| 4,104 | 120 | 220 |
| 下载次数 | 被引频次 | 阅读次数 |
针对目标检测特别是小目标检测问题,首先梳理了目标检测算法的发展与现状,系统性地总结了基于深度学习的目标检测算法研究进展,从精度和速度方面分析了两阶段及单阶段检测算法的优缺点及检测性能;其次作为参考梳理了小目标检测领域的典型数据集,分析在小目标检测难题上的改进算法,提升检测精度和速度,以及获得二者的平衡,成为小目标检测改进的方向和面临的挑战;最后对基于深度学习的目标检测应用领域做出展望和预测,为小目标检测问题提供了参考。
Abstract：Aimed at the target detection,especially the small target detection problem,the development and current status of the target detection algorithm are summarized; the research progress of the target detection algorithm based on deep learning is systematically summarized; the advantages and disadvantages of two-stage and single-stage detection algorithms and their detection performance are analyzed in terms of accuracy and speed. The typical data sets in the field of small target detection are combed as a reference,and the improved algorithm on the small target detection problem is analyzed. The detection accuracy and speed are improved,and obtaining the balance of them,which becomes the direction and challenge of small target detection improvement. Finally,the prospects and predictions of target detection based on deep learning are made,which provides a reference for small target detection.
[1] Lécun Y,Bottou L,Bengio Y,et al. Gradient-basedlearning applied to document recognition[J]. Proceed-ings of the IEEE,1998,86(11):2278-2324.
[2] Krizhevsky A,Sutskever I,Hinton G E. Image Net classi-fication with deep convolutional neural networks[C]. In-ternational Conference on Neural Information ProcessingSystems,2012:1097-1105.
[3] Girshick R,Donahue J,Darrell T,et al. Rich featurehierarchies for accurate object detection and semantic seg-mentation[C]. IEEE Conference on Computer Vision andPattern Recognition,2014,580-587.
[4] He K,Zhang X,Ren S,et al. Spatial pyramid pooling indeep convolutional networks for visual recognition[J].IEEE Transactions on Pattern Analysis&Machine Intelli-gence,2015,37(9):1904-1916.
[5] Girshick R. Fast R-CNN[J]. Computer Science,2015,(3):5-7.
[6] Ren S,He K,Girshick R,et al. Faster R-CNN:towardsreal-time object detection with region proposal networks[J]. IEEE Trans Pattern Anal Mach Intell,2015,39(6):1137-1149.
[7] Redmon J,Divvala S,Girshick R,et al. You only lookonce:unified,real-time object detection[C]. IEEE Con-ference on Computer Vision and Pattern Recognition,2016:779-788.
[8] Redmon J,Farhadi A. YOLO9000:better,faster,stron-ger[J]. IEEE Conference on Computer Vision and Pat-tern Recognition(CVPR),2016:6517-6525.
[9] Redmon J,Farhadi A. YOLOv3:an incremental im-provement[J]. ar Xiv Preprint,2018.
[10] Liu W,Anguelov D,Erhan D,et al. SSD:single shotmultibox detector[C]. European Conference on Comput-er Vision,2016:21-37.
[11] Bell S,Zitnick C L,Bala K,et al. Inside-outside net:detecting objects in context with skip pooling and recur-rent neural networks[J]. IEEE Conference on ComputerVision and Pattern Recognition(CVPR),2015:2874-2883.
[12] M¨a enp¨a¨a T,Pietik¨a inen M. Texture analysis withlocalbinary patterns[J]. Handbook of Pattern Recognition&Computer Vision,2005,3540:115-118.
[13]李大伟.固定翼无人机地面车辆目标自动检测与跟踪[J].电子设计工程,2017,26(8):13-17.
[14] Zhu H,Chen X,Dai W,et al. Orientation robust ob-ject detection in aerial images using deep convolutionalneural network[C]. IEEE International Conference onImage Processing,2015:3735-3739.
[15] Xiang Y,Choi W,Lin Y,et al. Subcategory-awareconvolutional neural networks for object proposals anddetection[C]. Applications of Computer Vision,2017.924-933.
[16] Starck J L. A combined approach for object detectionand deconvolution[J]. Astronomy&Astrophysics Sup-plement,2000,147(1):139-149.
[17] Razakarivony S,Jurie F. Vehicle detection in aerial im-agery:A small target detection benchmark[J]. Journalof Visual Communication&Image Representation,2015,34:187-203.
[18] Takeki A,Tu T T,Yoshihashi R,et al. Combining deepfeatures for object detection at various scales:findingsmall birds in landscape images[J]. Ipsj Transactionson Computer Vision&Applications,2016,8(1):5.
[19] Kampffmeyer M,Salberg A B,Jenssen R. Semanticsegmentation of small objects and modeling of uncertain-ty in urban remote sensing images using deep convolu-tional neural networks[C]. Computer Vision and PatternRecognition Workshops,2016:680-688.
[20] Mundhenk T N,Konjevod G,Sakla W A,et al. A largecontextual dataset for classification,detection and count-ing of cars with deep learning[M]. Springer Internation-al Publishing,2016.
[21] Lin T Y,Dollar P,Girshick R,et al. Feature pyramidnetworks for object detection[C]. IEEE Conference onComputer Vision and Pattern Recognition,2017:936-944.
[22] Fu C Y,Liu W,Ranga A,et al. DSSD:Deconvolu-tional single shot detector[J]. ar Xiv Preprint,2017,(6):12-15.
[23] Ren J,Chen X,Liu J,et al. Accurate single stage de-tector using recurrent rolling convolution[J]. IEEEConference on Computer Vision and Pattern Recognition(CVPR),2017,(5):752-760.
[24] Ren Y,Zhu C,Xiao S. Small object detection in opticalremote sensing images via modified faster R-CNN[J].Applied Sciences,2018,(2):134-136.
[25] Yi K,Jian Z,Chen S,et al. Knowledge-based recur-rent attentive neural network for small object detection[J]. ar Xiv Preprint,2018,(4):15-18.
基本信息:
DOI：10.16358/j.issn.1009-1300.2019.8.522
中图分类号:TP391.41;TP18
引用信息:
[1]刘晓楠,王正平,贺云涛,等.基于深度学习的小目标检测研究综述[J].战术导弹技术,2019,No.193(01):100-107.DOI:10.16358/j.issn.1009-1300.2019.8.522.