Fish classification has significant importance in the field of biodiversity and study of aquaculture. The aim of this project is to use a Voice Controlled Remotely Operated Vehicle (ROV) for underwater exploration. The ROV will use a camera to capture images of fish at user defined time interval and transmit them to the remote server via wireless connection. The remote server will classify the fish species into one of five categories of fish using Convolutional Neural Networks (CNN). CNN is employed as it is data driven, convenient to use and accurate in terms of its strength in automatically identifying the features from a diverse data set. One CNN model is used and trained with images from the ground truth dataset consisting of 24971 fish images belonging to 5 fish species which were collected from a video. The preprocessed images of the fish were used to adjust the trained deep neural network in order to identify the network architecture resulting in the highest accuracy of classification performance. The results illustrate that the system performed with high accuracy, reaching test accuracy of 98.122% and this approach is able to classify fish species effectively. The motivation of this project is to provide economical underwater services such as environmental survey & observation and oceanographic research that can be conducted directly from the ships at sea. This ROV can be employed in aquaculture to provide a cost effective solution for fish farmers to assure efficient harvest, healthy fish crop and environmental protection.
Project Objectives
- Use a remote or voice controlled boat for underwater exploration.
- Use a camera to capture fish images and transmit them to the user’s laptop via wireless connection.
- Fish images will be classified using Convolutional Neural Networks.
Presentation
