Fish species classification in underwater video monitoring using Convolutional Neural Networks

This report presents a case study for automatic fish species classification in underwater video monitoring of fish passes. Although the presented approach is based on the FishCam monitoring system, it can be used with any video-based monitoring system. The presented classification scheme in this study, is based on Convolutional Neural Networks that do not require the calculation of any hand-engineered image features. Instead, these networks use the raw video image as input. Additionally, this study investigates, if the classification accuracy can be increased by adding additional meta-information (date of migration and fish length) to the network. The approach is tested on a subset of 10 fish species (8099~individuals) occurring in Austrian river. On an independent test set, the presented approach achieves a classification accuracy of 93 %.

本报告呈现了一项针对水下鱼类通行监控视频的自动鱼类物种分类案例研究。尽管所提出的方法基于FishCam监控系统，但其可应用于任何基于视频的监控系统。本研究中提出的分类方案基于卷积神经网络（Convolutional Neural Networks），该网络无需计算任何手工设计的图像特征。相反，这些网络直接以原始视频图像作为输入。此外，本研究还探讨了通过向网络添加额外的元信息（如迁徙日期和鱼类长度）是否能够提高分类精度。该方法在奥地利河流中出现的10种鱼类（共计8099个个体）的子集上进行了测试。在独立的测试集上，所提出的方法实现了93%的分类准确率。