Tiny YOLOV3目标检测改进

巩笑天; 欧阳航空

doi:10.3788/OPE.20202804.0988

您当前的位置：

首页 >

文章列表页 >

Tiny YOLOV3目标检测改进

信息科学 | 更新时间：2020-08-13

- Tiny YOLOV3目标检测改进
- Improvement of Tiny YOLOV3 target detection
- 光学精密工程 2020年28卷第4期页码：988-995
- 作者机构：
  
  上海大学机电工程与自动化学院，上海 200444
- 作者简介：
  
  欧阳航空(1976-)，男，湖南人，工程师，2005年于上海大学获得硕士学位，现为上海大学机电工程与自动化学院教师，主要研究方向为精密运动控制技术。E-mail:hankow@shu.edu.cn OUYANG Hang-kong, E-mail:hankow@shu.edu.cn
- 基金信息：
  
  国家自然科学基金资助项目(No.61573238);国家重点研发计划资助项目(No. 2018YFB1309200)
- DOI：10.3788/OPE.20202804.0988
  中图分类号： TP391
- 修回日期：2020-01-01，
  
  录用日期：2020-1-1，
  
  纸质出版日期：2020-04-15
- 稿件说明：
移动端阅览
巩笑天, 欧阳航空. Tiny YOLOV3目标检测改进[J]. 光学精密工程, 2020,28(4):988-995.

Xiao-tian GONG, Hang-kong OUYANG. Improvement of Tiny YOLOV3 target detection[J]. Optics and precision engineering, 2020, 28(4): 988-995.
巩笑天, 欧阳航空. Tiny YOLOV3目标检测改进[J]. 光学精密工程, 2020,28(4):988-995. DOI： 10.3788/OPE.20202804.0988.

Xiao-tian GONG, Hang-kong OUYANG. Improvement of Tiny YOLOV3 target detection[J]. Optics and precision engineering, 2020, 28(4): 988-995. DOI： 10.3788/OPE.20202804.0988.

摘要

针对Tiny YOLOV3目标检测算法在实时检测中对行人等小目标漏检率高的问题，对该算法的特征提取网络、预测网络、损失函数等进行研究改进。首先，在特征提取网络中增加2步长的卷积层，代替原网络中的最大池化层进行下采样；接着，使用深度可分离卷积构造反残差块替换传统卷积，降低模型尺寸和参数量，增加高维特征提取；然后，在原网络两尺度预测的基础上增加一尺度，形成三尺度预测；最后，对损失函数中的边界框位置误差项进行优化。实验结果表明，改进后的Tiny YOLOV3算法的目标检测准确率比原算法提高了9.8%，满足实时性要求，具有一定鲁棒性。本文方法能够更好地提取目标特征，多尺度预测和边界框位置误差的改进能更准确地对目标进行检测。

Abstract

The Tiny YOLOV3 target detection algorithm has a high error rate for small targets

such as pedestrians

in real-time detection.Therefore

this study aimed to improve the feature extraction network

prediction network

and loss function of the algorithm.First

a two-step convolution layer was added to the feature extraction network to replace the maximum pooling layer in the original network for downsampling. Second

the traditional convolution was replaced with an anti-residual block constructed by a deep convolutional convolution to reduce the model size as well as number of parameters and increase the high-dimensional feature extraction.Third

based on the original two-scale prediction of the network

a scale was added to form a three-scale prediction. Finally

the boundary box position error in the loss function was optimized.The experimental results demonstrat that the improved Tiny YOLOV3 algorithm achieve a target detection accuracy that IS 9.8% higher than the original algorithm

satisfied the real-time requirement

and demonstratedrobustness.The proposed method can better extract target features

and the multi-scale prediction and improvement of the boundary box position error can detect targets more accurately.

关键词

Keywords

references

G RÄTSCH , T ONODA , K R MÜLLER . Soft margins for Ada-Boost . Machine Learning , 2001 . 42 ( 3 ): 287 - 320 . http://cn.bing.com/academic/profile?id=e560f7f42dc14e6f528e7a197b035ac5&encoded=0&v=paper_preview&mkt=zh-cn http://cn.bing.com/academic/profile?id=e560f7f42dc14e6f528e7a197b035ac5&encoded=0&v=paper_preview&mkt=zh-cn .

丁鹏 , 张叶 , 刘让 , 等 . 结合形态学和Canny算法的红外弱小目标检测 . 液晶与显示 , 2016 . 31 ( 8 ): 793 - 800 . http://d.old.wanfangdata.com.cn/Periodical/yjyxs201608010 http://d.old.wanfangdata.com.cn/Periodical/yjyxs201608010 .

P DING , Y ZHANG , R LIU , 等 . Infrared small target detection based on adaptive Canny algorithm and morphology . Chinese Journal of Liquid Crystal and Displays , 2016 . 31 ( 8 ): 793 - 800 . http://d.old.wanfangdata.com.cn/Periodical/yjyxs201608010 http://d.old.wanfangdata.com.cn/Periodical/yjyxs201608010 .

张国梁 , 贾松敏 , 张祥银 , 等 . 采用自适应变异粒子群优化SVM的行为识别 . 光学精密工程 , 2017 . 25 ( 6 ): 1669 - 1678 . http://www.eope.net/CN/abstract/abstract17080.shtml http://www.eope.net/CN/abstract/abstract17080.shtml .

G L ZHANG , S M JIA , X Y ZHANG , 等 . Action recognition based on adaptive mutation particle swarm optimization for SVM . Editorial Office of Optics and Precision Engineering , 2017 . 25 ( 6 ): 1669 - 1678 . http://www.eope.net/CN/abstract/abstract17080.shtml http://www.eope.net/CN/abstract/abstract17080.shtml .

梁华 , 宋玉龙 , 钱锋 , 等 . 基于深度学习的航空对地小目标检测 . 液晶与显示 , 2018 . 33 ( 9 ): 793 - 800 . http://d.old.wanfangdata.com.cn/Periodical/yjyxs201809011 http://d.old.wanfangdata.com.cn/Periodical/yjyxs201809011 .

H LIANG , Y L SONG , F QIAN , 等 . Detection of small target in aerial photography based on deep learning . Chinese Journal of Liquid Crystal and Displays , 2018 . 33 ( 9 ): 793 - 800 . http://d.old.wanfangdata.com.cn/Periodical/yjyxs201809011 http://d.old.wanfangdata.com.cn/Periodical/yjyxs201809011 .

GIRSHICK R. Fast R-CNN[C]. Proceedings of the IEEE international conference on computer vision . 2015: 1440-1448.

REN S, HE K, GIRSHICK R, et al . Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks[C]. Proceedings of International Conference on Neural Information Processing Systems , 2015: 91-99.

HE K, GKIOXARI G, DOLLÁR P, et al . Mask R-CNN[C]. Computer Vision ( ICCV ), 2017 IEEE International Conference on. IEEE , 2017: 2980-2988.

FU C Y, LIU W, RANGA A, et al . DSSD: Deconvolutional single shot detector[J]. arXiv preprint arXiv: 1701.06659, 2017.

REDMON J, DIVVALA S, GIRSHICK R, et al . You Only Look Once: Unified, Real-Time Object Detection[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2016: 779-788.

REDMON J, FARHADI A. YOLO9000: Better, faster, stronger[J]. arXiv preprint arXiv , 1612.08242, 2017.

REDMON J, FARHADI A. Yolov3: An incremental improvement[J]. arXiv preprint arXiv , 1804.02767, 2018.

LIN T Y, DOLL?R P, GIRSHICK R, et al . Feature pyramid networks for object detection[J]. arXiv preprint arXiv: 1612.03144, 2016.

NOH H, HONG S, HAN B. Learning Deconvolution Network for Semantic Segmentation[C]. IEEE International Conference on Computer Vision , 2016: 1520-1528.

刘智 , 黄江涛 , 冯欣 . 构建多尺度深度卷积神经网络行为识别模型 . 光学精密工程 , 2017 . 25 ( 3 ): 799 - 805 . http://www.eope.net/CN/abstract/abstract16952.shtml http://www.eope.net/CN/abstract/abstract16952.shtml .

ZH LIU , J T HUANG , X FENG . Construction of multi-scale deep convolutional neural network behavior recognition model . Opt. Precision Eng , 2017 . 25 ( 3 ): 799 - 805 . http://www.eope.net/CN/abstract/abstract16952.shtml http://www.eope.net/CN/abstract/abstract16952.shtml .

杨州 , 慕晓冬 , 王舒洋 , 等 . 基于多尺度特征融合的遥感图像场景分类 . 光学精密工程 , 2018 . 26 ( 12 ): 232 - 240 . http://www.eope.net/CN/abstract/abstract17866.shtml http://www.eope.net/CN/abstract/abstract17866.shtml .

YANG ZH YANG , X D MU , SH Y WANG , 等 . Remote sensing image scene classification based on multiscale feature fusion . Opt. Precision Eng , 2018 . 26 ( 12 ): 232 - 240 . http://www.eope.net/CN/abstract/abstract17866.shtml http://www.eope.net/CN/abstract/abstract17866.shtml .

刘伟宁 . 基于小波域扩散滤波的弱小目标检测 . 中国光学 , 2011 . 4 ( 5 ): 503 - 508 . DOI: 10.3969/j.issn.2095-1531.2011.05.015 http://doi.org/10.3969/j.issn.2095-1531.2011.05.015 .

W N LIU . Dim target detection based on wavelet field diffusion filter . Chinese Optics , 2011 . 4 ( 5 ): 503 - 508 . DOI: 10.3969/j.issn.2095-1531.2011.05.015 http://doi.org/10.3969/j.issn.2095-1531.2011.05.015 .

刘希佳 , 陈宇 , 王文生 , 等 . 小目标识别的小波阈值去噪方法 . 中国光学 , 2012 . 5 ( 3 ): 248 - 256 . DOI: 10.3969/j.issn.2095-1531.2012.03.009 http://doi.org/10.3969/j.issn.2095-1531.2012.03.009 .

X J LIU , Y CHEN , W SH WANG , 等 . De-noising algorithm of wavelet threshold for small target detection . Chinese Optics , 2012 . 5 ( 3 ): 248 - 256 . DOI: 10.3969/j.issn.2095-1531.2012.03.009 http://doi.org/10.3969/j.issn.2095-1531.2012.03.009 .

浏览量

314

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

跨层注意力交互下的多特征交叉无人机图像检测

适应大范围星场密度变化的恒星辨识与抑制

利用光谱解混合的目标检测

强化特征提取能力的下颌骨骨折检测3M-YOLOv5网络

面向无人机平台的轻量化目标检测网络