| Coastal zone monitoring has practical significance for land and resources exploration,sea use management and other fields.While remote sensing provides significant means for coastal zone monitoring.Among them,optical remote sensing and Synthetic Aperture Radar(SAR)images can have a wide detection range at medium or low resolution.On this premise,this paper focuses on two problems of coastal zone,namely coastline detection and coastal zone classification.In terms of coastline detection,SAR images have the ability of all-weather detection,however,due to the coherent speckle noise,ocean waves caused by sea wind,or other complexity factors,the coastline detection research is still facing great difficulities.In terms of coastal zone classification,there is still much room for improvement in the interpretation performance of optical or SAR images alone.Thus,this paper studies the coastline detection algorithm based on SAR images,and the fiusion based classification algorithm with SAR images and optical images in coastal zones from the perspective of the image fusion.The main work of this paper is as follows:(1)A modified SAR images coastline detection algorithm based on superpixel with line detection is proposed.This algorithm solves the coastline detection problem from the perspective of classification.Since existing superpixel algorithms misfitting the lines,this paper presents a Improved line finder(ILF)for SAR image to extract straight lines,which introduces a bilateral filter into the existing Fast line finder(FLF)algorithm.Then a modified superpixel generation algorithm with line detection is given,in which the parameters of modified superpixel algorithm is controlled by ILF line map.Finallly,hidden markov model is used to classify the obtained superpixels.In our experiments,SAR images under relatively complex sea condiitons were detected,the results show that the mean offset index of our coastline detection algorithm can be controlled within two pixels,which is smaller than that of comparison algorithms.In short,our algorithm can solve the problem of coastline detection in SAR images with complex area.(2)A modified fusion based coastal zone classification algorithm for remote sensing image with the improved reliability factor,measured by homogeneous degree,is given.The motivation lies in the fact that the existing fusion based classification algorithms measure sensor data with uncertainty,in which the advantages in interpreting different ground objects of the two different kinds of sensors are not fully utilized.As a result,the classification accuracy is only about 70%,leaving much more room for improvement.To solve this problem,the gray level co-occurrence matrix texture of SAR image and the intensity value of the optical image is combined together to extract artificial building areas,then different reliability factor is defined in different regions.Finally,the fusion classification of SAR and optical image in coastal zone is relized by markov random field.In the experiments,Sentinel-1 and Landsat-8 images are used to evaluate our algorithm.Results show that the improved fusion based classification algorithm has better classification performance than those of the existing fusion based classification and the single image based classification methods.Specifically,the overall accuracy of our method can reach 87.86%,which is about 20%higher than the existing comparison algorithms of fusion based classificaiton.Thus,the problem of insufficient utilization of sensor data by existing algorithms is solved.In other words,the classification performance of coastal zone is obviously improved in our algorithm. |
PDF Full Text Request