To address the core challenge that the existing object detection algorithms were constrained by issues such as low light, low image resolution, and dense small objects in underwater environments, an improved YOLOv8 algorithm for underwater garbage object detection was proposed. The algorithm introduced self‑calibrated convolution into the feature fusion module of YOLOv8, used the method of grouping convolution for multi scale feature extraction, and expanded the receptive field of the network through down-sampling operation, to improve the multi scale feature fusion and detection ability of the model, and more accurately identify underwater garbage. The experiments were carried out on the Seaclear Marine Debris data set and TrashCan data set. Compared with YOLOv8, the detection accuracy of the improved model on the Seaclear Marine Debris data set was improved by 1.5 percent point, mAP values increased by 1.1 percent point. On the TrashCan dataset, the detection accuracy was increased by 2.4 percent point, and the mAP value was increased by 0.7 percent point. Experimental results showed that the proposed method could maintain high detection accuracy in complex underwater environment, and could meet the actual needs of underwater garbage detection.
用于目标识别与定位的深度学习算法均以卷积神经网络为核心框架[8]。例如,CHEN等[9]在URPC2017数据集上提出SWIPE‑Net网络架构,但检测精度较低。随着URPC2018中目标类别的扩展,JIANG等[10]设计的增强型SSD、HAN等[11]将平均精度均值(Mean Average Precision,mAP)分别提升至66.9%和91.2%,但检测速度较慢,无法满足多类别检测需求。ZHU等[12]提出一种基于YOLOv8的改进算法,重点强化对小型水下碎片的检测能力,该方法在TrashCan数据集[13]上实现了63.6%的精度和47.1%的mAP值。ĐURAŠ等[14]在Seaclear Marine Debris数据集上采用YOLOv6模型开展实验,取得了68.9%的mAP值,虽保障了一定的检测速度,但精度较低。
精确度(Precision,P)、召回率(Recall,R)、平均准确率(Average Precision,AP)、各类别平均精度均值(Mean Average Precision,mAP)、参数量(Params)和浮点运算数(FLOPs)是目标检测领域常用的核心评估指标,采用P、R、mAP、Params和FLOPs作为评估指标来全面评估所提方法的综合性能。P、R、AP和mAP的计算公式如下所示:
WUChunming, SUNYiqian, WANGTiaojun, et al. Underwater trash detection algorithm based on improved YOLOv5s[J]. Journal of Real-Time Image Processing, 2022,19(5): 911‑920.
[2]
AGAMUTHUP, MEHRANS B, NORKHAIRAHA, et al. Marine debris: A review of impacts and global initiatives[J]. Waste Management & Research, 2019, 37(10): 987‑1002.
HANLi, MAChunhai, LINZhihao, et al. Underwater trash detection algorithm for low-resolution small targets[J]. Science Technology and Engineering, 2024, 24(35): 15126‑15136.
[5]
SIMONH. Marine debris: Understanding, preventing and mitigating the significant adverse impacts on marine and coastal biodiversity[M]. Montreal, Canada: Secretariat of the Convention on Biological Diversity, 2016.
[6]
SHINDEA, SWATIS. Computer vision‑based autonomous underwater vehicle with robotic arm for garbage detection and cleaning[M]//CHOUDHURY T, KOLEY B, NAT A, et al. Geo‑Environmental Hazards using AI‑enabled Geospatial Techniques and Earth Observation Systems. Advances in Geographic Information Science, Cham: Springer, 2024: 265‑288.
[7]
ZHANGYanghai, HUANGZan, CHENChanglin, et al. A spiral‑propulsion amphibious intelligent robot for land garbage cleaning and sea garbage cleaning[J]. Journal of Marine Science and Engineering, 2023, 11(8): 1482.
[8]
KONGShihan, TIANManjun, QIUChanglin, et al. IWSCR: An intelligent water surface cleaner robot for collecting floating garbage[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2020, 10: 6358‑6368.
LIGuojin, YAODongyi, AIJiaoyan, et al. Detection and localization of floating objects via improved Faster R‑CNN[J]. Journal of Xinyang Normal University(Natural Science Edition), 2021, 34(2): 292‑299.
[11]
CHENLong, LIUZhihua, TONGLei, et al. Underwater object detection using Invert Multi‑Class Adaboost with deep learning[C]//2020 International Joint Conference on Neural Networks (IJCNN), Glasgow, UK, 2020: 1‑8.
[12]
JIANGZhongyun, WANGRongrong. Underwater object detection based on improved single shot multibox detector[C]//Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence, New York, NY, USA, 2020 : 1‑7.
[13]
HANFenglei, YAOJingzheng, ZHUHaitao, et al. Marine organism detection and classification from underwater vision based on the deep CNN method[J]. Mathematical Problems in Engineering, 2020, 2020: 3937580.
[14]
ZHUJin, HUTao, ZHENGLinhan, et al. YOLOv8‑C2f‑Faster‑EMA: An improved underwater trash detection model based on YOLOv8[J]. Sensors, 2024, 24(8): 2483.
[15]
HONGJ, FULTONM, SATTARJ. TrashCan: A semantically‑segmented dataset towards visual detection of Marine debris[EB/OL].(2020‑07‑16)[2025‑01-04].
[16]
ĐURAŠA, WOLFB J, ILIOUDIA, et al.A dataset for detection and segmentation of underwater Marine debris in shallow waters[J]. Scientific Data, 2024, 11(1): 921.
[17]
LIUJiangjiang, HOUQibin, CHENGMingming, et al. Improving convolutional networks with self-calibrated convolutions[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 2020: 10096‑10105.