SAR Building Detection Research And Application

With the development of city,the need of acquiring city environment information is increasing.Buildings are important part of the city environment,so judging whether there are buildings and acquiring buildings' position are very important.Therefore building detection of SAR image becomes an important research subject of city environment information interpretation.In this paper,through analyzing the reflection of radar waves,the characters of building in SAR image are summarized and two methods of building detection are proposed.The first method is based on segmentation and edge outline threshold.Because of overlapping feature of radar wave,the building in SAR image is of high gray value,as well as,the exist of shadow makes it easy to separate building from background.So making use of MRF model whose feature field weight is variable,the building object can be segregated from SAR image.Then the building object's edge is detected by multi-scale and multi-direction wavelet transform.Through tracing the outline of object's edge,the outline's pixel number is known.If the number of outline pixel is beyond the standard threshold of building,the object is added into detection result.The method which judges building by gray value and outline pixel number can avoid the influence of SAR image speckle.The second method is based on segmentation and radon transform.The building in SAR image has rectangle and L shape,so radon transform is adopted to get direction and angle information of building skeleton.The object which has same direction or has right angle is judged as building object and added into detection result.The second method which makes use of gray value and geometrical information can improve accuracy of building detection.To prove the two methods are effective,it is applied into actual SAR image.The result shows that the method based on segmentation and edge outline threshold can delete small plaque object,as well as,it has fast operation speed.The method based on segmentation and radon transform can judge whether the object is building or not by direction and angle information.And the angle condition can be changed with practical SAR image,which improves the accuracy of building detection and makes it become adaptable to different imaging situation.