SIA OpenIR  > 光电信息技术研究室
改进的YOLO V3及其在小目标检测中的应用
Alternative TitleImproved YOLO V3 and its application in small target detection
鞠默然1,2,3,4,5; 罗海波1,2,4,5; 王仲博1,2,3,4,5; 何淼1,2,3,4,5; 常铮1,2,4,5; 惠斌1,2,4,5
Department光电信息技术研究室
Source Publication光学学报
ISSN0253-2239
2019
Volume39Issue:7Pages:1-8
Indexed ByEI
EI Accession number20193607400976
Contribution Rank1
Keyword机器视觉 小目标检测 YOLO V3 VEDAI数据集 K-means
Abstract针对图像中小目标检测率低、虚警率高等问题,提出了一种YOLO V3的改进方法,并将其专用于小目标的检测。由于小目标所占的像素少,特征不明显,提出对原网络输出的8倍降采样特征图进行2倍的上采样,与第二个残差块(Res block)输出的特征图进行拼接,建立输出为4倍降采样的特征融合目标检测层。为了获取更多的小目标特征信息,在YOLO V3网络结构Darknet53的第二个残差块中增加2个残差单元(Resnet unit)。利用K-means聚类算法对目标候选框的个数和宽高比维度进行聚类分析。用改进的网络在VEDAI航拍车辆数据集上与YOLO V3进行对比试验,结果表明改进后的网络能有效检测小目标,对小目标的召回率和检测的平均准确率都有明显的提升。
Other AbstractAiming at the problems of low detection rate and high false alarm rate of the small targets in the image, an improved detection algorithm of YOLO V3 is proposed and specially applied in small target detection. Because the resolution of small targets is low and the features are not obvious, it is proposed to upsample by 2× the feature map which is downsampled by 8× of the previous network, and concatenate it with the output of the second Res block unit. The feature fusion target detection layer whose feature map is downsampled by 4× is established. To obtain more features of the small target, add two Resnet units in the second Res block unit of Darknet53 in YOLO V3 network structure. K-means clustering algorithm is used to select the number of the candidate anchor boxes and the aspect ratio dimensions. The comparative experiment on VEDAI dataset with YOLO V3 shows that the improved network can detect small targets efficiently and improve the accuracy rate and the recall rate of the small targets.
Language中文
Document Type期刊论文
Identifierhttp://ir.sia.cn/handle/173321/24460
Collection光电信息技术研究室
Corresponding Author鞠默然
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院机器人与智能制造创新研究院
3.中国科学院大学
4.中国科学院光电信息处理重点实验室
5.辽宁省图像理解与视觉计算重点实验室
Recommended Citation
GB/T 7714
鞠默然,罗海波,王仲博,等. 改进的YOLO V3及其在小目标检测中的应用[J]. 光学学报,2019,39(7):1-8.
APA 鞠默然,罗海波,王仲博,何淼,常铮,&惠斌.(2019).改进的YOLO V3及其在小目标检测中的应用.光学学报,39(7),1-8.
MLA 鞠默然,et al."改进的YOLO V3及其在小目标检测中的应用".光学学报 39.7(2019):1-8.
Files in This Item: Download All
File Name/Size DocType Version Access License
改进的YOLO_V3算法及其在小目标检测(2348KB)期刊论文作者接受稿开放获取CC BY-NC-SAView Application Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[鞠默然]'s Articles
[罗海波]'s Articles
[王仲博]'s Articles
Baidu academic
Similar articles in Baidu academic
[鞠默然]'s Articles
[罗海波]'s Articles
[王仲博]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[鞠默然]'s Articles
[罗海波]'s Articles
[王仲博]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: 改进的YOLO_V3算法及其在小目标检测中的应用.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

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