Maritime radar image processing method based on deep learning

Abstract

The deep learning-based maritime radar image processing method disclosed by the present invention is a combination of image processing algorithms and convolutional neural networks to identify ship targets in maritime radar images, specifically: first using grayscale, median filtering, and connected domains Marking and double-threshold connected domain processing algorithm preprocess the maritime radar image, and then use the inter-frame addition and skeleton thinning algorithm to obtain the initial motion vector of the ship target, and obtain the motion vector of the ship target in different frames by the frame difference method, according to These two motion vectors get the optimal number of frames of the frame difference method; then use the frame difference and the shape, contour and motion features of the ship to extract the ship target and non-ship target; finally use the convolutional neural network to analyze the ship target and non-ship The target is trained and tested, and the trained convolutional neural network is used for ship target recognition in maritime radar images. The invention makes full use of the motion characteristics of the ship to identify the ship target more conveniently and better than the traditional method.

Description

technical field [0001] The invention relates to the fields of maritime radar image processing and ship target detection, in particular to a deep learning-based maritime radar image processing method. Background technique [0002] With the rapid development of social economy and science and technology, water transport, as a low-cost, large-capacity transport mode, has an absolute advantage in cargo transportation. The rise of water transport has led to more and more busy water traffic. The real-time monitoring and scheduling of ships underway is very important. [0003] The traditional frame difference methods mainly include the background difference method and the three-frame difference method. The background difference method requires the detection point to be fixed, because the background difference method can be used to remove the background only when the detection point is fixed, because the ship moves relatively slowly relative to the vehicle , and the three-frame diff...

AI Technical Summary

Problems solved by technology

[0003] The traditional frame difference methods mainly include the background difference method and the three-frame difference method. The background difference method requires the detection point to be fixed, because the background difference method can be used to remove the background only when the detection point is fixed, because the ship moves relatively slowly relative to the vehicle , and the three-frame difference method is prone to errors for slow-moving targets, so the three-frame difference method cannot be directly applied to enhance the motion characteristics of ships. Due to the particularity of maritime radar images, ships move slowly, and an adaptive frame-by-frame processing is required Algorithms are used to enhance ship motion features

Images