multi-scale infrared pedestrian detection method is proposed with the ability of background-awareness, which can improve the detection performance of small objects and reduce the redundant computation. Firstly, a four scales feature pyramid network is constructed to predict object independently and supplement detail features with higher resolution. Secondly, attention module is integrated into the horizontal connection of feature pyramid structure to generate salient features, suppress feature response of irrelevant areas and enhance the object features. Finally, the anchor mask generation subnetwork is constructed on the basis of salient coefficient to the location of the anchors, to eliminate the flat background, and to improve the processing efficiency. The experimental results show that the salient generation subnetwork only increases the processing time by 5.94%, and has the lightweight characteristic. The Average-Precision is 93.20% on the U-FOV infrared pedestrian dataset, 26.49% higher than that of YOLOv3. Anchor box constraint strategy can save 18.05% of processing time. The proposed method is lightweight and accurate, which is suitable for detecting multi-scale infrared objects in the U-FOV camera.
- Infrared pedestrian detection /
- Ultrawide Field Of View(U-FOV) /
- Convolutional Neural Network(CNN) /
- Background-awareness /
- Multi-scale

HTML全文

图 1 超大视场红外图像行人特性

下载: 全尺寸图片幻灯片

图 2 多尺度红外行人检测网络结构

下载: 全尺寸图片幻灯片

图 3 注意力模块结构

下载: 全尺寸图片幻灯片

图 4 显著性特征与卷积特征融合方法

下载: 全尺寸图片幻灯片

图 5 锚框掩膜生成过程

下载: 全尺寸图片幻灯片

图 6 不同输入图像的锚框掩膜

下载: 全尺寸图片幻灯片

图 7 不同二值化阈值下的锚框掩膜

下载: 全尺寸图片幻灯片

图 8 红外行人检测可视化结果

下载: 全尺寸图片幻灯片

表 1 不同IoU阈值下的行人检测平均准确率

方法	主干网络	训练集	平均准确率(AP)
方法	主干网络	训练集	IoU=0.3	IoU=0.45	IoU=0.5	IoU=0.7
Faster R-CNN	ResNet101	U-FOV	–	–	0.5932	–
SSD	Mobilenet_v1	U-FOV	–	–	0.5584	–
R-FCN	ResNet101	U-FOV	–	–	0.6312	–
CSP	Resnet50	U-FOV	–	–	0.8414	–
YOLOv3	Darknet53	U-FOV	0.6595	0.6671	0.6628	0.6461
YOLOv3+FS	Darknet53	U-FOV	0.8880	0.8870	0.8828	0.8511
YOLOv3+FS	Darknet53	Caltech+U-FOV	0.9057	0.9078	0.9084	0.8961
本文方法	Darknet53	Caltech+U-FOV	0.9201	0.9320	0.9315	0.9107

下载: 导出CSV

表 2 参数量对比

方法	总参数量	可训练参数量	不可训练参数量
YOLOv3	61576342	61523734	52608
本文方法	64861976	64806296	55680

下载: 导出CSV

表 3 U-FOV测试集图像总处理时间及处理帧速

方法	YOLOv3	YOLOv3+Attention	FS+Attention	本文方法
总时间(s)	90.35	95.72	125.39	107.25
处理帧率	7.32	6.91	5.27	6.16

下载: 导出CSV

参考文献(25)

BLOISI D D, PREVITALI F, PENNISI A, et al. Enhancing automatic maritime surveillance systems with visual information[J]. IEEE Transactions on Intelligent Transportation Systems, 2017, 18(4): 824–833. doi: 10.1109/TITS.2016.2591321

KANG J K, HONG H G, and PARK K R. Pedestrian detection based on adaptive selection of visible light or far-infrared light camera image by fuzzy inference system and convolutional neural network-based verification[J]. Sensors, 2017, 17(7): 1598. doi: 10.3390/s17071598

KIM S, SONG W J, and KIM S H. Infrared variation optimized deep convolutional neural network for robust automatic ground target recognition[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops, Honolulu, USA, 2017: 195–202. doi: 10.1109/CVPRW.2017.30.

王晨, 汤心溢, 高思莉. 基于人眼视觉的红外图像增强算法研究[J]. 激光与红外, 2017, 47(1): 114–118. doi: 10.3969/j.issn.1001-5078.2017.01.022

WANG Chen, TANG Xinyi, and GAO Sili. Infrared image enhancement algorithm based on human vision[J]. Laser &Infrared, 2017, 47(1): 114–118. doi: 10.3969/j.issn.1001-5078.2017.01.022

MUNDER S, SCHNORR C, and GAVRILA D M. Pedestrian detection and tracking using a mixture of view-based shape-texture models[J]. IEEE Transactions on Intelligent Transportation Systems, 2008, 9(2): 333–343. doi: 10.1109/TITS.2008.922943

DALAL N and TRIGGS B. Histograms of oriented gradients for human detection[C]. IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Diego, USA, 2005: 886–893. doi: 10.1109/CVPR.2005.177.

ZHANG Shanshan, BAUCKHAGE C, and CREMERS A B. Informed haar-like features improve pedestrian detection[C]. 2014 IEEE Conference on Computer Vision and Pattern Recognition, Columbus, USA, 2014: 947–954. doi: 10.1109/CVPR.2014.126.

WATANABE T and ITO S. Two co-occurrence histogram features using gradient orientations and local binary patterns for pedestrian detection[C]. The 2nd IAPR Asian Conference on Pattern Recognition, Naha, Japan, 2013: 415–419. doi: 10.1109/ACPR.2013.117.

余春艳, 徐小丹, 钟诗俊. 面向显著性目标检测的SSD改进模型[J]. 电子与信息学报, 2018, 40(11): 2554–2561. doi: 10.11999/JEIT180118

YU Chunyan, XU Xiaodan, and ZHONG Shijun. An improved SSD model for saliency object detection[J]. Journal of Electronics &Information Technology, 2018, 40(11): 2554–2561. doi: 10.11999/JEIT180118

LIU Songtao, HUANG Di, and WANG Yunhong. Adaptive NMS: Refining pedestrian detection in a crowd[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, USA, 2019: 6452–6461. doi: 10.1109/CVPR.2019.00662.

LIU Wei, LIAO Shengcai, REN Weiqiang, et al. Center and scale prediction: A box-free approach for pedestrian and face detection[C]. The IEEE Conference on Computer Vision and Pattern Recognition, Los Angeles, USA, 2019: 5187–5196.

车凯, 向郑涛, 陈宇峰, 等. 基于改进Fast R-CNN的红外图像行人检测研究[J]. 红外技术, 2018, 40(6): 578–584. doi: 10.11846/j.issn.1001_8891.201806010

CHE Kai, XIANG Zhengtao, CHEN Yufeng, et al. Research on infrared image pedestrian detection based on improved fast R-CNN[J]. Infrared Technology, 2018, 40(6): 578–584. doi: 10.11846/j.issn.1001_8891.201806010

王殿伟, 何衍辉, 李大湘, 等. 改进的YOLOv3红外视频图像行人检测算法[J]. 西安邮电大学学报, 2018, 23(4): 48–52. doi: 10.13682/j.issn.2095-6533.2018.04.008

WANG Dianwei, HE Yanhui, LI Daxiang, et al. An improved infrared video image pedestrian detection algorithm[J]. Journal of Xi'an University of Posts and Telecommunications, 2018, 23(4): 48–52. doi: 10.13682/j.issn.2095-6533.2018.04.008

GIRSHICK R. Fast R-CNN[C]. 2015 IEEE International Conference on Computer Vision, Santiago, Chile, 2015: 1440–1448. doi: 10.1109/ICCV.2015.169.

REDMON J and FARHADI A. YOLOv3: An incremental improvement[EB/OL]. http://arxiv.org/abs/1804.02767, 2018.

郭智, 宋萍, 张义, 等. 基于深度卷积神经网络的遥感图像飞机目标检测方法[J]. 电子与信息学报, 2018, 40(11): 2684–2690. doi: 10.11999/JEIT180117

GUO Zhi, SONG Ping, ZHANG Yi, et al. Aircraft detection method based on deep convolutional neural network for remote sensing images[J]. Journal of Electronics &Information Technology, 2018, 40(11): 2684–2690. doi: 10.11999/JEIT180117

CHEN Long, ZHANG Hanwang, XIAO Jun, et al. SCA-CNN: Spatial and channel-wise attention in convolutional networks for image captioning[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, 2017: 6298–6306. doi: 10.1109/CVPR.2017.667.

WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[C]. Proceedings of the 15th European Conference on Computer Vision, Munich, Germany, 2018: 3–19. doi: 10.1007/978-3-030-01234-2_1.

DOLLÁR P, WOJEK C, SCHIELE B, et al. Pedestrian detection: An evaluation of the state of the art[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(4): 743–761. doi: 10.1109/TPAMI.2011.155

FU Chengyang, LIU Wei, RANGA A, et al. DSSD: Deconvolutional single shot detector[J]. arXiv, 2017, 1701.06659.

HE Kaiming, GKIOXARI G, DOLLÁR P, et al. Mask R-CNN[C]. 2017 IEEE International Conference on Computer Vision, Venice, Italy, 2017: 2980–2988. doi: 10.1109/ICCV.2017.322.

BERG A, AHLBERG J, and FELSBERG M. A thermal object tracking benchmark[C]. The 2015 12th IEEE International Conference on Advanced Video and Signal Based Surveillance, Karlsruhe, Germany, 2015: 1–6. doi: 10.1109/AVSS.2015.7301772.

LIU Wei, ANGUELOV D, ERHAN D, et al. SSD: Single shot multibox detector[C]. Proceedings of the 14th European Conference on Computer Vision, Amsterdam, Netherlands, 2016: 21–37. doi: 10.1007/978-3-319-46448-0_2.

REN Shaoqing, HE Kaiming, 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, 2017, 39(6): 1137–1149. doi: 10.1109/TPAMI.2016.2577031

DAI Jifeng, LI Yi, HE Kaiming, et al. R-FCN: Object detection via region-based fully convolutional networks[C]. Advances in Neural Information Processing Systems, Barcelona, Spain, 2016: 379–387.

施引文献

资源附件(0)

访问统计