Behavior based classification of aquatic invasive fish species in underwater video
Major Goals and Objectives
Aquatic invasive species pose a serious threat to aquatic ecosystems. In the Great Lakes, new fish species that are non-native to the environment have been introduced due to transoceanic activities. These species compete with native species causing a heavy toll on the environment and economy by shifting ecological balances and disrupting food chains. While many methods exist to sample fish, due to the increasing availability of underwater video, fish can now be sampled by crowd sourcing or using Image recognition techniques. These methods have highlighted the challenges associated with identifying fish based on their appearance against a cluttered background. This graduate student scholars project aims to identify and model fish behavior in underwater video using machine learning methods that are suited for classification of classification of time series data. Video data of round goby (Neogobius melanostomus) from literature will be used to test the approach. The methods proposed here will significantly increase the usability of existing datasets and enable the creation of life-like animations for use in virtual training environments.