Intelligent ship identity recognition method and system based on twin network

Abstract

The invention discloses an intelligent ship identity recognition method and system based on a twin network. The method comprises steps: constructing the twin network based on a deep convolutional neural network, manufacturing a ship name picture sample data set, and training the twin network by using the data set. The ship name information of the ship board position can be accurately matched and recognized, and the ship identity information can be intelligently recognized. According to the invention, the ship identity can be intelligently identified based on the ship video image, water supervisors are helped to better identify the real identity of the ship and acquire detailed information of the ship, and compared with other methods using artificial intelligence, such as a ship shape feature extraction method and an optical character-based identification method, the provided method is characterized in that the required requirements on definition of required training picture set are low, quantity requirements are low, the training picture set is easier to obtain, the detection accuracy is higher, the ship identification range can be conveniently expanded, and the method has the capability of being applied to actual engineering.

Description

technical field [0001] The invention relates to the technical field of ship navigation, in particular to a twin network-based intelligent ship identification method and system. Background technique [0002] With the increase of human commercial production activities, the number of ships navigating in major ports and waterways is increasing, and the supervision of water traffic mainly relies on Vessel Traffic System (VTS) and Automatic Identification System (Automatic Identification System) , AIS), especially in the identification of ships on the water, mainly rely on AIS to obtain information such as ship name, ship identification code, ship type, ship length, ship width, and ship maximum draft independently reported by the ship. However, once the ship turns off the AIS equipment, or the AIS equipment malfunctions or the information is tampered with, it will lead to the inability to effectively identify the ship's target, which will further lead to certain difficulties in th...

AI Technical Summary

Problems solved by technology

Among them, the method of using deep neural network to extract ship shape features has the problem that ships produced by the same shipyard may have similar shapes but different ship names, making it difficult to identify different ships of the same ship type; in addition, each ship is viewed from different angles, and its The extracted shape features are also different. If the extraction of all angles of the ship is not completed in the early stage, it will also cause misidentification of the ship’s identity in the actual detection and identification, which is difficult to support practical applications.

In theory, OCR can directly recognize the name of the ship, but it has high requirements for the clarity of the detected pictures, and there are many requirements for the standardization of the detected text, such as uniform background color, uniform font size, uniform Behavioral format and limited character collection, therefore, are commonly used in license plate detection, ID card recognition, invoice recognition and other fields; although ship names are often printed on the ship's side, there is no uniform printing standard, font size, background color , The number of displayed lines is different, and the characters involved in ship names in different countries are also different, including Chinese, English, numbers and even special characters, which leads to low accuracy of text recognition when OCR technology is directly used, which is difficult to support practical application

Images