首页 > 论文 > 激光与光电子学进展 > 56卷 > 18期(pp:180402--1)

基于深度学习的管制物品自动检测算法研究

Automatic Detection Algorithm for Controlled Items Based on Deep Learning

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

摘要

提出一种对图片分区域检测的特征融合目标检测算法。利用角度旋转方法对数据集进行扩增;在Single Shot MultiBox Detector(SSD)算法的基础上,采用多尺度特征融合的方法在浅层特征图中融合更深层的特征,以扩大浅层特征图的感受野,提高小目标的检测精度;当输入图片较大时,如大于1024 pixel×1024 pixel,对目标图像进行分区域检测。为了验证该算法的精度,选择VOC2007+2012通用数据集和SDCI2018管制物品数据集对所提算法的精度进行测试。结果表明:所提算法在VOC2007+2012通用数据集上的检测精度为80.3%,比SSD算法提高了1.4%;在SDCI2018管制物品数据集上的检测精度为97.9%,比SSD算法提高了2.2%。所提算法能够实时准确地检测出安检图片中的管制物品,特别是对于大图片中的小目标检测效果较好。

Abstract

This paper proposes a feature-fused object detection algorithm for image sub-region detection. First, a multi-angle rotation method is used to amplify the dataset. Then, based on the Single-Shot MultiBox Detector (SSD) algorithm, the deep features are fused to shallow features using the method of multi-scale feature fusion to enlarge the receptive field of the shallow feature map and improve the detection accuracy for small objects. If the input image is large, i.e., more than 1024 pixel×1024 pixel, it is detected in separated regions. To verify the performance of the algorithm, the VOC2007+2012 and SDCI2018 datasets are selected to test the algorithm. The detection accuracy of the proposed algorithm on the VOC2007+2012 dataset is 80.3%, which is 1.4% higher than that of the SSD algorithm. The detection accuracy on SDCI2018 dataset is 97.9%, which is 2.2% higher than that of the SSD algorithm. The proposed algorithm can accurately detect the controlled items in security check images in real time, particularly for the small objects in large images.

Newport宣传-MKS新实验室计划
补充资料

DOI:10.3788/LOP56.180402

所属栏目:探测器

基金项目:国家重点研发计划、江西省自然科学基金、江西省教育厅科技项目、江西理工大学科研基金;

收稿日期:2019-03-05

修改稿日期:2019-04-15

网络出版日期:2019-09-01

作者单位    点击查看

吉祥凌:江西理工大学信息工程学院, 江西 赣州 341000
吴军:江西理工大学信息工程学院, 江西 赣州 341000
易见兵:江西理工大学信息工程学院, 江西 赣州 341000
张晓光:上海英迈吉东影图像设备有限公司, 上海 200120

联系人作者:易见兵(yijianbing8@163.com)

备注:国家重点研发计划、江西省自然科学基金、江西省教育厅科技项目、江西理工大学科研基金;

【1】Liu W, Anguelov D, Erhan D et al. SSD: single shot MultiBox detector. //Leibe B, Matas J, Sebe N, et al. Computer vision-ECCV 2016. Lecture notes in computer science. Cham: Springer. 9905, 21-37(2016).

【2】Zhang H Y, Wang S N and Hu W B. Improved method for estimating number of people based on convolution neural network. Laser & Optoelectronics Progress. 55(12), (2018).
张红颖, 王赛男, 胡文博. 改进的基于卷积神经网络的人数估计方法. 激光与光电子学进展. 55(12), (2018).

【3】Ma Y J, Li X Y and Song X F. Traffic sign recognition based on improved deep convolution neural network. Laser & Optoelectronics Progress. 55(12), (2018).
马永杰, 李雪燕, 宋晓凤. 基于改进深度卷积神经网络的交通标志识别. 激光与光电子学进展. 55(12), (2018).

【4】Ou P, Zhang Z, Lu K et al. Object detection of remote sensing images based on convolutional neural networks. Laser & Optoelectronics Progress. 56(5), (2019).
欧攀, 张正, 路奎 等. 基于卷积神经网络的遥感图像目标检测. 激光与光电子学进展. 56(5), (2019).

【5】Girshick R, Donahue J, Darrell T et al. Rich feature hierarchies for accurate object detection and semantic segmentation. [C]//2014 IEEE Conference on Computer Vision and Pattern Recognition, June 23-28, 2014, Columbus, OH, USA. New York: IEEE. 580-587(2014).

【6】Uijlings J R R, Gevers T et al. Selective search for object recognition. International Journal of Computer Vision. 104(2), 154-171(2013).

【7】Zitnick C L and Dollár P. Edge boxes: locating object proposals from edges. //Fleet D, Pajdla T, Schiele B, et al. Computer vision-ECCV 2014. Lecture notes in computer science. Cham: Springer. 8693, 391-405(2014).

【8】He K M, Zhang X Y, Ren S Q et al. Spatial pyramid pooling in deep convolutional networks for visual recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence. 37(9), 1904-1916(2015).

【9】Girshick R. Fast R-CNN. [C]//2015 IEEE International Conference on Computer Vision (ICCV), December 7-13, 2015, Santiago, Chile. New York: IEEE. 1440-1448(2015).

【10】Ren S Q, He K M, Girshick R et al. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence. 39(6), 1137-1149(2017).

【11】Redmon J, Divvala S, Girshick R et al. You only look once: unified, real-time object detection. [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 27-30, 2016, Las Vegas, NV, USA. New York: IEEE. 779-788(2016).

【12】Simonyan K. -04-10) [2019-01-25]. Zisserman A. Very deep convolutional networks for large-scale image recognition. org/abs/1409, (2015).

【13】Fu C Y, Liu W, Ranga A et al. -01-23)[2019-01-25]. org/abs/1701, (2017).

【14】He K M, Zhang X Y, Ren S Q et al. Deep residual learning for image recognition. [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 27-30, 2016, Las Vegas, NV, USA. New York: IEEE. 770-778(2016).

【15】Chang H T, Gou J N and Li X M. Application of Faster R-CNN in image defect detection of industrial CT. Journal of Image and Graphics. 23(7), 1061-1071(2018).
常海涛, 苟军年, 李晓梅. Faster R-CNN在工业CT图像缺陷检测中的应用. 中国图象图形学报. 23(7), 1061-1071(2018).

【16】Cao S Y, Liu Y H and Li X Z. Vehicle detection method based on Fast R-CNN. Journal of Image and Graphics. 22(5), 671-677(2017).
曹诗雨, 刘跃虎, 李辛昭. 基于Fast R-CNN的车辆目标检测. 中国图象图形学报. 22(5), 671-677(2017).

【17】Woo S and Hwang S. -09-18)[2019-01-25]. Kweon I S. Stairnet: top-down semantic aggregation for accurate one shot detection. 1709, (2017).

【18】Lin T Y, Goyal P, Girshick R et al. -02-07) [2019-01-25]. org/abs/1708, (2018).

【19】Jeong J and Park H. -05-26) [2019-01-25]. Kwak N. Enhancement of SSD by concatenating feature maps for object detection. org/abs/1705, (2017).

【20】Kong T, Yao A B, Chen Y R et al. HyperNet: towards accurate region proposal generation and joint object detection. [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 27-30, 2016, Las Vegas, NV, USA. New York: IEEE. 845-853(2016).

【21】Liu W and Rabinovich A. -11-19)[2019-01-25]. Berg A C. ParseNet: looking wider to see better. org/abs/1506, (2015).

【22】Dai J, Li Y, He K M et al. -06-21) [2019-01-25]. org/abs/1605, (2016).

【23】Lin T Y, Dollár P, Girshick R et al. Feature pyramid networks for object detection. [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), July 21-26, 2017, Honolulu, HI, USA. New York: IEEE. 936-944(2017).

【24】Ren J, Chen X H, Liu J B et al. -04-19)[2019-01-25]. org/abs/1704, (2017).

【25】Zheng L W and Fu C M. -01-18)[2019-01-25]. Zhao Y. Extend the shallow part of single shot MultiBox detector via convolutional neural network. org/abs/1801, (2018).

【26】Ioffe S and Szegedy C. Batch normalization: accelerating deep network train by reducing internal covariate shift. [C]//Proceedings of the 32nd International Conference on Machine Learning, July 6-11, 2015, Lille, France. Cmabridge: PMLR. 448-456(2015).

【27】Li Z X. -05-17)[2019-01-25]. Zhou F Q. FSSD: feature fusion single shot MulitiBox detector. org/abs/1712, (2018).

【28】Everingham M, van Gool L, Williams C K I et al. . The Pascal visual object classes (VOC) challenge. International Journal of Computer Vision. 88(2), 303-338(2010).

引用该论文

Xiangling Ji,Jun Wu,Jianbing Yi,Xiaoguang Zhang. Automatic Detection Algorithm for Controlled Items Based on Deep Learning[J]. Laser & Optoelectronics Progress, 2019, 56(18): 180402

吉祥凌,吴军,易见兵,张晓光. 基于深度学习的管制物品自动检测算法研究[J]. 激光与光电子学进展, 2019, 56(18): 180402

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