Development of labor saving and automation support technology for agricultural water management

2019 Fiscal Year Final Research Report

• Project/Area Number: 17K08001
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Rural environmental engineering/Planning
• Research Institution: Mie University
• Principal Investigator: ITO Ryoei 三重大学, 生物資源学研究科, 助教 (30232490)
• Project Period (FY): 2017-04-01 – 2020-03-31
• Keywords: 農業IoT / 水田灌漑 / 見える化 / 画像処理 / 水管理

Outline of Final Research Achievements

The meter image of the pump was image-processed to read the operating time of the pump. In FY2018, we tried to recognize numbers by deep learning and achieved a recognition rate of over 90%. As a result of creating a program that automatically corrects the numerical value that was erroneously recognized and setting the difference between the visually read value and the corrected value within ± 1 as an allowable range, the ratio of the corrected numerical value within this range is 99.99%. Thus, reasonable numbers were obtained at almost all times. In FY2019, we changed to a high-definition digital camera to obtain a high-resolution image and changed to number recognition using template matching. As a result, in the target pumping station, the optimal threshold for binarization was 60 to 65, and the recognition rate was 100%, which eliminated erroneous recognition.

Free Research Field

地域環境工学・計画学

Academic Significance and Societal Importance of the Research Achievements

予算面から高額な流量計の導入が困難なことが多い農業用水システムにおいて，高精細なカメラを用いて得られる高解像度の揚水機場ポンプメータ画像から画像処理により数字認識して積算時間を読み取ることが可能となった。画像処理に用いたテンプレートマッチングは深層学習などと比べて計算負荷が圧倒的に軽いため，Raspberry Pi等のSBC上で実行可能となった。以上より，現段階でオンサイトでのリアルタイム処理実行のための基幹技術の開発ができた。今後5Gなどの通信網が農業現場に普及するにつれて揚水機場単位での農業用水の利用実態が上水道のように見える化されることにより，農家の節水意識が向上されることが期待される。