Perceptual Fusion Of Electronic Navigation Image And Navigation Radar Image Based On WGAN

Modern ships put forward higher requirements for the accuracy and reliability of the navigation system.As an important navigation instrument for ships,the ship navigation radar can detect targets above the water surface and is the main source of information for ships to avoid collisions.However,radar images cannot provide specific hydrographic information of the navigation waters;Electronic Navigation Image(ENC)can display coastline,land,and navigation signs,providing accurate marine environmental information for crew members.It is impossible to ensure the safe navigation of a ship only by relying on one kind of equipment.The integration of electronic navigation images and navigation radar images can help ship drivers better understand the marine environment they are in,so as to better understand the ship’s navigation status and improve the ship’s navigation safety.The navigation radar is prone to be affected by weather and sea conditions in the actual navigation process of the ship,resulting in different levels of noise spots in the obtained navigation radar image.It is not only difficult to accurately distinguish the image information,but also has certain risks and hidden dangers.This article proposes a generative adversarial network(GAN)denoising method based on Wasserstein distance and perceptual similarity.Utilizing Wasserstein distance as an evaluation parameter between distribution and perceptual loss,adjust the network structure in a timely manner to improve the training ability of the denoising network.Input the original navigation radar image and the navigation radar image with noise into the convolutional neural network generator at the same time,use the depth convolution network to extract image features and calculate the perception loss,judge the similarity between the generated image and the original radar image by network analysis,so as to obtain a clearer and more accurate navigation radar image.Input the denoised navigation radar image into the local adaptive Canny operator,design local high threshold and low threshold target edge detection,and realize the edge detection of navigation radar image;Then the scale invariant feature transform(SIFT)method is used to complete the registration of navigation radar image and electronic navigation image image;High frequency and low frequency subband coefficients of navigation radar image are obtained by parallel processing of fast Fourier transform based on sparse theory.High frequency fusion rules are used to fuse high frequency subband coefficients containing important information of navigation radar image.Low frequency subband coefficients of approximate image of navigation radar image are trained to obtain dictionary.Fusion image of electronic navigation image and navigation radar image is obtained through inverse Fourier transform,Finally,the quality of the fused image is evaluated using the corner overlap algorithm.The research results show that the data fusion of electronic navigation image and navigation radar image has high consistency,and the algorithm has good performance of fast training speed,short registration time and good fusion effect.