Road Detection Of Sar Image Based On Muti-window And Connected Domain Analysis

As an artificial land,roads have important geographical,political and economic significance.The data of road information is an important part of geographic information data,and is also the basic data of GIS application.Synthetic aperture radar(SAR)is a high resolution coherent imaging radar,which can use the antenna to transmit the pulse signal to the target and the electromagnetic wave energy returned by the target to form the image.Compared with the traditional visible and infrared remote sensing,SAR has some imaging features,such as all-weather,all-weather and penetration.Therefore,SAR image road detection is of great research and application value.At present,road detection in SAR images is mainly divided into three steps: edge detection,line primitive's extraction and line element connection.Although many domestic and foreign scholars have done a lot of research on road detection in SAR images,there are still two problems:(1)In the edge detection stage,the speckle noise is more serious for road extraction from SAR images.(2)The extraction algorithm of line primitives is highly dependent on the direction information and the calculation is time-consuming.Therefore,this paper mainly studies and improves the related algorithm of SAR image road detection from two aspects: edge detection and line primitive's extraction.The main contributions are as follows:(1)In view of the serious interference of speckle noise on road extraction from SAR images,a new SAR image multi window road edge detection algorithm is proposed in this paper.The algorithm uses the existing SAR image edge detection operator to detect the image and detect the results in weighted fusion.The weighted local entropy is used as the basis for evaluating the road points.The fusion weight is determined by normalization of the inverse cotangent function transformation form.Experiments show that the algorithm can improve the integrity of road edge detection.(2)In order to solve the problem of high dependence on direction information and time consuming,a new linear base element extraction algorithm based on connected domain is proposed in this paper.The algorithm detects segmentation points for every connected domain and segmenting the connected domain by segmentation points.A linear fitting is done for every part of the connected domain and the initial element is connected locally to get the line primitives.Experiments show that the algorithm can extract line elements accurately and efficiently without relying on the information of edge direction.