摘 要：为掌握水下海珍品分布情况，本文结合 YOLOv5s 算法和注意力机制，得到一种新的轻量化目标检测模型——SEYOLO 模型。实验结果显示，相较于原 YOLOv5s 模型，该模型的准确率提升了 1.1%、召回率提升了 0.7%，并且在设计对比实验的过程中，发现传统图像增强算法并不具备提升目标检测准确度的可能。由此可见，本文提出的改进模型符合轻量化模型标准并兼具检测准确度高的优点，能够很好地完成对水下海珍品资源评估的任务。
关键词：深度学习；海珍品检测；YOLOv5
DOI:10.19850/j.cnki.2096-4706.2021.18.021
中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2021）18-0080-06
Sea Treasure Target Detection Algorithm Based on Improved YOLOv5
MA Zhiqiang
(Dalian Ocean University, Dalian 116023, China)
Abstract: In order to master the distribution of underwater treasures, a new lightweight target detection model, SE-YOLO model, is obtained by combining YOLOv5s algorithm and attention mechanism. The experimental results show that compared with the original YOLOv5s model, the accuracy of the model is increased by 1.1% and the recall rate is increased by 0.7%. And in the process of designing the comparison experiment, it is found that the traditional image enhancement algorithm does not have the possibility to improve the accuracy of the target detection. It can be seen that the improved model proposed in this paper conforms to the lightweight model standard and has the advantages of high detection accuracy, and can well complete the task of evaluating underwater treasure resources.
Keywords: deep learning; sea treasure detection; YOLOv5
参考文献：
[1] AGGARWAL C C.Neural Networks and Deep Learning [M]. Cham：Springer，2018.
[2] LECUN Y，BOTTOU L，BENGIO Y，et al. Gradient-based learning applied to document recognition [J].Proceedings of the IEEE， 1998，86（11）：2278-2324.
[3] REDMON J，DIVVALA S，GIRSHICK R，et al. You Only Look Once：Unified，Real-Time Object Detection [C]//IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Las Vegas：IEEE，2016：8418.
[4] REDMON J，FARHADI A.YOLO9000：Better，Faster， Stronger [C]//IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Honolulu：IEEE，2017：6517-6525.
[5] REDMON J，FARHADI A. YOLOv3：An Incremental Improvement [EB/OL].[2021-09-02] http://alumni.soe.ucsc. edu/~czczycz/src/YOLOv3.pdf.
[6] LIU W，ANGUELOV D，ERHAN D，et al. SSD：Single Shot MultiBox Detector [C]//Computer Vision – ECCV 2016.Cham：Springer，2016：21-37.
[7] REN S Q，HE K M，GIRSHICK R，et al. Faster R-CNN： Towards Real-Time Object Detection with Region Proposal Networks [J].IEEE Transactions on Pattern Analysis and Machine Intelligence， 2017，39（6）：1137-1149.
[8] 王柯力 . 深度学习在渔业图像识别中的应用研究 [D]. 上海：上海海洋大学，2018.
[9] 闫建伟，赵源，张乐伟，等 . 改进 Faster-RCNN 自然环境下识别刺梨果实 [J]. 农业工程学报，2019，35（18）：143-150.
[10] 易诗，李欣荣，吴志娟，等 . 基于红外热成像与改进YOLOV3 的夜间野兔监测方法 [J]. 农业工程学报，2019，35（19）： 223-229.
[11] 李善军，胡定一，高淑敏，等 . 基于改进 SSD 的柑橘实时分类检测 [J]. 农业工程学报，2019，35（24）：307-313.
[12] KAVETI P，SINGH H. Towards Automated Fish Detection Using Convolutional Neural Networks [C]//OCEANS - MTS/IEEE Kobe Techno-Oceans.Kobe：IEEE，2018：1-6.
[13] 郭祥云，王文胜，刘亚辉，等 . 基于深度残差网络的水下海参自动识别研究 [J]. 数学的实践与认识，2019，49（13）： 203-212.
[14] LI X，SHANG M，HAO J，et al. Accelerating fish detection and recognition by sharing CNNs with objectness learning [C]//OCEANS 2016 - Shanghai.Shanghai：IEEE，2016：1-5.
[15] 袁利毫，昝英飞，钟声华，等 . 基于 YOLOv3 的水下小目标自主识别 [J]. 海洋工程装备与技术，2018，5（S1）：118-123.
[16] 赵德安，刘晓洋，孙月平，等 . 基于机器视觉的水下河蟹识别方法 [J]. 农业机械学报，2019，50（3）：151-158.
[17] CAI K W，MIAO X Y，WANG W，et al. A modified YOLOv3 model for fish detection based on MobileNetv1 as backbone [J]. Aquacultural Engineering，2020：91：102117.
[18] ROY A G，NAVAB N，WACHINGER C. Concurrent Spatial and Channel Squeeze & Excitation in Fully Convolutional Networks [C]// Cham：Springer，2018：421-429.
[19] LIU S，QI L，QIN H F，et al. Path Aggregation Network for Instance Segmentation [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City：IEEE，2018：8759- 8768.
[20] RAHMAN Z，JOBSON D J，WOODELL G A. Multi-scale retinex for color image enhancement [C]//.Proceedings of 3rd IEEE International Conference on Image Processing.Lausanne：IEEE， 1996：1003-1006.
[21] KASHIF I，SALAM R A，AZAM O，et al. Underwater Image Enhancement Using an Integrated Colour Model [J].Iaeng International Journal of Computer Science，2007，34（2）：239-244.
[22] HUANG D M，WANG Y，SONG W，et al. Shallow-water Image Enhancement Using Relative Global Histogram Stretching Based on Adaptive Parameter Acquisition [C]//Multimedia Modeling.Cham： Springer，2018：453-465.
作者简介：马志强（1996—），男，满族，辽宁辽阳人，硕士研究生在读，研究方向：图像识别。