改进的YOLOv3红外视频图像行人检测算法An improved infrared video image pedestrian detection algorithm
王殿伟,何衍辉,李大湘,刘颖,许志杰,王晶
摘要(Abstract)：
针对YOLOv3检测红外视频图像行人时存在准确率低、漏检率高的问题,提出一种改进的YOLOv3红外视频图像行人检测算法。根据行人在红外图像中呈现宽高比相对固定的特点,利用k-means聚类方法选取目标候选框个数和宽高比维度,调整网络参数并提高输入图像分辨率,最后进行多尺度训练得到最优检测模型,从而检测红外视频图像序列中的行人目标,并通过候选框标注行人位置。在CVC-09红外行人数据集上进行对比实验,结果表明,改进的YOLOv3算法在红外行人检测中的准确率高达90.63%,明显优于Faster-rcnn和YOLOv3算法,且改进后的网络能够同时检测到更多目标,降低了漏检率。
关键词(KeyWords)： 行人检测;红外图像;YOLOv3;聚类分析
基金项目(Foundation): 陕西省自然科学基础研究计划资助项目(2018JM6118);; 西安邮电大学创新创业项目(2018SC-08)
作者(Author): 王殿伟,何衍辉,李大湘,刘颖,许志杰,王晶
DOI: 10.13682/j.issn.2095-6533.2018.04.008
参考文献(References)：
- [1]王殿伟,闫伟超,刘颖,等.基于暗原色先验的快速单幅图像去雾算法[J/OL].计算机应用研究,2019(1):1-7[2018-02-01].http://kns.cnki.net/kcms/detail/51.1196.TP.20180208.1715.120.html.
- [2] LUO J H,LIN C H.Pure FPGA implementation of an HOG based real time pedestrian detection system[J/OL].Sensors,2018,18(4):1174[2018-03-01].http://dx.doi.org/10.3390/s18041174.
- [3] TRICHET R,BREMOND F.LBP channelsfor pedestrian detection[C/OL]//2018IEEE Winter Conference on Applications of Computer Vision,Lake Tahoe:IEEE,2018:1066-1074[2018-03-01].http://doi.ieeecomputersociety.org/10.1109/WACV.2018.00122.
- [4]刘颖,权婉,伍世虔.基于边缘增强的正则化超分辨率图像重建[J/OL].西安邮电大学学报,2016,21(6):14-19[2018-03-01].http://dx.doi.org/10.13682/j.issn.2095-6533.2016.06.003.html.
- [5] GIRSHICK R,DONAHUE J,DARRELLT,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C/OL]//2014IEEE Conference on Computer Vision and Pattern Recognition,Columbus:IEEE,2014:580-587[2018-05-01].http://dx.doi.org/10.1109/CVPR.2014.81.
- [6] GIRSHICK R.Fast R-CNN[C/OL]//2015IEEE International Conference on Computer Vision(ICCV),Santiago,Chile:IEEE,2015:1440-1448[2018-03-01].http://dx.doi.org/10.1109/ICCV.2015.169.
- [7] LIU W,ANGUELOV D,ERHAN D,et al.SSD:Single Shot MultiBox Detector[C/OL]//Computer Vision-ECCV 2016.ECCV 2016.Lecture Notes in Computer Science(vol 9905).Cham:Springer,2016:21-37[2018-05-01].http://dx.doi.org/10.1007/978-3-319-46448-0_2.
- [8] REDMON J,DIVVALA S,GIRSHICK R,et al.You only look once:Unified,Real-Time Object Detection[C/OL]//2016IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas:IEEE,2016:779-788[2018-02-01].http://doi.ieeecomputersociety.org/10.1109/CVPR.2016.91.
- [9] REDMON J,FARHADI A.YOLO9000:Better,Faster,Stronger[C/OL]//2017 IEEE Conference on Computer Vision and Pattern Recognition,Honolulu:IEEE,2017:6517-6525[2018-04-01].http://doi.ieeecomputersociety.org/10.1109/CVPR.2017.690.
- [10]REDMON J,FARHADI A.YOLOv3:An Incremental Improvement[Z/OL].Computer Vision and Pattern Recognition.(2018-04-08)[2018-04-01].https://arxiv.org/abs/1804.02767v1.
- [11]HE K,ZHANG X,REN S,et al.Deep residual learning for image recognition[C/OL]//2016IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas:IEEE,2015:770-778[2018-03-01].http://doi.ieeecomputersociety.org/10.1109/CVPR.2016.90.
- [12]SIMONYAN K,ZISSERMAN A.Very deep convolutional networks for large scaleImage recognition[J/OL].Computer Vision and Pattern Recognition,2014:1-14[2018-03-01].http://adsabs.harvard.edu/abs/2014arXiv1409.1556S.
- [13]张素洁,赵怀慈.最优聚类个数和初始聚类中心点选取算法研究[J/OL].计算机应用研究,2017,34(6):1617-1620[2018-03-01]. http://dx.doi.org/10.3969/j.issn.1001-3695.2017.06.004.
- [14]REN S,HE K,GIRSHICK R,et al.FasterR-CNN:towards real-time object detection with region proposal networks[J/OL]//IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,39(6):1137-1149[2018-03-01].http://dx.doi.org/10.1109/TPAMI.2016.2577031.