RIASSUNTO
A remote fish finder system for effective set-net fishery, which enable to visualize as well as recognize catch amount and the fish kind, is proposed in this paper. In order to improve fishing efficiency and obtain the highest possible profit of set-net fishery, set-net monitoring is needed by coastal fishermen. Set-net monitoring can be used to estimate fish catch amount within the set-net, fish kind trapped or to estimate the adequate timing to lift the set-net. Set-net monitoring has been practiced for several decades using conventional equipment. However, the equipment has many weaknesses, namely, high priced, large sized, troublesome maintenance and poor data record (paper based data record). This study aimed to develop a remote fish finder system for set-net monitoring to meet those problems. The development of remote fish finder system is based on a floating echo sounder for data acquisition, a cloud server for data storage, and an iPad application for data display. The experiment for this study has been conducted from June 28 until December 25, 2013 in two experimental sites in Hakodate, Hokkaido, Japan. The algorithm development for estimations in this study was based on empirical estimation of fishermen when they observed the printed-out sounder reflection data. The empirical estimation was practiced by observing the intensity of signs of fish in certain depth. Statistics of reflection data in several layers depth during sunrise were used as indicators to estimate. Catch amount within set-net could be estimated through multiple regression analysis, while fish kind classification was through linear discriminant analysis. The results of catch estimation via multiple regression analysis has been obtained with multiple R2 = 0.89. Yet, single dominant fish kind could be estimated better than by catch using this algorithm. On the other hand, fish kind within set-net could be classified with 83% of correctness. In order to improve the accuracy, the data sampling must be continued at least a whole fishing season.