An object-based SVM method incorporating optimal segmentation scale estimation using Bhattacharyya Distance for mapping salt cedar (Tamarisk spp.) with QuickBird imagery

Salt cedar, a predominant arid type of vegetation in Western China, provides invaluable ecosystem services and is important for economic sustainability. Most previous attempts to map salt cedar have been focused at the pixel level and not at the object level. A particular problem for object-based classification is determining the optimal scale for the segmentation. In this study, we proposed a new object-based image analysis method for classifying high resolution satellite image with support vector machine (SVM). Specifically, we set forth three objectives: (1) to choose the optimal multiscale parameters for different cover types with the aid of the Bhattacharyya Distance index; (2) to extract the class specific features for different classes to feed into the SVM classification procedure; and (3) to compare three different classification methods: the integration of object-SVM, SVM at the pixel level and nearest-neighbor at the object level. The result of the case study demonstrated that the multiscale object-SVM method, which employed spectral, texture and shadow features, produced the best overall accuracy (91.6%).