Convolutional Neural Network-based Fish Recognition Algorithm on Intelligent Vehicles

The species recognition algorithms have been paid more attentions recently. In this paper a convolutional neural network-based fish recognition algorithm was proposed on intelligent vehicles. The algorithm determines the designated vehicle position with the fixed-point position recognition function, and removes it to the corresponding position through the intelligent control system. Finally, to meet the computational requirements and accuracy in the embedded situation, a lightweight convolutional neural network was designed to complete fish identification. This method was trained on a Large Scale Fish Dataset (LSFD), and the accuracy rate reached above 97.2%. This algorithm has certain research significance in intelligent vehicles and fish recognition area.