Fish Aggregation Devices （FADs）are simply man-made, floating devices, which make use of the natural habit by aggregating pelagic fishes for subsistence, recreational and commercial fishing. The introduction of anchored FADs （aFADs）has mainly been promoted throughout the world to assist smallscale fisheries. This paper details information of East Asia and Indian Ocean regions where aFADs are heavily, traditionary used or successfully implemented. We focused on the aFAD design, fishing gear used, target species and others. In addition, we highlighted the issues and management measures related to aFADs. The main aim was to give supplementary information since aFADs are an important tool for promoting, managing artisanal and small-scale commercial fisheries all over the world, increasing localized catches at reduced costs thereby improving food security and livelihoods of the coastal communities.

View full abstract

Identifying pufferfish species for human consumption is conducted by experts who hold a license to cook pufferfishes. Nevertheless, it is difficult to identify the parental species and the poisoned parts of interspecific hybrid pufferfishes. Therefore, putative hybrids are completely excluded from the distribution process. Developing a system to identify hybrid pufferfishes will decrease the erroneous identification of pufferfish species. In this study, to apply to such identification system, pufferfish skin patterns were replicated using a cellular automata （CA）model. Here the CA model was based on Turing patterns through the exchange of binary values between neighboring cells. Despite the simplicity of the model, which uses five parameters （three parameters related to basic color pattern and two parameters for creating a large black spot）to produce skin patterns, it can produce characteristic skin patterns of all edible species of Takifugu.

View full abstract

[in Japanese]

View full abstract

Models which ensure accurate non-destructive estimates of the freshness of fish meat （K value）in real time are proposed to improve quality control and maintain the skill level of distributers of dressed puffers （Migaki）. Seven kinds of dressed puffers were used to construct the models for estimating fish meat freshness. Relationships between fish coloration and K values from sample acquisition until 72 hours later under refrigeration at −2℃, ＋2℃, and ＋6℃were investigated. The statistical analysis revealed that fish coloration does reflect its K value, although the strength of the relationship differs according to fish species. Two models were designed on the basis of these results, and the usefulness of each model was evaluated. The models are as follows: （1）Model to infer the K value of fish meat based on the coloration of the fish body surface by using fuzzy inference （Model 1）, and （2）Model to estimate the K value of fish meat after several hours for the same fish for which the K value was inferred with Model 1 （Model 2）. For these models, a high estimation accuracy was confirmed, demonstrating their potential usefulness for quality control in the distribution of dressed puffers.

View full abstract

This research aims to efficiently create detailed bathymetric charts. Our approach is to obtain fine seafloor details from coarse depth measurements only, making full use of existing data and minimizing new observation. To this end, treating gridded bathymetric data as digital images, we propose to apply super resolution, which is a technique to enhance image resolution, to bathymetry. Specifically, we employ learning-based super resolution to automatically extract characteristic features of bathymetric images. In experiments, we prepared pairs of low and high-resolution images, and let a deep neural network learn their relationship and estimate a high-resolution image from each low-resolution one. Then, we evaluated results in terms of numerical error and visual quality, and confirmed that the proposed method can recover detailed seafloor structures more plausibly than naive interpolation.

View full abstract

Most of marine accidents are caused by human errors, and hence the prevention of marine accidents is essentially difficult as long as ships are operated by human beings. In addition, further congestion of seaborne routes and harbors is apprehended, and also it will be seriously difficult to secure required number of seafarers in near future in Japan. Therefore automation of ship navigation is an urgent issue in marine transportation. As the automatic collision avoidance is a key technology for realizing autonomous ships, various algorithms and approaches related to hazard detection and maneuver for collision avoidance have been studied nowadays. In this report, we try to demonstrate the potential of automatic ship handling by AI （Artificial Intelligence）through comparisons with that by a conventional geometric model as well as a dedicated experiment using three self-propelled ship models.

View full abstract

Fishery prediction using satellite data has been reported by many studies. However, many of these studies use catches data sets from fishing boats and research vessels. In this study, using night-time visible images from satellite data instead of catches information, we predicted potential fishing grounds for saury using random forest, support vector machine, maximum entropy of machine learning. From September to December, the fishing ground predicted by machine learning showed moving from the north to the south as in the past catches reports. The predicted fishing ground distribution pattern was consistent with past reports. Some ships like as fleet of fishing boats were also located in the predicted fishing ground outside EEZ. The fishing ground zone predicted by using random forest showed the most reasonable in the three machine learning models. We suggest that it is possible to predict the potential fishing grounds only from satellite data sets.

View full abstract

Many of the coastal embankments have been maintained for over 50 years, and inspections for deterioration diagnosis are becoming increasingly important. In addition, for the inspection of coastal embankments is conducted by visual inspection, the inspecrtion efficiency is poor, and oversights are unavoidable, and survey results show artificial variation. Therefore, improvement of the inspection method is required. 　　A public-private research platform was constructed to investigate the method of cost-saving or laborsaving of stock management of public facilities such as coastal embankments. We, a public-private research platform developed a degradation diagnostic system using a Unmanned Aerial Vehicle （UAV） -mounted digital camera to inspect the damage of degraded coastal embankments. The information necessary for the maintenance and management of the facility is automatically extracted using artificial intelligence （AI）, using the visualization information from the aerial image by the UAV.

View full abstract

Amitori Bay, located in northwestern Iriomote Island, is characterized by its varied physical environments such as geographical features, wave height, and current in spite of its small size. It also exhibits a diverse distribution of coral reefs in response. Physical data acquired through numerical analysis, although including errors from actual measurements, provide much information by being interpolated spatially, and are useful for the understanding of phenomena. However, the spatial distribution of corals is difficult to estimate using an ecosystem model because the coral ecology has numerous and important unknown characteristics. We proposed to use AI to estimate the horizontal distribution of coral cover based on the association between coral cover and physical data.

View full abstract