SIA OpenIR  > 光电信息技术研究室
Dual-NMS: A method for autonomously removing false detection boxes from aerial image object detection results
Lin ZY(林智远)1,2,3,4,5; Wu QX(吴清潇)1,2,4,5; Fu SF(付双飞)1,2,4,5; Wang SK(王思奎)1,2,3,4,5; Zhang ZY(张钟毓)1,2,3,4,5; Kong YZ(孔研自)1,2,3,4,5
Department光电信息技术研究室
Source PublicationSensors (Switzerland)
ISSN1424-8220
2019
Volume19Issue:21Pages:1-18
Indexed BySCI ; EI
EI Accession number20194507625841
WOS IDWOS:000498834000088
Contribution Rank1
Funding OrganizationNational Natural Science Foundation of China (U1713216)
Keywordfalse detection boxes density of detection boxes dual-NMS object detection aerial image deep learning
Abstract

In the field of aerial image object detection based on deep learning, it’s difficult to extract features because the images are obtained from a top-down perspective. Therefore, there are numerous false detection boxes. The existing post-processing methods mainly remove overlapped detection boxes, but it’s hard to eliminate false detection boxes. The proposed dual non-maximum suppression (dual-NMS) combines the density of detection boxes that are generated for each detected object with the corresponding classification confidence to autonomously remove the false detection boxes. With the dual-NMS as a post-processing method, the precision is greatly improved under the premise of keeping recall unchanged. In vehicle detection in aerial imagery (VEDAI) and dataset for object detection in aerial images (DOTA) datasets, the removal rate of false detection boxes is over 50%. Additionally, according to the characteristics of aerial images, the correlation calculation layer for feature channel separation and the dilated convolution guidance structure are proposed to enhance the feature extraction ability of the network, and these structures constitute the correlation network (CorrNet). Compared with you only look once (YOLOv3), the mean average precision (mAP) of the CorrNet for DOTA increased by 9.78%. Commingled with dual-NMS, the detection effect in aerial images is significantly improved.

Language英语
Citation statistics
Document Type期刊论文
Identifierhttp://ir.sia.cn/handle/173321/25870
Collection光电信息技术研究室
Affiliation1.Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, China
2.Institutes for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang 110169, China
3.University of Chinese Academy of Sciences, Beijing 100049, China
4.Key Laboratory of Opto-Electronic Information Processing, Chinese Academy of Sciences, Shenyang 110016, China
5.The Key Lab of Image Understanding and Computer Vision, Shenyang 110016, China
Recommended Citation
GB/T 7714
Lin ZY,Wu QX,Fu SF,et al. Dual-NMS: A method for autonomously removing false detection boxes from aerial image object detection results[J]. Sensors (Switzerland),2019,19(21):1-18.
APA Lin ZY,Wu QX,Fu SF,Wang SK,Zhang ZY,&Kong YZ.(2019).Dual-NMS: A method for autonomously removing false detection boxes from aerial image object detection results.Sensors (Switzerland),19(21),1-18.
MLA Lin ZY,et al."Dual-NMS: A method for autonomously removing false detection boxes from aerial image object detection results".Sensors (Switzerland) 19.21(2019):1-18.
Files in This Item: Download All
File Name/Size DocType Version Access License
Dual-NMS_ A method f(4996KB)期刊论文出版稿开放获取CC BY-NC-SAView Download
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Lin ZY(林智远)]'s Articles
[Wu QX(吴清潇)]'s Articles
[Fu SF(付双飞)]'s Articles
Baidu academic
Similar articles in Baidu academic
[Lin ZY(林智远)]'s Articles
[Wu QX(吴清潇)]'s Articles
[Fu SF(付双飞)]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Lin ZY(林智远)]'s Articles
[Wu QX(吴清潇)]'s Articles
[Fu SF(付双飞)]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: Dual-NMS_ A method for autonomously removing false detection boxes from aerial image object detection results.pdf
Format: Adobe PDF
This file does not support browsing at this time
All comments (0)
No comment.
 

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