Exploring deep learning techniques for the extraction of lit fishing vessels from Luojia1-01
文献类型： 外文期刊
作者： Hu, Huijuan 1 ; Zhou, Weifeng 1 ; Jiang, Bohui 2 ; Zhang, Jiaze 2 ; Cheng, Tianfei 1 ;
作者机构： 1.Chinese Acad Fishery Sci, East China Sea Fishery Res Inst, Shanghai 200090, Peoples R China
2.Shanghai Ocean Univ, Coll Marine Sci, Shanghai 201306, Peoples R China
关键词： Identification of lit fishing vessel; Deep learning; YOLO-V5; Object detection
期刊名称：ECOLOGICAL INDICATORS （ 影响因子：6.9； 五年影响因子：6.6 ）
ISSN： 1470-160X
年卷期： 2024 年 159 卷
页码：
收录情况： SCI
摘要： Marine fishery resources are linked to global food security and the livelihoods of millions of people, but their sustainability is seriously affected by the overexpansion of fishing activities. The regulation of fishing activities of fishing vessels has been based on ineffective data sources. Fortunately, rapid developments in nighttime light remote sensing have provided effective data for the formulation of marine fisheries regulations. We conducted a series of experiments to explore the application of Luojia1-01 (LJ-01) nighttime light images for the extraction of information on fishing vessels to provide effective data support for the formulation of marine fisheries regulations. First, a series of preprocessing operations, including radiation correction and masking of the offshore oil and gas platforms were performed on the original image. The preprocessed single -band nighttime light image was combined with two stretching methods to form a triband image. Sample labelling of lit fishing boats was performed. Then, by comparing YOLOv5 metrics, YOLO-V5s was chosen as the base framework of the model, and a small-target detection layer was added to improve the detection of lit fishing vessels. The method developed in this study accurately and effectively detected lit fishing boats in LJ1-01 images with strong robustness and generalizability, and the accuracy of detection rate reached 96.6 %. The precision was improved by 10 % compared to the original YOLO-V5s model that used raw data. The recall rate was improved from 85.4 to 93 %, and the average accuracy(map@0.5) was improved from 85.4 to 93.1 %. The developed model can be successfully used for the extraction of lit fishing boats and for other applications based on medium- and highresolution nighttime satellite data.
- 相关文献
作者其他论文 更多>>
-
Screening and Analysis of Potential Aquaculture Spaces for Larimichthys crocea in China's Surrounding Waters Based on Environmental Temperature Suitability
作者：Yang, Ling;Zhou, Weifeng;Cui, Xuesen;Lu, Yanan;Liu, Qin;Yang, Ling
关键词：
Larimichthys crocea ; deep-sea aquaculture; potential spaces; spatial analysis; China -
Effects of western boundary currents and sea surface temperature anomalies on interannual variability of chub mackerel abundance in the Northwest Pacific
作者：Li, Jiasheng;Zhou, Weifeng;Dai, Yang;Tang, Fenghua;Wu, Yumei;Zhang, Heng;Fan, Xiumei;Cui, Xuesen;Li, Jiasheng
关键词：Chub mackerel; Abundance; Kuroshio; Oyashio; Sea surface temperature anomaly; Northwest Pacific
-
Unsupervised Classification of Global Temperature Profiles Based on Gaussian Mixture Models
作者：Ye, Xiaotian;Zhou, Weifeng;Ye, Xiaotian
关键词：ocean temperature; Gaussian Mixture Models; the optimal model; global distribution
-
A Data Cleaning Method for the Identification of Outliers in Fishing Vessel Trajectories Based on a Geocoding Algorithm
作者：Zhang, Li;Zhou, Weifeng;Zhang, Li
关键词：Geohash; fishing vessel; trajectory data; outliers; data cleaning; data mining
-
Multiscale variation analysis of sea surface temperature in the fishing grounds of pelagic fisheries
作者：Lai, Qixiang;Zhou, Weifeng;Lai, Qixiang
关键词：pelagic fisheries; sea surface temperature; Ocean Nino Index; trend decomposition; variance analysis; the interquartile ranges; change point analysis
-
Fishing operation type recognition based on multi-branch convolutional neural network using trajectory data
作者：Jiang, Bohui;Zhou, Weifeng;Jiang, Bohui
关键词：Geohash; Vessel trajectory; Deep convolutional neural network; Deep learning; Spatiotemporal context; Fishing operation type; Embedding
-
Identification of glass eel capture equipment in the Yangtze River estuary based on high-spatial-resolution imagery and an improved YOLOv8 model
作者：Zhu, Pengfei;Zhou, Weifeng;Zhu, Pengfei
关键词：YOLOv8; Small target detection; Glass eel; Deep learning
