High-resolution multispectral image classification over urban areas by image segmentation and extended morphological profile

In this study, classification of multispectral data with high resolution from urban areas by combining image segmentation and morphological characteristics is investigated. Traditional watershed segmentation defined for gray level image was extended to multispectral image segmentation by computing multispectral gradient image through a vector based approach, which uses extended dilation and erosion operations. The extended morphological profile was used to extract multiscale structural information from multispectral image, which was then used in image classification. An extended morphological profile is constructed based on the repeated use of geodesic openings and closings with a structuring element of increasing size, starting with the original multispectral image. Since the profile includes a range of increasing opening and closing by reconstruction operation, the resulting profile can be high-dimensional. The Support Vector Machines (SVM) were selected as classifier in this study. The per-pixel classification by SVM using both spectral data and structural information derived from extended morphological profiles was first conducted. The obtained per-pixel classification results were then combined with image segmentation results by an overlay operation for object based image classification. The proposed method was evaluated using QuickBird multispectral images over urban areas. The results, show that the proposed classification method significantly improves the image classification results, compared to per-pixel spectral classification.