Challenge to species identification with high-resolution echosounders -analysis of fish-school echoes with machine learning-

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K21329
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Review Section: Medium-sized Section 40:Forestry and forest products science, applied aquatic science, and related fields
• Research Institution: Tokyo University of Marine Science and Technology
• Principal Investigator: Amakasu Kazuo 東京海洋大学, 学術研究院, 教授 (80452043)
• Project Period (FY): 2020-07-30 – 2023-03-31
• Keywords: 魚種判別 / 魚群探知機 / スペクトル

Outline of Final Research Achievements

In this study, the internal structures of fish-schools were observed with a high-resolution echosounder. Amplitude spectra, echo-intensity spectra, variation spectra of echo intensity, and morphology of fish-schools were extracted as the feature quantity of fish-schools and were confirmed that some of them were useful information for species identification. The window sizes of fast Fourie transform to obtain accurate echo-intensity spectra were clarified and the echo processing method was improved. The internal structures of fish-schools could be observed, but it was also clarified that such observations were difficult for deeper fish-schools and higher densities. Although clustering of fish-school echoes was tried with an unsupervised machine learning, it was difficult to determine appropriate number of clusters.

Free Research Field

水産音響学

Academic Significance and Societal Importance of the Research Achievements

高分解能魚探は魚種判別に道を開くと期待されてきたが，特徴量の抽出についてはエコーデータの処理方法も含めて未だ十分に研究がなされていない。その点，本研究は複数の特徴量を抽出し，魚種判別に有用な情報となることを確かめた。また，エコー処理方法の高度化もしている。魚種との対応付けが不十分であったが，魚種判別問題の解決に資する研究成果であり学術的意義がある。
魚種判別手法は，魚群探知機を使用した効率的な漁獲および正確な資源量推定に必要である。魚探による魚種判別技術の確立は「水産資源の保全（管理）と持続的な利用」すなわち食料の安定供給に寄与するものであり，ここに本研究成果の社会的意義がある。