基于RGHS图像增强和改进YOLOv5网络的黄颡鱼目标提取

doi:10.13229/j.cnki.jdxbgxb.20230446

摘要/Abstract

摘要：

针对水下能见度不佳，黄颡鱼目标提取精度低、速度慢等问题，提出了基于相对全局直方图拉伸（RGHS）算法和改进YOLOv5的黄颡鱼目标提取模型。首先，为解决光照不均、噪声大等因素带来的图像质量问题，采用RGHS算法对黄颡鱼图像进行亮度增强。然后，在YOLOv5主干网络中引入C3ghost模块和坐标注意力（CA）机制，在颈部网络中用 $g n$ Conv替换普通卷积，建立改进YOLOv5模型，提升黄颡鱼目标提取精度。结果表明，改进模型的AP值、准确率、召回率比YOLOv5模型分别提升了2.76%、3.16%、3.1%，F1值提升了0.03，所占内存减少了2.3 MB，单张图片推理时间减少了0.001 s。同时，在与YOLOv4、SSD、Faster-RCNN、YOLOx模型的对比实验中，改进模型的AP值分别提升了3.27%、8.63%、2.48%、2.52%。基于RGHS图像增强的改进YOLOv5模型在保证较快速度的情况下，显著提高了黄颡鱼目标提取精度，可为鱼类状态监测方法的研究提供有益参考。

关键词: 计算机应用, 目标提取, 亮度增强, 注意力机制, 深度学习

Abstract:

Aiming at the problems of poor underwater visibility， low accuracy and slow speed of object extraction， a yellow catfish object extraction model based on RGHS algorithm and improved YOLOv5 was proposed. Firstly， in order to solve the image quality problems caused by uneven illumination and high noise， RGHS algorithm was used to enhance the brightness of yellow catfish image. Then， C3ghost and CA attention mechanisms were introduced into the YOLOv5 backbone network， and $g n$ Conv was used to replace the common convolution in the neck part， so as to establish an improved YOLOv5 model and improve the target extraction accuracy of yellow catfish. The results show that compared with YOLOv5， the AP value， accuracy rate and recall rate of the improved model are increased by 2.76%， 3.16% and 3.1 %respectively， the F1 value is increased by 0.03， the memory occupied by the improved model is reduced by 2.3 MB， and the reasoning time of a single image is reduced by 0.001 s. Meanwhile， compared with the YOLOv4， SSD， Faster-RCNN and YOLOx models， the AP values of the improved models are increased by 3.27%， 8.63%， 2.48% and 2.52% respectively. The improved YOLOv5 model based on RGHS image enhancement can significantly improve the target extraction accuracy of yellow catfish while maintaining a fast speed， which can provide useful reference for the study of fish status monitoring methods.

Key words: computer application, target extraction, brightness enhancement, attention mechanism, deep learning

中图分类号:

TP391

李路,宋均琦,朱明,谭鹤群,周玉凡,孙超奇,周铖钰. 基于RGHS图像增强和改进YOLOv5网络的黄颡鱼目标提取[J]. 吉林大学学报(工学版), 2024, 54(9): 2638-2645.

Lu Li,Jun-qi Song,Ming Zhu,He-qun Tan,Yu-fan Zhou,Chao-qi Sun,Cheng-yu Zhou. Object extraction of yellow catfish based on RGHS image enhancement and improved YOLOv5 network[J]. Journal of Jilin University(Engineering and Technology Edition), 2024, 54(9): 2638-2645.

图/表 14

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

表1

图11

表2

图12

参考文献 25

1	Jiao L C, Zhang R H, Liu F, et al. New generation deep learning for video object detection: a survey[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022, 33(8): 3195-3215.
2	Zhang L, Li W, Liu C, et al. Automatic fish counting method using image density grading and local regression[J]. Computers and Electronics in Agriculture, 2020, 179: No.105844.
3	Aliyu I, Gana K J, Musa A A, et al. Incorporating recognition in catfish counting algorithm using artificial neural network and geometry[J]. KSII Transactions on Internet and Information Systems, 2020, 14(12): 4866-4888.
4	Garcia R, Prados R, Quintana J, et al. Automatic segmentation of fish using deep learning with application to fish size measurement[J]. ICES Journal of Marine Science, 2020, 77(4): 1354-1366.
5	Yu C, Liu Y, Hu Z, et al. Precise segmentation and measurement of inclined fish's features based on U-net and fish morphological characteristics[J]. Applied Engineering in Agriculture, 2022, 38(1): 37-48.
6	Yasruddin M L, Ismail M A H, Husin Z, et al. Feasibility study of fish disease detection using computer vision and deep convolutional neural network (dcnn) algorithm[C]∥2022 IEEE 18th International Colloquium on Signal Processing & Applications, Selangor, Malaysia, 2022: 272-276.
7	Zhang J, Pang H, Cai W, et al. Using image processing technology to create a novel fry counting algorithm[J]. Aquaculture and Fisheries, 2022, 7(4): 441-449.
8	Guillaud A, Troadec H, Benzinou A, et al. A multiagent system for edge detection and continuity perception on fish otolith images[J]. EURASIP Journal on Advances in Signal Processing, 2002, 7: 746-753.
9	Deka J, Laskar S. Comparative Analysis of FOD based prewitt, sobel & laplacian operators for edge detection on freshwater fish images[C]∥2020 International Conference on Emerging Smart Computing and Informatics, Madurai, India, 2020: 65-70.
10	Gao T S, Sheng D B, Nguyen T H, et al. Measurement of the fish body wound depth based on a depth map inpainting method[C]∥AETA 2016: Recent Advances in Electrical Engineering and Related Sciences: Theory and Application, Busan, Korea, 2017: 289-299.
11	Abdeldaim A M, Houssein E H, Hassanien A E. Color image segmentation of fishes with complex background in water[C]∥The International Conference on Advanced Machine Learning Technologies and Applications,Cairo, Egypt, 2018: 634-643.
12	Xu G, Chen Q, Yoshida T, et al. Detection of bluefin tuna by cascade classifier and deep learning for monitoring fish resources[C]∥Global Oceans 2020: Singapore-US Gulf Coast, Biloxi, MS, USA, 2020: 1-4.
13	谭鹤群,李玉祥,朱明,等. 通过图像增强与改进Faster-RCNN网络的重叠鱼群尾数检测[J]. 农业工程学报,2022,38(13):167-176.
	TAN He-qun, Li Yu-xiang, ZHU Ming, et al. Detecting overlapping fish population using image enhancement and improved Faster-RCNN networks[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2022, 38(13):167-176.
14	Zhang W, Dong L, Pan X, et al. Single image defogging based on multi-channel convolutional MSRCR[J]. IEEE Access, 2019, 7: 72492-72504.
15	郝琨,王阔,赵璐,等. 基于图像增强与改进YOLOv3的水下生物检测算法[J]. 吉林大学学报: 工学版, 2022, 52(5):1088-1097.
	Hao Kun, Wang Kuo, Zhao Lu, et al. Underwater biological detection algorithm based on image enhancement and improved YOLOv3[J]. Journal of Jilin University (Engineering and Technology Edition), 2022, 52(5):1088-1097.
16	Cai K, Miao X, Wang W, et al. A modified YOLOv3 model for fish detection based on MobileNetv1 as backbone[J]. Aquacultural Engineering, 2020, 91: No.102117.
17	Li J, Liu C, Lu X, et al. CME-YOLOv5: an efficient object detection network for densely spaced fish and small targets[J]. Water, 2022, 14(15): 2412.
18	Huang D, Wang Y, Song W, et al. Shallow-water image enhancement using relative global histogram stretching based on adaptive parameter acquisition[C]∥MultiMedia Modeling: 24th International Conference, Bangkok, Thailand, 2018: 453-465.
19	刘彦呈,董张伟,朱鹏莅,等. 基于特征解耦的无监督水下图像增强[J]. 电子与信息学报,2022,44(10):3389-3398.
	Liu Yan-chen, Dong Zhang-wei, Zhu Peng-li, et al. Unsupervised underwater image enhancement based on feature decoupling[J]. Journal of Electronics & Information Technology, 2022, 44(10):3389-3398.
20	Bochkovskiy A, Wang C Y, Liao H Y M. Yolov4: Optimal speed and accuracy of object detection[J/OL]. [2023-04-20].
21	Han K, Wang Y, Tian Q, et al. Ghostnet: more features from cheap operations[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle USA, 2020: 1580-1589.
22	Hou Q, Zhou D, Feng J. Coordinate attention for efficient mobile network design[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Online Conference, 2021: 13713-13722.
23	Rao Y M, Zhao W L, Tang Y S, et al. Hornet: efficient high-order spatial interactions with recursive gated convolutions[J/OL]. [2023-04-22].
24	Liu W, Anguelov D, Erhan D, et al. SSD: single shot multibox detector[C]∥Computer Vision-ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, 2016: 21-37.
25	Ge Z, Liu S, Wang F, et al. Yolox: exceeding yolo series in 2021[J/OL]. [2023-04-22].

相关文章 15

[1]	张磊,焦晶,李勃昕,周延杰. 融合机器学习和深度学习的大容量半结构化数据抽取算法[J]. 吉林大学学报(工学版), 2024, 54(9): 2631-2637.
[2]	余萍,赵康,曹洁. 基于优化A-BiLSTM的滚动轴承故障诊断[J]. 吉林大学学报(工学版), 2024, 54(8): 2156-2166.
[3]	郭昕刚,程超,沈紫琪. 基于卷积网络注意力机制的人脸表情识别[J]. 吉林大学学报(工学版), 2024, 54(8): 2319-2328.
[4]	赵宏伟,武鸿,马克,李海. 基于知识蒸馏的图像分类框架[J]. 吉林大学学报(工学版), 2024, 54(8): 2307-2312.
[5]	张云佐,郑宇鑫,武存宇,张天. 基于双特征提取网络的复杂环境车道线精准检测[J]. 吉林大学学报(工学版), 2024, 54(7): 1894-1902.
[6]	乔百友,武彤,杨璐,蒋有文. 一种基于BiGRU和胶囊网络的文本情感分析方法[J]. 吉林大学学报(工学版), 2024, 54(7): 2026-2037.
[7]	郭昕刚,何颖晨,程超. 抗噪声的分步式图像超分辨率重构算法[J]. 吉林大学学报(工学版), 2024, 54(7): 2063-2071.
[8]	张丽平,刘斌毓,李松,郝忠孝. 基于稀疏多头自注意力的轨迹kNN查询方法[J]. 吉林大学学报(工学版), 2024, 54(6): 1756-1766.
[9]	孙铭会,薛浩,金玉波,曲卫东,秦贵和. 联合时空注意力的视频显著性预测[J]. 吉林大学学报(工学版), 2024, 54(6): 1767-1776.
[10]	李延风,刘名扬,胡嘉明,孙华栋,孟婕妤,王奥颖,张涵玥,杨华民,韩开旭. 基于梯度转移和自编码器的红外与可见光图像融合[J]. 吉林大学学报(工学版), 2024, 54(6): 1777-1787.
[11]	陆玉凯,袁帅科,熊树生,朱绍鹏,张宁. 汽车漆面缺陷高精度检测系统[J]. 吉林大学学报(工学版), 2024, 54(5): 1205-1213.
[12]	高云龙,任明,吴川,高文. 基于注意力机制改进的无锚框舰船检测模型[J]. 吉林大学学报(工学版), 2024, 54(5): 1407-1416.
[13]	梁礼明,周珑颂,尹江,盛校棋. 融合多尺度Transformer的皮肤病变分割算法[J]. 吉林大学学报(工学版), 2024, 54(4): 1086-1098.
[14]	张云佐,郭威,李文博. 遥感图像密集小目标全方位精准检测算法[J]. 吉林大学学报(工学版), 2024, 54(4): 1105-1113.
[15]	李晓旭,安文娟,武继杰,李真,张珂,马占宇. 通道注意力双线性度量网络[J]. 吉林大学学报(工学版), 2024, 54(2): 524-532.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

模型Model	AP值/%	准确率/%	召回率/%	调和均值F1	模型所占内存/MB
YOLOv5	93.07	92.51	87.04	0.90	14.4
YOLOv5+①	91.36	94.38	82.82	0.88	11.3
YOLOv5+①+②	92.79	95.32	86.06	0.90	11.3
YOLO5+①+②+③	93.69	94.34	89.15	0.92	11.4
YOLOv5+①+②+③+④（Improved-YOLOv5）	95.83	95.67	90.14	0.93	12.1

模型	AP值/%	准确率/%	召回率/%	参数量/M	浮点运算量/G	推理时间/s
YOLOv4	92.65	95.10	84.79	64.36	60.53	0.171
YOLOv5	93.07	92.51	87.04	7.08	17.16	0.018
SSD	87.20	94.00	75.07	26.29	62.75	0.014
Faster-RCNN	93.35	93.50	89.15	137.10	370.21	0.153
YOLOx	93.31	86.90	91.95	8.97	26.93	0.090
Improved-YOLOv5	95.83	95.67	90.14	6.51	16.38	0.017