首页 > 论文 > 激光与光电子学进展 > 57卷 > 20期(pp:201009--1)

基于改进MDSSD的小目标实时检测算法

Small Target Real-Time Detection Algorithm Based on Improved MDSSD

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

摘要

与常规目标相比,小目标在图像中的像素占比较小且特征不明显,而基于卷积神经网络的目标检测算法对小目标的鲁棒性较差。因此,基于改进的多尺寸反卷积单次多目标检测(MDSSD)网络提出了一种多特征融合小目标实时检测算法,可在不影响实时性的前提下增强算法对小目标的检测能力。为了提高算法的实时性,删减了MDSSD网络中的一部分融合模块,并用双线性插值操作代替特征图尺寸变换过程中的反卷积操作。为了提高算法的准确率,在特征图预测阶段,对每个预测层添加残差预测模块。针对算法中存在正负样本严重失衡的问题,用Focal Loss作为分类损失函数。实验结果表明,本算法在PASCAL VOC2007数据集上的平均精度均值(mAP)为81.7%,在自制轮胎瑕疵数据集上的mAP为90.6%,每秒检测帧数为50 frame。

Abstract

Compared with common target, small target occupies smaller proportion of pixel in the image, the feature is not obvious, and target detection algorithm based on convolutional neural network is less robust to the small target. Therefore, this paper proposes a small target real-time detection algorithm based on multi-scale deconvolutional single shot detector (MDSSD) network, which can enhance the algorithm''s ability to detect small targets without affecting the real-time performance. In order to improve the real-time performance of the algorithm, a part of the fusion module in the MDSSD network is deleted, and the deconvolution operation in the process of the feature map size transformation is replaced by a bilinear interpolation operation. In order to improve the accuracy of the algorithm, in the feature map prediction stage, a residual prediction module is added to each prediction layer. Aiming at the problem of serious imbalance between positive and negative samples in the algorithm, Focal Loss is used as the classification loss function. Experimental results show the mean average precision (mAP) of this algorithm on the PASCAL VOC2007 dataset is 81.7%, the mAP on self-made tire defect dataset is 90.6%, and 50 frames per second is detected.

广告组1 - 空间光调制器+DMD
补充资料

中图分类号:TP391.4

DOI:10.3788/LOP57.201009

所属栏目:图像处理

基金项目:国家自然科学基金、国家重点研发计划、教育部-中国移动科研项目;

收稿日期:2019-12-17

修改稿日期:2020-02-25

网络出版日期:2020-10-01

作者单位    点击查看

奚琦:江南大学物联网工程学院, 江苏 无锡 214122
张正道:江南大学物联网工程学院, 江苏 无锡 214122江南大学物联网应用技术教育部工程研究中心, 江苏 无锡 214122
彭力:江南大学物联网工程学院, 江苏 无锡 214122江南大学物联网应用技术教育部工程研究中心, 江苏 无锡 214122

联系人作者:张正道(wxzzd@hotmail.com)

备注:国家自然科学基金、国家重点研发计划、教育部-中国移动科研项目;

【1】Chen L L, Zhang Z D, Peng L. Real-time detection based on improved single shot MultiBox detector [J]. Laser & Optoelectronics Progress. 2019, 56(1): 011002.
陈立里, 张正道, 彭力. 基于改进SSD的实时检测方法 [J]. 激光与光电子学进展. 2019, 56(1): 011002.

【2】Feng X Y, Mei W, Hu D S. Aerial target detection based on improved Faster R-CNN [J]. Acta Optica Sinica. 2018, 38(6): 0615004.
冯小雨, 梅卫, 胡大帅. 基于改进Faster R-CNN的空中目标检测 [J]. 光学学报. 2018, 38(6): 0615004.

【3】Liu W, Anguelov D, Erhan D, et al. SSD: single shot MultiBox detector [M]. ∥Leibe B, Matas J, Sebe N, et al. Computer Vision-ECCV 2016. Lecture Notes in Computer Science. Cham: Springer. 2016, 9905: 21-37.

【4】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. 2016, 779-788.

【5】He S, Yang S, Shao X, et al. Ship target detection on the sea surface based on natural measure feature of image block [J]. Infrared and Laser Engineering. 2011, 40(9): 1812-1817.

【6】Luo H B, Xu L Y, Hui B, et al. Status and prospect of target tracking based on deep learning [J]. Infrared and Laser Engineering. 2017, 46(5): 0502002.

【7】Lee D S. Effective Gaussian mixture learning for video background subtraction [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2005, 27(5): 827-832.

【8】Horn B K P, Schunck B G. Determining optical flow [J]. Artificial Intelligence. 1981, 17(1/2/3): 185-203.

【9】Fu C Y, Liu W, Ranga A, et al. DSSD: deconvolutional single shot detector[2019-12-03].https:∥arxiv.[2019-12-03]. 0, org/abs/1701: 06659.

【10】Jeong J, Park H, Kwak N. Enhancement of SSD by concatenating feature maps for object detection[2019-12-02].https://arxiv.[2019-12-02]. 0, org/abs/1705: 09587.

【11】Cui L S, Ma R, Lü P, et al. MDSSD: multi-scale deconvolutional single shot detector for small objects[2019-12-03].https://arxiv.[2019-12-03]. 0, org/abs/1805: 07009.

【12】Cao G M, Xie X M, Yang W Z, et al. Feature-fused SSD: fast detection for small objects [J]. Proceedings of SPIE. 2018, 1061: 106151E.

【13】Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[2019-11-30].https://arxiv.[2019-11-30]. 0, org/abs/1409: 1556.

【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. 2016, 770-778.

【15】Lin T Y, Goyal P, Girshick R, et al. Focal Loss for dense object detection . [C]∥2017 IEEE International Conference on Computer Vision (ICCV), October 22-29, 2017, Venice, Italy. New York: IEEE. 2017, 2999-3007.

引用该论文

Xi Qi,Zhang Zhengdao,Peng Li. Small Target Real-Time Detection Algorithm Based on Improved MDSSD[J]. Laser & Optoelectronics Progress, 2020, 57(20): 201009

奚琦,张正道,彭力. 基于改进MDSSD的小目标实时检测算法[J]. 激光与光电子学进展, 2020, 57(20): 201009

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