Underwater Target Recognition Based on Improved YOLOv4 Neural Network

Underwater Target Recognition Based on Improved YOLOv4 Neural Network

The YOLOv4 neural network is employed for underwater target recognition. To improve the accuracy and speed of recognition, the structure of YOLOv4 is modified by replacing the upsampling module with a deconvolution module and by incorporating depthwise separable convolution into the network. Moreove...

Full description

Bibliographic Details
Main Authors: Lingyu Chen, Meicheng Zheng, Shunqiang Duan, Weilin Luo, Ligang Yao
Format: Article
Language:English
Published: MDPI AG 2021-07-01
Series:Electronics
Subjects:
underwater image
enhancement
YOLO neural network
target recognition
mosaic data augmentation
Online Access:https://www.mdpi.com/2079-9292/10/14/1634
id doaj-88c011a19e2041199d8ba030f8657bf8
record_format Article
spelling doaj-88c011a19e2041199d8ba030f8657bf82021-07-23T13:37:59ZengMDPI AGElectronics2079-92922021-07-01101634163410.3390/electronics10141634Underwater Target Recognition Based on Improved YOLOv4 Neural NetworkLingyu Chen0Meicheng Zheng1Shunqiang Duan2Weilin Luo3Ligang Yao4College of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350108, ChinaCollege of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350108, ChinaCollege of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350108, ChinaCollege of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350108, ChinaCollege of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350108, ChinaThe YOLOv4 neural network is employed for underwater target recognition. To improve the accuracy and speed of recognition, the structure of YOLOv4 is modified by replacing the upsampling module with a deconvolution module and by incorporating depthwise separable convolution into the network. Moreover, the training set used in the YOLO network is preprocessed by using a modified mosaic augmentation, in which the gray world algorithm is used to derive two images when performing mosaic augmentation. The recognition results and the comparison with the other target detectors demonstrate the effectiveness of the proposed YOLOv4 structure and the method of data preprocessing. According to both subjective and objective evaluation, the proposed target recognition strategy can effectively improve the accuracy and speed of underwater target recognition and reduce the requirement of hardware performance as well.https://www.mdpi.com/2079-9292/10/14/1634underwater imageenhancementYOLO neural networktarget recognitionmosaic data augmentation
collection DOAJ
language English
format Article
sources DOAJ
author Lingyu Chen
Meicheng Zheng
Shunqiang Duan
Weilin Luo
Ligang Yao
spellingShingle Lingyu Chen
Meicheng Zheng
Shunqiang Duan
Weilin Luo
Ligang Yao
Underwater Target Recognition Based on Improved YOLOv4 Neural Network
Electronics
underwater image
enhancement
YOLO neural network
target recognition
mosaic data augmentation
author_facet Lingyu Chen
Meicheng Zheng
Shunqiang Duan
Weilin Luo
Ligang Yao
author_sort Lingyu Chen
title Underwater Target Recognition Based on Improved YOLOv4 Neural Network
title_short Underwater Target Recognition Based on Improved YOLOv4 Neural Network
title_full Underwater Target Recognition Based on Improved YOLOv4 Neural Network
title_fullStr Underwater Target Recognition Based on Improved YOLOv4 Neural Network
title_full_unstemmed Underwater Target Recognition Based on Improved YOLOv4 Neural Network
title_sort underwater target recognition based on improved yolov4 neural network
publisher MDPI AG
series Electronics
issn 2079-9292
publishDate 2021-07-01
description The YOLOv4 neural network is employed for underwater target recognition. To improve the accuracy and speed of recognition, the structure of YOLOv4 is modified by replacing the upsampling module with a deconvolution module and by incorporating depthwise separable convolution into the network. Moreover, the training set used in the YOLO network is preprocessed by using a modified mosaic augmentation, in which the gray world algorithm is used to derive two images when performing mosaic augmentation. The recognition results and the comparison with the other target detectors demonstrate the effectiveness of the proposed YOLOv4 structure and the method of data preprocessing. According to both subjective and objective evaluation, the proposed target recognition strategy can effectively improve the accuracy and speed of underwater target recognition and reduce the requirement of hardware performance as well.
topic underwater image
enhancement
YOLO neural network
target recognition
mosaic data augmentation
url https://www.mdpi.com/2079-9292/10/14/1634
work_keys_str_mv AT lingyuchen underwatertargetrecognitionbasedonimprovedyolov4neuralnetwork
AT meichengzheng underwatertargetrecognitionbasedonimprovedyolov4neuralnetwork
AT shunqiangduan underwatertargetrecognitionbasedonimprovedyolov4neuralnetwork
AT weilinluo underwatertargetrecognitionbasedonimprovedyolov4neuralnetwork
AT ligangyao underwatertargetrecognitionbasedonimprovedyolov4neuralnetwork
_version_ 1721288687093809152

Similar Items