An Improved Faster RCNN Marine Fish Classification Identification Algorithm

Abstract

For the efficient identification of marine fish species, a fast neural network recognition algorithm based on improved Faster RCNN is proposed.The algorithm first selects residual network (Resnet) with strong feature extraction capability for feature extraction; then generates candidate target regions through 12 different Anchors to further improve the accuracy of detection; finally, the resulting features are transmitted to two subnetworks to achieve classification and positioning respectively.The classification networks are based on the full connectivity structure, while the localization network is mainly composed of convolutional neural networks.This paper verifies the effectiveness of the algorithm on the marine fish (holothurian, echinus, scallop, starfish) image dataset. The results show that the proposed algorithm is more accurate recognition than Faster RCNN while efficiently detecting the target.