Extraction of Floating Raft Aquaculture Areas from Sentinel-1 SAR Images by a Dense Residual U-Net Model with Pre-Trained Resnet34 as the Encoder

Abstract

Marine floating raft aquaculture (FRA) monitoring is significant for marine ecological environment and food security assessment. Synthetic aperture radar-based monitoring is considered to be an effective means of FRA identification because of its capability for all-weather applications. Considering the poor generalization and extraction accuracy of traditional monitoring methods, a semantic segmentation model called D-ResUnet is proposed to extract FRA areas from Sentinel-1 images. The proposed model has a U-Net-like structure but combines the pre-trained ResNet34 as the encoder and adds dense residual units into the decoder. For this model, the final layer and cropping operation of the original U-Net model are removed to eliminate the model parameters. The mean and standard deviation of Precision, Recall, Intersection over Union (IoU), and F1 score are calculated under a five-fold training strategy to evaluate the model accuracy. The test experiments indicated that the proposed model performs well with the F1 of 92.6% and IoU of 86.24% in FRA extraction tasks. In particular, the ablation experiments and application experiments proved the effectiveness of the improvement strategy and the portability of the proposed D-ResUnet model, respectively. Compared with the other three state-of-the-art semantic segmentation models, the experiments demonstrate a clear accuracy advantage of the D-ResUnet model. For the FRA extraction task, this paper presents a promising approach that has refined extraction capability, high accuracy, and acceptable model complexity.

Marine Infrastructure Detection with Satellite Data—A Review

“…The largest share of Deep Learning approaches was represented by the encoder-decoder architecture UNet. With 12 applications [31,39,50,62,[67][68][69]72,73,80,83,115], UNet is the most frequently used technology of all the articles reviewed. The U-Net model, originally developed for biomedical image segmentation [116], gained considerable attention due to its outstanding performance when published and its clear, structured design.…”

“…OWFs were investigated six times from 2020 on Concentrating on thematic foci, aquaculture was investigated in 57 publications and was therefore the most commonly studied infrastructure type in this review. Raft aquaculture was the most studied (46%) [31,32,37,39,46,50,[55][56][57][58][59][60][61][62][63][64][65][66][67][68][69][70][71][72][73] and 12% of aquaculture studies looked specifically at cages [40,52,[74][75][76][77]. The remaining 42% investigated different types of aquaculture including raft, cage or longline in combination [33][34][35][47][48][49]53,54,[78][79][80][81][82][83][84][85][86]…”

“…The zoomed-in map (Figure 8) of East and Southeast Asia clearly shows that aquaculture is being studied along the entire Chinese coast. The hotspots here are Sansha Bay in Fujian Province (16 studies) [33][34][35]37,47,49,56,65,69,73,76,80,83,85,86,88], the offshore area around Dalian in Liaoning Province (13 studies) [37,39,49,50,53,[57][58][59][60]62,67,68,73], and Haizhou Bay in Jiangsu Province (9 studies) [31,32,55,61,63,66,[71][72][73]. Further site and local studies of aquaculture were carried out on the coast of Japan [59,70,93,106], on the south coast of South Korea [81,82], the south coast of Vietnam [43,…”

“…The fields of application and infrastructures detected with radar data are diverse. Of the total of 32 reviewed articles (35%), aquaculture was identified in 11 [39,49,50,59,62,67,68,70,93,114,115], platforms in 18 [14,17,38,43,[95][96][97][98][99][100][101][104][105][106], OWFs in five [41,42,51,107,108], and bridges in 2 [44,45]. The one study that used hyperspectral satellite data used them to detect aquaculture [64].…”

Evaluating the effect of the incidence angle of ALOS-2 PALSAR-2 on detecting aquaculture facilities for sustainable use of coastal space and resources

“…Consequently, several studies have employed SAR to detect aquaculture facilities. For example, while Zhang et al (2022) proposed a method for extracting marine raft aquaculture areas using C-band SAR Sentinel-1 images by analyzing the features of marine surface areas in China, Gao et al (2022) proposed the D-ResUnet model for extracting the floating raft information of aquaculture areas from Sentinel-1 images in China. Notwithstanding, these studies primarily focused on image analysis methods for detecting aquaculture areas and not on the observation conditions or the state of individual facilities.…”