Application Of Data Fusion Methods In Extract Road Information From SAR And SPOT

With the rapid development of pace, computer and sensor technology, remote sensing data in geography fields has been widely used day by day. This paper uses the RADARSAT -1 satellite SAR image and the SPOT multispectral images, after geometric correction, radiometric calibration, noise reduction and other pretreatment, through PCA,BT,HIS,MT, HPF and WT six methods to study the image fusion ,then extract linear feature Main contents and conclusions as follows:⑴Study of Image Fusion MethodsThe results show that: based on a structural transformation image fusion HPF and WT methods ,the feature of liner clearly, and have the same color as the original multispectr- al ;The PCA based on statistical information images fusion methold , not only preserve the good texture information, but also maintain the SAR image texture information; Image based on IHS,BT and MT, although the visual effects are better, but the distortion and spectral distortion of information is serious, the entropy is lowest. But in general they are better than the original SPOT data.⑵Study of Extracting Road InformationExtract road information using spectral signature,In the field of leakaging points, PCA and HPF are better for road extraction; Followed by BT and WT; IHS's leakaging poi- nts are serious as 43%. In the field of misclassification, PCA and WT's misclassification is lower; Followed by BT and HPF; MT's misclassification are serious.accuracy of the origin- al SPOT image which extracting the road is worst either leakaging or misclassification rate。Extract Road Information Using Maximum Likelihood,based on PCA and HPF fusion methods can retain the road linear features better and extract the road completely; Followed by BT and WT;MT's misclassification and IHS's leakaging and misclassification rate are serious.But,accuracy of the original SPOT image which extracting the road is lower than six fusion metholds.In summary, Based on the six fusion image, not only preserves the original SPOT multispectral information, while maintaining SAR image texture information to improve fusion results, to improve the accuracy of linear feature extraction, and their use value.