Romote Sensing Image Airport Target Segmentation Based On Co-Saliency Detection

The information of target is complex in remote sensing images,and how to identify them quickly and accurately has always been a challenging topic of remote sensing image analysis.The current remote sensing image detection methods and public data sets are primarily aimed at small targets,such as ships,automobiles,and airplanes.Few research for segmentation of airport target in remote sensing images.However,airport target detection has always played an important role in the military field.Some single segmentation algorithms are difficult to segment airport targets accurately,and the precision of segmentation needs to be further improved.Hence,it is worth researching the improvement and innovation of remote sensing image airport target segmentation algorithms in-depth.This paper researches the segmentation algorithm of airport objects in remote sensing images.Combined with the structure and runway characteristics of the airport,we propose the Parallel Line Segmentation Detection(PLSD)method,the co-saliency detection model and the contour extraction model.Through experiments,the segmentation results of this paper are compared with other methods from the subjective and objective aspects,and proves the effectiveness of this method.The main research content of this paper includes three aspects:(1)The remote sensing image used for the task of segmentation of airport target contains more interference information.The existing straight line segmentation method can detect all the straight line information in the remote sensing image,but it is not suitable for the problem of airport runway segmentation.For this problem,we combine the characteristics of the runway to improve it,and propose the Parallel Line Segmentation Detection method.(2)Existing remote sensing image airport target segmentation methods usually use runway features to locate the airport,and then use Support Vector Machines(SVM)for classification,finally get the airport target.But the airport target extracted by this method has the phenomenon of blurred edges.Responding to this phenomenon,We propose a contour extraction model,which improves the problem of edge blur during airport target segmentation in remote sensing images.(3)The saliency detection of a single image ignores the consistency between similar images.For the airport targets of remote sensing images,all airports are principally composed of thin,long,regular runways and buildings.It is different from the size and structure of the airport in each region,but they have similarities.According to the above-mentioned characteristics of the airport,this paper proposes a remote sensing image airport target segmentation method based on co-saliency detection.The images in this article data set are all from Google Earth.Through precise positioning,325 remote sensing images of different airports and 30 interference images are collected,we test the proposed method on this data set.In this paper,three evaluation indicators of MAE,DR,and FAR are used to evaluate the experimental results.The experimental results prove that the method proposed in this paper is ahead of the existing methods,easy to implement and understand,and has great practical significance.