Marine organism identification system and identification method based on deep neural network

Abstract

The invention provides a marine organism recognition system based on a deep neural network, and the system comprises a preprocessing module which collects an original image candidate region containinga target object in an original image of a marine organism, and inputs the original image candidate region into the deep neural network DNN; a deep neural network DNN; and an image synthesis module used for carrying out synthesis processing on the species classification, the species boundary frame and mask information and the preprocessed marine organism image to obtain a marine organism entity segmentation image. According to the method, original image information is collected and marked, then calculation processing is carried out in a deep neural network DNN, then marine organisms are recognized, segmented and tracked, and finally the marine organisms are combined with an original image to form visual picture information. In addition, the invention further provides an identification method based on the identification system, and the method effectively solves the technical problem that in the prior art, sea low species cannot be accurately identified, segmented and tracked.

Description

technical field [0001] The invention relates to a marine biological identification system and identification method, in particular to a marine biological identification system and identification method based on a deep neural network. Background technique [0002] A neural network is an algorithmic mathematical model that uses distribution and processes information, and is usually used for artificial intelligence recognition. Deep learning neural networks generally use sample data for deep learning. Traditional technology centers usually use sample data to input a fixed number of layers of deep learning neural network for unsupervised learning, and use labeled sample data to input fixed number of layers of deep learning for supervised learning to obtain output labels to complete the deep learning. Learn how to train a neural network and test it with test data after training. With the development of computer vision, the application of deep neural network in various fields is...

AI Technical Summary

Problems solved by technology

[0003] At present, in terms of marine biological identification, some marine biological identification systems cannot accurately identify, segment and track smaller underwater species, cannot display the ecological details of underwater species, and canno

Images