SIA OpenIR  > 光电信息技术研究室
基于改进Faster R-CNN的子弹外观缺陷检测技术研究
Alternative TitleStudy on Bullet Surface Defect Detection Technology Based on Improved Faster R-CNN
Thesis Advisor朱丹
Keyword目标检测 Faster R-CNN 子弹外观缺陷 K-means++ 卷积神经网络
Degree Discipline控制理论与控制工程
Degree Name硕士
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract本文主要围绕如何提高子弹外观缺陷检测模型的检测精度和速度进行研究和分析。首先介绍了四种基于候选区域的深度学习目标检测算法,通过实验对比四种算法在PASCAL VOC数据集的检测性能,基于对比结果,本文选择Faster R-CNN作为子弹外观缺陷检测的基础模型。以PASCAL VOC数据集格式标准制作子弹外观缺陷数据集,搭建实验所需软件和硬件环境。将检测模型分别与ZFNet、VGG_CNN_M_1024和VGG16三种不同深度卷积神经网络结合,在子弹缺陷数据集进行检测,检测结果表明,采用较深层网络VGG16的Faster R-CNN模型检测性能优于其余两种模型方案。针对子弹外观缺陷数据集,本文在Faster R-CNN基础上结合K-means++改进锚框的生成方法,得到基于改进Faster R-CNN的子弹外观缺陷检测模型。实验结果表明,改进模型在原文算法基础上检测精度与速度均有所提升。
Other AbstractThis dissertation mainly focuses on how to improve the detection accuracy and speed of the bullet surface defect detection model. Firstly, four algorithms based on region proposal for deep learning object detection are introduced. The detection performance of four algorithms in PASCAL VOC dataset is compared experimentally. Based on the comparison results, this dissertation chooses Faster R-CNN as the basic model for bullet surface defect detection. The bullet surface defect data set was created by PASCAL VOC dataset format standard, and the software and hardware environment required for the experiment were built. Then the detection model is combined with three different depth convolutional neural networks respectively, which are ZFNet, VGG_CNN_M_1024 and VGG16, and detected in the bullet surface defect dataset. The detection results show that the Faster R-CNN model with deeper network VGG16 has better detection performance than the other two model schemes. Aiming at the bullet surface defect dataset, this dissertation combines K-means++ to improve the anchor frame generation method based on Faster R-CNN, and obtains the bullet surface defect detection model based on improved Faster R-CNN. The experimental results show that the detection accuracy and speed of the improved model are increased based on the original algorithm.
Contribution Rank1
Document Type学位论文
Recommended Citation
GB/T 7714
马晓云. 基于改进Faster R-CNN的子弹外观缺陷检测技术研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
基于改进Faster R-CNN的子弹外(2830KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[马晓云]'s Articles
Baidu academic
Similar articles in Baidu academic
[马晓云]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[马晓云]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.