Design of a multi-view visual feature extraction system for ship imagery using deep learning
Feature extraction and image recognition schemes based on visual image processing are effective comple-ments to radar and AIS,effectively identifying other ship targets,reefs,and so on,which can effectively ensure the safety of ship navigation.This paper proposes a visual image feature deep extraction system based on CNN and Laplacian pyramid image fusion,designs the basic structure of the system,and analyzes the basic functions of each module within the system.A feature deep extraction scheme based on CNN is proposed,and features of ship and buoy targets are extracted.The Lapla-cian pyramid fusion is used to deeply integrate images obtained from different ships.The multi-directional visual image feature deep extraction system established in this paper can effectively acquire various target features and can significantly enhance the richness and accuracy of feature representation.
feature deep extractionCNNvisual imagingdeep fusion
NETL
NSTL
万方数据
