Research On Cloud Detection Using Support Vector Machine(SVM)and Multi-Features Fusion

The remote-sensing image is widely used in the resource investigation,agriculture production,land cover classification,environment monitoring,ecological protection,target recognition and so on,Due to the influence of the weather,the remote sensing images tend to have some cloud covering areas,but the presence of clouds hinders imagery interpretation work and impedes subsequent use of images.Detecting the cloud in the images accurately is important to the further procession and application of the image.For the present situation of traditional cloud detection's lower accuracy and lower feasibility,this paper plans to carry out cloud detection based on multi-source remote images and multi-region images.By analyzing the ability of different features in identifying clouds,this paper proposes a cloud detection method using support vector machine(Support Vector Machine,SVM)and multi-features fusion.Firstly,the spectral feature of the cloud samples was acquired on the radiation characteristics of cloud in the different bands.Then the texture features were extracted by Sum and difference histogram of the sample blocks.The normalized difference vegetation index(NDVI)was selected as a supplementary feature.Using the spectral features,texture features and NDVI feature of image to construct feature vector and use SVM to detect cloud,combining with improved average gradient and geometry feature to reduce the false positive in preliminary cloud detection.The method can be used in the Landsat image and Spot image to get a high precision of cloud detection,suggesting SVM and multi-features fusion cloud detection exhibits good scalability.The result shows that this method has good wide applicability and high precision,and can meet the requirements geomatics industry.